DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
The information disclosure statement(s) (IDS) was/were submitted on 07/02/2024 and 05/28/2026. The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, each information disclosure statement is being considered by the examiner.
Specification
The disclosure is objected to because of the following informalities:
On paragraph [0003], “The machine-learned encoder model is trained to separate a plurality of embeddings of positive data instances belong to the target category from a plurality of embeddings of negative data instances” should read “The machine-learned encoder model is trained to separate a plurality of embeddings of positive data instances belonging to the target category from a plurality of embeddings of negative data instances.”
On paragraph [0129], “The provide machine learning model may be similar to the ones described under the section Machine Learning Models below” should read “The provided machine learning model may be similar to the ones described under the section Machine Learning Models below.”
On paragraph [0131], “The features maybe broken into groups” should read “The features may be broken into groups.”
On paragraph [0135], “Recall include situations where some transactions worth overriding are flagged” should read “Recall includes situations where some transactions worth overriding are flagged.”
On paragraph [0151], “The trained machine-learned model 600 can be used for make inference or another suitable task for which the model is trained” should read “The trained machine-learned model 600 can be used for making inference or another suitable task for which the model is trained.”
On paragraph [0152], “training may be performed using an unsupervised learning techniques” should read “training may be performed using an unsupervised learning technique.”
On paragraph [0154], “The computing server 110 may an embedding for a transaction” should read “The computing server 110 may include an embedding for a transaction.”
Appropriate correction is required.
Claim Objections
Claims 1, 7, 15, and 20 are objected to because of the following informalities:
In claim 1, “wherein the machine-learned encoder model is trained to separate a plurality of embeddings of positive data instances belong to the target category from a plurality of embeddings of negative data instances” should read “wherein the machine-learned encoder model is trained to separate a plurality of embeddings of positive data instances belonging to the target category from a plurality of embeddings of negative data instances.”
In claim 1, "training the machine-learned encoder model that generates embeddings of data instances” should read "training the machine-learned encoder model to generate embeddings of data instances.”
In claim 7, “each laying performing a particular operation on data instances” should read “each layer performing a particular operation on data instances.”
In claim 15, “wherein the machine-learned encoder model is trained to separate a plurality of embeddings of positive data instances belong to the target category from a plurality of embeddings of negative data instances” should read “wherein the machine-learned encoder model is trained to separate a plurality of embeddings of positive data instances belonging to the target category from a plurality of embeddings of negative data instances.”
In claim 15, "train the machine-learned encoder model that generates embeddings of data instances” should read "train the machine-learned encoder model to generate embeddings of data instances.”
In claim 20, “wherein the machine-learned encoder model is trained to separate a plurality of embeddings of positive data instances belong to the target category from a plurality of embeddings of negative data instances” should read “wherein the machine-learned encoder model is trained to separate a plurality of embeddings of positive data instances belonging to the target category from a plurality of embeddings of negative data instances.”
In claim 20, "train the machine-learned encoder model that generates embeddings of data instances” should read "train the machine-learned encoder model to generate embeddings of data instances.”
Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 6-9 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claims 6 and 8 recite the limitation "loss function that calculates a relationship between embeddings of anchor, positive and negative data instances.” There is insufficient antecedent basis for this limitation in the claims. It is unclear what “anchor” is referring to. For the purposes of examination, the Examiner has interpreted these instances and all subsequent instances in the dependent claims as a data instance similar to the positive data instances and dissimilar to the negative data instances (interpreted from spec [0089]).
Claim(s) 7-9 are rejected for at least the same reasons as claim 6 since they depend on claim 6.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claim(s) 1-20 are rejected under 35 U.S.C. 101 because the claims are directed towards an abstract idea without significantly more.
Regarding Claim 1:
Subject Matter Eligibility Analysis Step 1:
Claim 1 recites a method and is thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 1 recites:
to separate a plurality of embeddings of positive data instances belong to the target category from a plurality of embeddings of negative data instances (This limitation is a mental process as it encompasses a human mentally separating embeddings and is thus an evaluation.)
generating features of the target data instance to prepare the target data instance for further processing (This limitation is a mental process as it encompasses a human mentally generating features and is thus an evaluation.)
to determine an assignment of a category from the list of custom-defined categories (This limitation is a mental process as it encompasses a human mentally determining an assignment of a category and is thus an evaluation.)
Therefore, claim 1 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 1 further recites additional elements of:
A computer-implemented method, comprising (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
retrieving a list of custom-defined categories of a database maintained by a third-party platform, the list of custom-defined categories defined by an entity who uses the third-party data platform (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of receiving data (see MPEP 2106.05(g)).)
accessing a plurality of training samples for training a machine-learned encoder model, a training sample comprises a positive data instance belonging to a target category from the list of custom-defined categories and a negative data instance outside of the target category (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of receiving data (see MPEP 2106.05(g)).)
training the machine-learned encoder model that generates embeddings of data instances (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of training a model (see MPEP 2106.05(g)).)
wherein the machine-learned encoder model is trained (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of training a model (see MPEP 2106.05(g)).)
receiving a target data instance that is to be imported to the third-party data platform (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of receiving data (see MPEP 2106.05(g)).)
applying the machine-learned encoder model to the target data instance (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
and exporting the target data instance including the assignment of the category to the third-party data platform (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of sending data (see MPEP 2106.05(g)).)
Therefore, claim 1 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 1 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
A computer-implemented method, comprising uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
retrieving a list of custom-defined categories of a database maintained by a third-party platform, the list of custom-defined categories defined by an entity who uses the third-party data platform is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
accessing a plurality of training samples for training a machine-learned encoder model, a training sample comprises a positive data instance belonging to a target category from the list of custom-defined categories and a negative data instance outside of the target category is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
training the machine-learned encoder model that generates embeddings of data instances is the well understood, routine, and conventional activity of iteratively training a model (US 2021/0125108 A1, Metzler et al. page 11, paragraph 0052, “a classifier is trained using a conventional iterative machine learning training process that determines weights for each result list position”).
wherein the machine-learned encoder model is trained is the well understood, routine, and conventional activity of iteratively training a model (US 2021/0125108 A1, Metzler et al. page 11, paragraph 0052, “a classifier is trained using a conventional iterative machine learning training process that determines weights for each result list position”).
receiving a target data instance that is to be imported to the third-party data platform is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
applying the machine-learned encoder model to the target data instance uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
and exporting the target data instance including the assignment of the category to the third-party data platform is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
Therefore, claim 1 is subject-matter ineligible.
Regarding Claim 2:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 2 recites:
generating a plurality of positive embeddings corresponding to the positive data instances and a plurality of negative embeddings corresponding to the negative data instances (This limitation is a mental process as it encompasses a human mentally defining a list of categories and is thus an evaluation.)
determining a loss function that measure distances for a plurality of embedding pairs, each embedding pair comprising at least one of the positive embeddings and one of the negative embeddings, and the distance for each embedding pair measuring a distance between the one of the positive embeddings and one of the negative embeddings (This limitation is a mental process as it encompasses a human mentally defining a list of categories and is thus an evaluation.)
Therefore, claim 2 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 2 further recites additional elements of:
wherein training of the machine-learned encoder model comprises (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of training a model (see MPEP 2106.05(g)).)
receiving the training samples comprising positive data instances belonging to the target category and negative data instances outside of the target category (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of receiving data (see MPEP 2106.05(g)).)
backpropagating the loss function through the machine-learned encoder model (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
and adjusting one or more parameters of the machine-learned encoder model through the backpropagation (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, claim 2 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 2 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
wherein training of the machine-learned encoder model comprises is the well understood, routine, and conventional activity of iteratively training a model (US 2021/0125108 A1, Metzler et al. page 11, paragraph 0052, “a classifier is trained using a conventional iterative machine learning training process that determines weights for each result list position”).
receiving the training samples comprising positive data instances belonging to the target category and negative data instances outside of the target category is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
backpropagating the loss function through the machine-learned encoder model uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
and adjusting one or more parameters of the machine-learned encoder model through the backpropagation uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, claim 2 is subject-matter ineligible.
Regarding Claim 3:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 3 recites:
creating the list of custom-defined categories of data instances (This limitation is a mental process as it encompasses a human mentally creating a list of categories and is thus an evaluation.)
Therefore, claim 3 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 3 further recites additional elements of:
wherein retrieving the list of custom-defined categories of the database maintained by the third-party platform comprises (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of receiving data (see MPEP 2106.05(g)).)
and maintaining the list of the custom-defined categories of data instances in a database of the third-party platform (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, claim 3 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 3 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
wherein retrieving the list of custom-defined categories of the database maintained by the third-party platform comprises is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
creating the list of custom-defined categories of data instances uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
and maintaining the list of the custom-defined categories of data instances in a database of the third-party platform uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, claim 3 is subject-matter ineligible.
Regarding Claim 4:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 4 recites:
wherein creating the list of custom-defined categories of data instances comprises (This limitation is a mental process as it encompasses a human mentally creating a list and is thus an evaluation.)
defining custom-defined categories of data instances (This limitation is a mental process as it encompasses a human mentally defining categories and is thus an evaluation.)
wherein defining the custom-defined categories of data instances (This limitation is a mental process as it encompasses a human mentally defining categories and is thus an evaluation.)
selecting…one or more data parameters that determine the category of a data instance (This limitation is a mental process as it encompasses a human mentally selecting parameters and is thus an evaluation.)
and assigning…the one or more data parameters to the defined custom-defined categories of data instances (This limitation is a mental process as it encompasses a human mentally assigning parameters to categories and is thus an evaluation.)
Therefore, claim 4 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 4 further recites additional elements of:
providing a category creation tool on a user interface of the third-party data platform (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
on the category creation tool (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
on the category creation tool comprises (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
on the category creation tool (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
wherein the data parameters comprise any one of: an origin of the data instance; an amount associated with the data instance; a type associated with the data instance; a user associated with the data instance; and contextual information associated with the data instance (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
on the category creation tool (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, claim 4 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 4 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
providing a category creation tool on a user interface of the third-party data platform uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
on the category creation tool uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
on the category creation tool comprises uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
on the category creation tool uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
wherein the data parameters comprise any one of: an origin of the data instance; an amount associated with the data instance; a type associated with the data instance; a user associated with the data instance; and contextual information associated with the data instance uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
on the category creation tool uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, claim 4 is subject-matter ineligible.
Regarding Claim 5:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 5 recites the same abstract idea as claim 1. Therefore, claim 5 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 5 further recites additional elements of:
wherein accessing the plurality of training samples for training the machine-learned encoder model comprises (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of receiving data (see MPEP 2106.05(g)).)
retrieving the training samples from a database of a third-party platform (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of receiving (see MPEP 2106.05(g)).)
or storing the training samples in the database of the third-party platform (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of storing information in memory (see MPEP 2106.05(g)).)
Therefore, claim 5 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 5 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
wherein accessing the plurality of training samples for training the machine-learned encoder model comprises is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
retrieving the training samples from a database of a third-party platform is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
or storing the training samples in the database of the third-party platform is the well understood, routine, and conventional activity of " Storing and retrieving information in memory" (see MPEP 2106.05(d)(II); Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); OIP Techs., 788 F.3d at 1363, 115 USPQ2d at 1092-93.
Therefore, claim 5 is subject-matter ineligible.
Regarding Claim 6:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 6 recites:
defining a loss function that calculates a relationship between embeddings of anchor, positive and negative data instances (This limitation is a mental process as it encompasses a human mentally defining a loss function and is thus an evaluation.)
to generate embeddings using the loss function (This limitation is a mental process as it encompasses a human mentally generating embeddings and is thus an evaluation.)
and evaluating the training of the machine-learned encoder (This limitation is a mental process as it encompasses a human mentally evaluating training and is thus an evaluation.)
Therefore, claim 6 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 6 further recites additional elements of:
wherein training the machine-learned encoder model that generates embeddings of data instances comprises (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of training a model (see MPEP 2106.05(g)).)
initializing the machine-learned encoder model with predetermined parameters (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
training the machine-learned encoder by processing each training sample (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of training a model (see MPEP 2106.05(g)).)
Therefore, claim 6 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 6 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
wherein training the machine-learned encoder model that generates embeddings of data instances comprises is the well understood, routine, and conventional activity of iteratively training a model (US 2021/0125108 A1, Metzler et al. page 11, paragraph 0052, “a classifier is trained using a conventional iterative machine learning training process that determines weights for each result list position”).
initializing the machine-learned encoder model with predetermined parameters uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
training the machine-learned encoder by processing each training sample is the well understood, routine, and conventional activity of iteratively training a model (US 2021/0125108 A1, Metzler et al. page 11, paragraph 0052, “a classifier is trained using a conventional iterative machine learning training process that determines weights for each result list position”).
Therefore, claim 6 is subject-matter ineligible.
Regarding Claim 7:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 7 recites:
defining an architecture of embedded spaces where data instances are mapped (This limitation is a mental process as it encompasses a human mentally defining an architecture and is thus an evaluation.)
and defining data flow through the layers of the machine-learned encoder model from input to embedded output (This limitation is a mental process as it encompasses a human mentally defining data flow and is thus an evaluation.)
Therefore, claim 7 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 7 further recites additional elements of:
wherein initializing the machine-learned encoder model with the predetermined parameters comprises (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
wherein the architecture includes multiple layers (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
each laying performing a particular operation on data instances (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, claim 7 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 7 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
wherein initializing the machine-learned encoder model with the predetermined parameters comprises uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
wherein the architecture includes multiple layers uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
each laying performing a particular operation on data instances uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, claim 7 is subject-matter ineligible.
Regarding Claim 8:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 8 recites:
wherein defining the loss function that calculates a relationship between embeddings of anchor, positive and negative data instances comprises (This limitation is a mental process as it encompasses a human mentally defining a loss function and is thus an evaluation.)
defining a triplet loss function to minimize the relative distance between embeddings of positive data instances and maximize the relative distance between embeddings of negative data instances (This limitation is a mental process as it encompasses a human mentally defining a triplet loss function and is thus an evaluation.)
Therefore, claim 8 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 8 does not further recite any additional elements. Therefore, claim 8 is not
integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
Since there are no additional elements, claim 8 does not provide significantly more than
the abstract idea itself, taken alone and in combination. Therefore, claim 8 is subject-matter
ineligible.
Regarding Claim 9:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 9 recites:
wherein evaluating the training of the machine-learned encoder comprises (This limitation is a mental process as it encompasses a human mentally evaluating training and is thus an evaluation.)
determining a metric for the machine-learned encoder model (This limitation is a mental process as it encompasses a human mentally determining a metric and is thus an evaluation.)
and categorizing new data instances correctly through a comparison of model-predicted categories and actual categories (This limitation is a mental process as it encompasses a human mentally categorizing data instances and is thus an evaluation.)
Therefore, claim 9 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 9 further recites additional elements of:
applying a validation dataset to the machine-learned encoder model (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
the validation dataset comprising a plurality of data instances representing diverse categories from a list of custom categories in a database (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
the metric measuring a performance of the model in minimizing the calculated loss function (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, claim 9 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 9 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
applying a validation dataset to the machine-learned encoder model uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
the validation dataset comprising a plurality of data instances representing diverse categories from a list of custom categories in a database uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
the metric measuring a performance of the model in minimizing the calculated loss function uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, claim 9 is subject-matter ineligible.
Regarding Claim 10:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 10 recites:
and validating the target data instance (This limitation is a mental process as it encompasses a human mentally validating a data instance and is thus an evaluation.)
Therefore, claim 10 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 10 further recites additional elements of:
wherein receiving the target data instance that is to be imported to the third-party data platform comprises (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of receiving data (see MPEP 2106.05(g)).)
receiving the target data instance from a third party computer interface (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of receiving data (see MPEP 2106.05(g)).)
and importing the target data instance in the third-party data platform (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of transmitting data (see MPEP 2106.05(g)).)
Therefore, claim 10 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 10 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
wherein receiving the target data instance that is to be imported to the third-party data platform comprises is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
receiving the target data instance from a third party computer interface is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
and importing the target data instance in the third-party data platform is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
Therefore, claim 10 is subject-matter ineligible.
Regarding Claim 11:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 11 recites:
wherein generating the features of the target data instance to prepare the target data instance for further processing comprises (This limitation is a mental process as it encompasses a human mentally generating features and is thus an evaluation.)
identifying or extracting the features from the target data instance (This limitation is a mental process as it encompasses a human mentally identifying features and is thus an evaluation.)
Therefore, claim 11 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 11 further recites additional elements of:
wherein the features comprise any one of: an origin of the target data instance; an amount associated with the target data instance; a type associated with the target data instance; a user associated with the target data instance; and contextual information associated with the target data instance (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, claim 11 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 11 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
wherein the features comprise any one of: an origin of the target data instance; an amount associated with the target data instance; a type associated with the target data instance; a user associated with the target data instance; and contextual information associated with the target data instance uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, claim 11 is subject-matter ineligible.
Regarding Claim 12:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 12 recites:
to determine an assignment of a category from the list of custom-defined categories comprises (This limitation is a mental process as it encompasses a human mentally determining an assignment of a category and is thus an evaluation.)
generating…embeddings from the features of the target data instance (This limitation is a mental process as it encompasses a human mentally generating embeddings and is thus an evaluation.)
and comparing…the target data instance embeddings with embeddings that the model has learned for each data category during training (This limitation is a mental process as it encompasses a human mentally comparing embeddings and is thus an evaluation.)
assigning…a category to the target data instance based at least on the comparing (This limitation is a mental process as it encompasses a human mentally assigning a category and is thus an evaluation.)
Therefore, claim 12 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 12 further recites additional elements of:
wherein applying the machine-learned encoder model to the target data instance (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
accessing, by the machine-learned encoder model, features of the target data instance (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of retrieving information in memory (see MPEP 2106.05(g)).)
by the machine-learned encoder model (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
the embeddings being in multiple layers in a latent space of the machine-learned encoder model (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
by the machine-learned encoder model (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
by the machine-learned encoder model (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, claim 12 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 12 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
wherein applying the machine-learned encoder model to the target data instance uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
accessing, by the machine-learned encoder model, features of the target data instance is the well understood, routine, and conventional activity of " Storing and retrieving information in memory" (see MPEP 2106.05(d)(II); Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); OIP Techs., 788 F.3d at 1363, 115 USPQ2d at 1092-93.
by the machine-learned encoder model uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
the embeddings being in multiple layers in a latent space of the machine-learned encoder model uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
by the machine-learned encoder model uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
by the machine-learned encoder model uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, claim 12 is subject-matter ineligible.
Regarding Claim 13:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 13 recites:
wherein comparing…the target data instance embeddings with the embeddings that the model has learned for each data category during training comprises (This limitation is a mental process as it encompasses a human mentally comparing embedding as and is thus an evaluation.)
computing, for each data category, a distance between the embeddings of the target data instance and the embeddings that the model has learned for each data category during training (This limitation is a mental process as it encompasses a human mentally computing distances and is thus an evaluation.)
and comparing the computed distances, wherein the category with the shortest distance to the target data instance's embeddings is considered a match (This limitation is a mental process as it encompasses a human mentally comparing distances and is thus an evaluation.)
Therefore, claim 13 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 13 further recites additional elements of:
by the machine-learned encoder model (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, claim 13 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 13 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
by the machine-learned encoder model uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, claim 13 is subject-matter ineligible.
Regarding Claim 14:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 14 recites:
further comprising: generating a message to a user, wherein generating the message to the user comprises (This limitation is a mental process as it encompasses a human mentally generating a message and is thus an evaluation.)
applying a natural language generation process to the target data instance and its category assignment to provide one or more sentences that present them to the user (This limitation is a mental process as it encompasses a human mentally applying a process and is thus an evaluation.)
Therefore, claim 14 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 14 further recites additional elements of:
transmitting the message to the user through a communication channel comprising a short message service (SMS) message, email, or a software as a service (SaaS) platform (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of sending data (see MPEP 2106.05(g)).)
in response to transmitting the message to the user, receiving feedback from the user (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of receiving data (see MPEP 2106.05(g)).)
and updating the category assignment for the target data instance based on the feedback received from the user (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, claim 14 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 14 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
transmitting the message to the user through a communication channel comprising a short message service (SMS) message, email, or a software as a service (SaaS) platform is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
in response to transmitting the message to the user, receiving feedback from the user is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
and updating the category assignment for the target data instance based on the feedback received from the user uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, claim 14 is subject-matter ineligible.
Regarding Claim 15:
Subject Matter Eligibility Analysis Step 1:
Claim 15 recites a storage medium and is thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 15 recites:
generate features of the target data instance to prepare the target data instance for further processing (This limitation is a mental process as it encompasses a human mentally generating features and is thus an evaluation.)
to separate a plurality of embeddings of positive data instances belong to the target category from a plurality of embeddings of negative data instances (This limitation is a mental process as it encompasses a human mentally separate embeddings and is thus an evaluation.)
apply the machine-learned encoder model to the target data instance to determine an assignment of a category from the list of custom-defined categories (This limitation is a mental process as it encompasses a human mentally determine an assignment of a category and is thus an evaluation.)
Therefore, claim 15 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 15 further recites additional elements of:
A non-transitory computer-readable storage medium configured to store computer code comprising instructions, the instructions, when executed by one or more processors, cause the one or more processors to (This element does not integrate the abstract idea into a practical application because it recites generic computing components on which to perform the abstract idea (see MPEP 2106.05(f)).)
retrieve a list of custom-defined categories of a database maintained by a third-party platform, the list of custom-defined categories defined by an entity who uses the third-party data platform (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of receiving data (see MPEP 2106.05(g)).)
access a plurality of training samples for training a machine-learned encoder model, a training sample comprises a positive data instance belonging to a target category from the list of custom-defined categories and a negative data instance outside of the target category (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
train the machine-learned encoder model that generates embeddings of data instances (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of training a model (see MPEP 2106.05(g)).)
wherein the machine-learned encoder model is trained (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of training a model (see MPEP 2106.05(g)).)
receive a target data instance that is to be imported to the third-party data platform (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
and export the target data instance including the assignment of the category to the third-party data platform (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of transmitting data (see MPEP 2106.05(g)).)
Therefore, claim 15 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 15 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
A non-transitory computer-readable storage medium configured to store computer code comprising instructions, the instructions, when executed by one or more processors, cause the one or more processors to uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
retrieve a list of custom-defined categories of a database maintained by a third-party platform, the list of custom-defined categories defined by an entity who uses the third-party data platform is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
access a plurality of training samples for training a machine-learned encoder model, a training sample comprises a positive data instance belonging to a target category from the list of custom-defined categories and a negative data instance outside of the target category uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
train the machine-learned encoder model that generates embeddings of data instances is the well understood, routine, and conventional activity of iteratively training a model (US 2021/0125108 A1, Metzler et al. page 11, paragraph 0052, “a classifier is trained using a conventional iterative machine learning training process that determines weights for each result list position”).
wherein the machine-learned encoder model is trained is the well understood, routine, and conventional activity of iteratively training a model (US 2021/0125108 A1, Metzler et al. page 11, paragraph 0052, “a classifier is trained using a conventional iterative machine learning training process that determines weights for each result list position”).
receive a target data instance that is to be imported to the third-party data platform uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
and export the target data instance including the assignment of the category to the third-party data platform is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
Therefore, claim 15 is subject-matter ineligible.
Regarding Claim 16:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 16 recites:
generating a plurality of positive embeddings corresponding to the positive data instances and a plurality of negative embeddings corresponding to the negative data instances (This limitation is a mental process as it encompasses a human mentally generating embeddings and is thus an evaluation.)
determining a loss function that measure distances for a plurality of embedding pairs, each embedding pair comprising at least one of the positive embeddings and one of the negative embeddings, and the distance for each embedding pair measuring a distance between the one of the positive embeddings and one of the negative embeddings (This limitation is a mental process as it encompasses a human mentally determining a loss function and is thus an evaluation.)
Therefore, claim 16 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 16 further recites additional elements of:
wherein training of the machine-learned encoder model comprises (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of training a model (see MPEP 2106.05(g)).)
receiving the training samples comprising positive data instances belonging to the target category and negative data instances outside of the target category (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of receiving data (see MPEP 2106.05(g)).)
backpropagating the loss function through the machine-learned encoder model (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
and adjusting one or more parameters of the machine-learned encoder model through the backpropagation (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, claim 16 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 16 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
wherein training of the machine-learned encoder model comprises is the well understood, routine, and conventional activity of iteratively training a model (US 2021/0125108 A1, Metzler et al. page 11, paragraph 0052, “a classifier is trained using a conventional iterative machine learning training process that determines weights for each result list position”).
receiving the training samples comprising positive data instances belonging to the target category and negative data instances outside of the target category is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
backpropagating the loss function through the machine-learned encoder model uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
and adjusting one or more parameters of the machine-learned encoder model through the backpropagation uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, claim 16 is subject-matter ineligible.
Regarding Claim 17:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 17 recites:
creating the list of custom-defined categories of data instances (This limitation is a mental process as it encompasses a human mentally creating a list of custom-defined categories and is thus an evaluation.)
Therefore, claim 17 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 17 further recites additional elements of:
wherein retrieving the list of custom-defined categories of the database maintained by the third-party platform comprises (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of receiving data (see MPEP 2106.05(g)).)
and maintaining the list of the custom-defined categories of data instances in a database of the third-party platform (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, claim 17 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 17 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
wherein retrieving the list of custom-defined categories of the database maintained by the third-party platform comprises is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
creating the list of custom-defined categories of data instances uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
and maintaining the list of the custom-defined categories of data instances in a database of the third-party platform uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, claim 17 is subject-matter ineligible.
Regarding Claim 18:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 18 recites:
defining custom-defined categories of data instances (This limitation is a mental process as it encompasses a human mentally defining categories and is thus an evaluation.)
wherein defining the custom-defined categories of data instances (This limitation is a mental process as it encompasses a human mentally defining categories and is thus an evaluation.)
selecting…one or more data parameters that determine the category of a data instance (This limitation is a mental process as it encompasses a human mentally selecting parameters and is thus an evaluation.)
and assigning…the one or more data parameters to the defined custom-defined categories of data instances (This limitation is a mental process as it encompasses a human mentally assigning parameters to categories and is thus an evaluation.)
Therefore, claim 18 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 18 further recites additional elements of:
wherein creating the list of custom-defined categories of data instances comprises (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
providing a category creation tool on a user interface of the third-party data platform (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
on the category creation tool (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
on the category creation tool comprises (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
on the category creation tool (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
wherein the data parameters comprise any one of: an origin of the data instance; an amount associated with the data instance; a type associated with the data instance; a user associated with the data instance; and contextual information associated with the data instance (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
on the category creation tool (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, claim 18 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 18 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
wherein creating the list of custom-defined categories of data instances comprises uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
providing a category creation tool on a user interface of the third-party data platform uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
on the category creation tool uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
on the category creation tool comprises uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
on the category creation tool uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
wherein the data parameters comprise any one of: an origin of the data instance; an amount associated with the data instance; a type associated with the data instance; a user associated with the data instance; and contextual information associated with the data instance uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
on the category creation tool uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, claim 18 is subject-matter ineligible.
Regarding Claim 19:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 19 recites the same abstract idea as claim 15. Therefore, claim 19 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 19 further recites additional elements of:
wherein accessing the plurality of training samples for training the machine-learned encoder model comprises (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of transmitting data (see MPEP 2106.05(g)).)
retrieving the training samples from a database of a third-party platform (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of receiving data (see MPEP 2106.05(g)).)
or storing the training samples in the database of the third-party platform (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of storing information in memory (see MPEP 2106.05(g)).)
Therefore, claim 19 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 19 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
wherein accessing the plurality of training samples for training the machine-learned encoder model comprises is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
retrieving the training samples from a database of a third-party platform is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
or storing the training samples in the database of the third-party platform is the well understood, routine, and conventional activity of " Storing and retrieving information in memory" (see MPEP 2106.05(d)(II); Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); OIP Techs., 788 F.3d at 1363, 115 USPQ2d at 1092-93.
Therefore, claim 19 is subject-matter ineligible.
Regarding Claim 20:
Subject Matter Eligibility Analysis Step 1:
Claim 20 recites a system and is thus a machine, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 20 recites:
generate features of the target data instance to prepare the target data instance for further processing (This limitation is a mental process as it encompasses a human mentally generating features and is thus an evaluation.)
to separate a plurality of embeddings of positive data instances belong to the target category from a plurality of embeddings of negative data instances (This limitation is a mental process as it encompasses a human mentally separate embeddings and is thus an evaluation.)
to determine an assignment of a category from the list of custom-defined categories (This limitation is a mental process as it encompasses a human mentally determine an assignment of a category and is thus an evaluation.)
Therefore, claim 20 recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 20 further recites additional elements of:
A system, comprising: one or more processors; and memory configured to store instructions, the instructions, when executed by the one or more processors, cause the one or more processors to (This element does not integrate the abstract idea into a practical application because it recites generic computing components on which to perform the abstract idea (see MPEP 2106.05(f)).)
retrieve a list of custom-defined categories of a database maintained by a third-party platform, the list of custom-defined categories defined by an entity who uses the third-party data platform (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of receiving data (see MPEP 2106.05(g)).)
access a plurality of training samples for training a machine-learned encoder model, a training sample comprises a positive data instance belonging to a target category from the list of custom-defined categories and a negative data instance outside of the target category (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of retrieving information in memory (see MPEP 2106.05(g)).)
train the machine-learned encoder model that generates embeddings of data instances (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of training a model (see MPEP 2106.05(g)).)
wherein the machine-learned encoder model is trained (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of training a model (see MPEP 2106.05(g)).)
receive a target data instance that is to be imported to the third-party data platform (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
apply the machine-learned encoder model to the target data instance (This element does not integrate the abstract idea into a practical application because it amounts to mere “apply it on a computer” (see MPEP 2106.05(f)).)
and export the target data instance including the assignment of the category to the third-party data platform (This element does not integrate the abstract idea into a practical application because it recites the insignificant extra-solution activity of transmitting data (see MPEP 2106.05(g)).)
Therefore, claim 20 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional elements of claim 20 do not provide significantly more than the abstract
idea itself, taken alone and in combination because
A system, comprising: one or more processors; and memory configured to store instructions, the instructions, when executed by the one or more processors, cause the one or more processors to uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
retrieve a list of custom-defined categories of a database maintained by a third-party platform, the list of custom-defined categories defined by an entity who uses the third-party data platform is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
access a plurality of training samples for training a machine-learned encoder model, a training sample comprises a positive data instance belonging to a target category from the list of custom-defined categories and a negative data instance outside of the target category is the well understood, routine, and conventional activity of " Storing and retrieving information in memory" (see MPEP 2106.05(d)(II); Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); OIP Techs., 788 F.3d at 1363, 115 USPQ2d at 1092-93.
train the machine-learned encoder model that generates embeddings of data instances is the well understood, routine, and conventional activity of iteratively training a model (US 2021/0125108 A1, Metzler et al. page 11, paragraph 0052, “a classifier is trained using a conventional iterative machine learning training process that determines weights for each result list position”).
wherein the machine-learned encoder model is trained is the well understood, routine, and conventional activity of iteratively training a model (US 2021/0125108 A1, Metzler et al. page 11, paragraph 0052, “a classifier is trained using a conventional iterative machine learning training process that determines weights for each result list position”).
receive a target data instance that is to be imported to the third-party data platform uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
apply the machine-learned encoder model to the target data instance uses a computer as a tool to perform the abstract idea and cannot provide significantly more (see MPEP 2106.05(f)).
and export the target data instance including the assignment of the category to the third-party data platform is the well understood, routine, and conventional activity of "transmitting or receiving data over a network" (see MPEP 2106.05(d)(II); OIP Techs., Inc., V. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
Therefore, claim 20 is subject-matter ineligible.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claim(s) 1, 3, 5, 7, 10, 11, 13, 15, 17, 19, and 20 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Martins (US20240013220).
Regarding claim 1, Martins teaches
A computer-implemented method, comprising: retrieving a list of custom-defined categories of a database maintained by a third-party platform (Martins, paragraph 0067, “At 702, method 400 may include querying or retrieving a database of entities with their associated fraudulent labels. For example, method 700 carried out by application server 120 may query a labeled data set that has been processed and stored in database 118. The data set may consist of multiple populations of entities, for example, fraudulent, high-risk, and not fraudulent merchants.” Martins, paragraph 0028, “According to some aspects of this disclosure, the computing device 126 may include a server, cloud-based device, third-party system, and/or the like that performs one or more functions associated with the user device 102 (and/or a user of the user device 102, etc.), organization system 108, and/or the like. Examiner notes the custom-defined categories of the database are the labels of either fraudulent or non-fraudulent for each entity.)
the list of custom-defined categories defined by an entity who uses the third-party data platform (Martins, paragraph 0068, “According to some aspects, at 704, method 700 may include identifying a first group of entities (e.g., merchants) that have a high fraud risk level. This may entail having application server 120 search and detect the respective label of each entity and identify the entities that are labeled as high-risk or risky entities.” Examiner notes the entity is the application server 120, which defines and labels whether the entities are fraudulent or not.)
accessing a plurality of training samples for training a machine-learned encoder model (Martins, paragraph 0008, “FIG. 4 illustrates an example process flow for a predictive model trained for embedding analysis for entity classification, according to some aspects of this disclosure.” Martins, paragraph 0056, “FIG. 4 illustrates an input and output representation for leveraging embeddings (e.g., embedding analysis for entity classification), according to some aspects of this disclosure. According to some aspects, in reference to natural language processing (NLP), a word embedding is a term used for the representation of words for text analysis, typically in the form of a real-valued vector that encodes the meaning of the word such that the words that are closer in the vector space are expected to be similar in meaning. FIG. 4 illustrates methodologies whereby positive and negative samples of merchant associates are used as input to a neural network to generate an output vector graph.” Martins, paragraph 0062, “According to some aspects of this disclosure, the ML model 122 may be trained in a dataset that includes labels for fraudulent merchants, potentially risky (e.g. high risk) merchants, non-fraudulent merchants, and/or any other type of classification based on aggregations of historical data, such as historical transactions and fraud reported thereupon.” Examiner notes the plurality of training samples are the positive and negative samples of merchant associates.)
a training sample comprises a positive data instance belonging to a target category from the list of custom-defined categories and a negative data instance outside of the target category (Martins, paragraph 0056, “FIG. 4 illustrates methodologies whereby positive and negative samples of merchant associates are used as input to a neural network to generate an output vector graph. A positive sample may be one where merchants are deemed sufficiently proximate in a transaction space, thereby requiring further analysis. A negative sample may be one where merchants are deemed insufficiently proximate in a transaction space, thereby not requiring further analysis. For example, an association between Merchant 1 and Merchant 2 may be detected. If a consumer transacts at Merchant 1 and within a short period of time transacts at Merchant 2, it can be determined that Merchant 1 and Merchant 2 have a positive association (e.g., deemed sufficiently proximate for further analysis).”)
training the machine-learned encoder model that generates embeddings of data instances (Martins, paragraph 0059, “According to some aspects, the model may be trained on a link prediction task. That is, given a merchant pair (e.g., Merchant 1 and Merchant 2), the model may be trained to predict whether this pair has at least one account in common. In this regard, the model produces embeddings that encode accounts in common between merchants—thereby allowing organization system 108 to link merchants to detect fraudulent patterns.”)
wherein the machine-learned encoder model is trained to separate a plurality of embeddings of positive data instances belong to the target category from a plurality of embeddings of negative data instances (Martins, paragraph 0058, “One example of training is the SVM, where features having the smallest weights are removed and the algorithm is re-trained with the remaining weights, wherein the process is repeated until features remain that are able to accurately separate the data into different patterns or classes…Such trends may correspond to increasing similarities between close merchants (e.g., Merchant 1 and Merchant 2) and decreasing similarities between distant merchants (e.g., Merchant 1 and Merchant 3). A vector graph may be generated as an output representing a mapping of the merchants whereby Merchant 1 and Merchant 2 are closely related (due to similarities in transactions, users, etc.), and Merchant 1 and Merchant 3 are distantly related.” Examiner notes the positive data instances are the close merchants and negative instances are the distant merchants.)
receiving a target data instance that is to be imported to the third-party data platform (Martins, FIG. 4, 702. Martins, FIG. 1. Martins, paragraph 0047, “According to some aspects of this disclosure, application server 120 may comprise one or more computer systems configured to compile data from a plurality of sources, such as location server 110, communication server 112, and transaction server 114, correlate and analyze the compiled data in real-time (continuously), arrange the compiled data, generate derived data based on the compiled data, and store the compiled and derived in a database such as database 118. According to some aspects of this disclosure, application server 120 may perform merchant detection and prediction operations, for example, via a trained machine learning (ML) model 122 included with the application server 120. Example merchant detection and prediction operations performed by application server 120 (e.g., the ML model 122, etc.) are further described herein with reference to FIGS. 2-6.” Examiner notes the target data instance is the retrieved labeled entity in step 702, which may be stored (imported) into a database, as described in paragraph 0047 and FIG. 1.)
generating features of the target data instance to prepare the target data instance for further processing (Martins, FIG. 4, 708. Martins, paragraph 0070, “At 708, method 700 may include joining (or assigning) entities to their associated embeddings. Martins, paragraph 0064, “According to some aspects of this disclosure, information such as embeddings describing fraudulent merchants may be used to seed a nearest neighbor search.” Examiner notes the features of the target data instance are the embeddings generated by the model for the entities, which is used for classifying the entities.)
applying the machine-learned encoder model to the target data instance to determine an assignment of a category from the list of custom-defined categories (Martins, FIG. 4, step 712.)
and exporting the target data instance including the assignment of the category to the third-party data platform (Martins, FIG. 4, step 712. Martins, FIG. 1. Martins, paragraph 0047, “According to some aspects of this disclosure, application server 120 may comprise one or more computer systems configured to compile data from a plurality of sources, such as location server 110, communication server 112, and transaction server 114, correlate and analyze the compiled data in real-time (continuously), arrange the compiled data, generate derived data based on the compiled data, and store the compiled and derived in a database such as database 118. According to some aspects of this disclosure, application server 120 may perform merchant detection and prediction operations, for example, via a trained machine learning (ML) model 122 included with the application server 120. Example merchant detection and prediction operations performed by application server 120 (e.g., the ML model 122, etc.) are further described herein with reference to FIGS. 2-6.” Examiner notes the classified (labeled) target data instance is the entity in step 712, which is assigned a category of either fraudulent or non-fraudulent. Examiner further notes the entity than has been assigned a category may be stored (imported) into a database, as described in paragraph 0047 and FIG. 1.)
Regarding claim 3, Martins teaches
The computer-implemented method of claim 1, wherein retrieving the list of custom-defined categories of the database maintained by the third-party platform comprises: creating the list of custom-defined categories of data instances (Martins, paragraph 0037, “According to some aspects of this disclosure, web server 110 may generate and/or store a list of nefarious actors, for example, such as identified malicious and/or fraudulent merchant websites. Examiner notes that the data instances are the actors and the custom-defined categories are whether the actors are fraudulent or not.)
and maintaining the list of the custom-defined categories of data instances in a database of the third-party platform (Martins, paragraph 0037, “According to some aspects of this disclosure, web server 110 may generate and/or store a list of nefarious actors, for example, such as identified malicious and/or fraudulent merchant websites. According to some aspects of this disclosure, web server 110 may enable application server 120, for example, to access the detected list and utilize it as an initial training data set. According to some aspects of this disclosure, detected transactions that are deemed to be in close proximity to the detected fraudulent merchants may then be flagged as potential fraudulent merchants. It can be appreciated that the list of detected merchants may include merchants previously flagged as being fraudulent.” Martins, paragraph 0036, “According to some aspects of this disclosure, the web server 110 may include one or more processors 132 and one or more databases 134 (e.g., any suitable repository of website data, etc.).” Examiner notes that the list of custom-defined categories is the list of nefarious actors which is stored in databases 134 by the web server 110.)
Regarding claim 5, Martins teaches
The computer-implemented method of claim 1, wherein accessing the plurality of training samples for training the machine-learned encoder model comprises: retrieving the training samples from a database of a third-party platform; or storing the training samples in the database of the third-party platform (Martins, paragraph 0037, “According to some aspects of this disclosure, web server 110 may generate and/or store a list of nefarious actors, for example, such as identified malicious and/or fraudulent merchant websites. According to some aspects of this disclosure, web server 110 may enable application server 120, for example, to access the detected list and utilize it as an initial training data set.” Martins, paragraph 0036, “According to some aspects of this disclosure, the web server 110 may include one or more processors 132 and one or more databases 134 (e.g., any suitable repository of website data, etc.).” Examiner notes the training samples are the detected list, which can be used as a training data set. Examiner further notes the list is accessed (retrieving) by the application server and stored by the web server using database 134.)
Regarding claim 7, Martins teaches
The computer-implemented method of claim 6, wherein initializing the machine-learned encoder model with the predetermined parameters comprises: defining an architecture of embedded spaces where data instances are mapped (Martins, paragraph 0056, “FIG. 4 illustrates an input and output representation for leveraging embeddings (e.g., embedding analysis for entity classification), according to some aspects of this disclosure. According to some aspects, in reference to natural language processing (NLP), a word embedding is a term used for the representation of words for text analysis, typically in the form of a real-valued vector that encodes the meaning of the word such that the words that are closer in the vector space are expected to be similar in meaning. FIG. 4 illustrates methodologies whereby positive and negative samples of merchant associates are used as input to a neural network to generate an output vector graph. A positive sample may be one where merchants are deemed sufficiently proximate in a transaction space, thereby requiring further analysis. A negative sample may be one where merchants are deemed insufficiently proximate in a transaction space, thereby not requiring further analysis.” Examiner notes the embedded spaces are the transaction spaces and the data instances are the samples of merchant associates.)
wherein the architecture includes multiple layers (Martins, paragraph 0057, “a multi-layer neural network with back-propagation, or other algorithms may be used with several associating factors to adjust weights and learn target data within the tables.”)
each laying performing a particular operation on data instances (Martins, paragraph 0058, “In forward propagation, the input signals may be multiplied by the weights in the weight matrices for each layer, and activation functions may transform the output at each layer, wherein the end output may be calculated.”)
and defining data flow through the layers of the machine-learned encoder model from input to embedded output (Martins, paragraph 0058, “Alternately, a neural network type algorithm is used, such as a back-propagation neural network, where there may be a weight matrix for each layer of the neural network, wherein for each layer, a bias vector is defined. The model may first undergo forward propagation. In forward propagation, the input signals may be multiplied by the weights in the weight matrices for each layer, and activation functions may transform the output at each layer, wherein the end output may be calculated. Back propagation aids in computing the error of partial derivatives, which can then be minimized across layers and can form the central mechanism by which the neural network learns. This may aid in discovering trends for classification wherein resources of a particular input may be more likely to be used. Such trends may correspond to increasing similarities between close merchants (e.g., Merchant 1 and Merchant 2) and decreasing similarities between distant merchants (e.g., Merchant 1 and Merchant 3). A vector graph may be generated as an output representing a mapping of the merchants whereby Merchant 1 and Merchant 2 are closely related (due to similarities in transactions, users, etc.), and Merchant 1 and Merchant 3 are distantly related.” Examiner notes the data flow is defined through the layers by forward propagation, where the input embedded into a vector graph output.)
Regarding claim 10, Martins teaches
The computer-implemented method of claim 1, wherein receiving the target data instance that is to be imported to the third-party data platform comprises: receiving the target data instance from a third party computer interface (Martins, FIG. 1. Examiner notes the target data instance is the Merchant Device (124) information, and this is received by the Network (106).)
and validating the target data instance (Martins, FIG. 1. Martins, paragraph 0031, “According to some aspects of this disclosure, the organization system 108 may facilitate, validate, secure, support, and/or enable transactions between the user device 102 and the merchant device 124.” Examiner notes Merchant Device (124) must go through Network (106) and be validated as non-fraudulent before being sent to the User Device (102).)
and importing the target data instance in the third-party data platform (Martins, FIG. 1. Martins, paragraph 0047, “According to some aspects of this disclosure, application server 120 may comprise one or more computer systems configured to compile data from a plurality of sources, such as location server 110, communication server 112, and transaction server 114, correlate and analyze the compiled data in real-time (continuously), arrange the compiled data, generate derived data based on the compiled data, and store the compiled and derived in a database such as database 118. Examiner notes the target data instance is sent from the Merchant Device to the Network, to any of the servers (110, 112, 114), before being analyzed and stored in a database (118).)
Regarding claim 11, Martins teaches
The computer-implemented method of claim 1, wherein generating the features of the target data instance to prepare the target data instance for further processing comprises: identifying or extracting the features from the target data instance; wherein the features comprise any one of: an origin of the target data instance; an amount associated with the target data instance; a type associated with the target data instance; a user associated with the target data instance; and contextual information associated with the target data instance (Martins, paragraph 0017, “The systems, methods, and computer program products described herein enable and/or facilitate embedding analysis for entity classification. Entity (e.g., device, system, user, etc.) interactions (e.g., transactions, communications, engagements, exchanges of information, etc.) may be modeled as a graph (e.g., a heterogeneous graph, of interaction participants where entities are connected by edges the represent the interactions.” Examiner notes the target data are the entities and features of the target data are the embeddings of the entity interactions, which includes contextual information associated with the target data instance (entity).)
Regarding claim 13, Martins teaches
The computer-implemented method of claim 12, wherein comparing, by the machine-learned encoder model, the target data instance embeddings with the embeddings that the model has learned for each data category during training comprises: computing, for each data category, a distance between the embeddings of the target data instance and the embeddings that the model has learned for each data category during training (Martins, FIG. 4, step 710. Martins, paragraph 0071, “At 710, method 700 carried out by application server 120 may perform a nearest neighbor search of fraudulent and high-risk entity embeddings. More specifically, taking a fraudulent merchant and a high-risk merchant into consideration, application server 120 may generate a similarity score between the fraudulent merchant and high-risk merchant. Similarity can be defined by, for example, a cosine similarity or other distance measures.” Examiner notes the embedding of the target data instance is the embedding of the labeled entity (the potential high risk entity), the category is either fraudulent or non-fraudulent merchant, and the embeddings the model learns for the data category are the embeddings of the fraudulent entities.)
and comparing the computed distances, wherein the category with the shortest distance to the target data instance's embeddings is considered a match (Martins, FIG. 4, step 712. Martins, paragraph 0072, “According to some aspects, at 712, the nearest neighbors of the fraudulent merchant may be deemed also fraudulent. The threshold for similarity that warrants a fraudulent label can be determined by business needs, taking into account the desire to prevent fraud as well as the tolerance for false identification. For example, application server 120 may perform a nearest neighbor search of the high-risk merchants to select those most proximal for further monitoring.” Examiner notes nearest neighbor search identifies the nearest point(s) given a target, and the target data instances are the potential high-risk merchants and the category is the fraudulent merchant.)
Regarding claim 15, Martins teaches
A non-transitory computer-readable storage medium configured to store computer code comprising instructions, the instructions, when executed by one or more processors, cause the one or more processors to (Martins, paragraph 0115, “In some aspects of this disclosure, a tangible, non-transitory apparatus or article of manufacture comprising a tangible, non-transitory computer usable or readable medium having control logic (software) stored thereon may also be referred to herein as a computer program product or program storage device. This includes, but is not limited to, computer system 1100, main memory 1108, secondary memory 1110, and removable storage units 1118 and 1122, as well as tangible articles of manufacture embodying any combination of the foregoing. Such control logic, when executed by one or more data processing devices (such as computer system 1100), may cause such data processing devices to operate as described herein.”)
retrieve a list of custom-defined categories of a database maintained by a third-party platform (Martins, paragraph 0067, “At 702, method 400 may include querying or retrieving a database of entities with their associated fraudulent labels. For example, method 700 carried out by application server 120 may query a labeled data set that has been processed and stored in database 118. The data set may consist of multiple populations of entities, for example, fraudulent, high-risk, and not fraudulent merchants.” Examiner notes the custom-defined categories of the database are the labels of either fraudulent or non-fraudulent for each entity.)
the list of custom-defined categories defined by an entity who uses the third-party data platform (Martins, paragraph 0068, “According to some aspects, at 704, method 700 may include identifying a first group of entities (e.g., merchants) that have a high fraud risk level. This may entail having application server 120 search and detect the respective label of each entity and identify the entities that are labeled as high-risk or risky entities.” Examiner notes the entity is the application server 120, which defines and labels whether the entities are fraudulent or not.)
access a plurality of training samples for training a machine-learned encoder model (Martins, paragraph 0008, “FIG. 4 illustrates an example process flow for a predictive model trained for embedding analysis for entity classification, according to some aspects of this disclosure.” Martins, paragraph 0056, “FIG. 4 illustrates an input and output representation for leveraging embeddings (e.g., embedding analysis for entity classification), according to some aspects of this disclosure. According to some aspects, in reference to natural language processing (NLP), a word embedding is a term used for the representation of words for text analysis, typically in the form of a real-valued vector that encodes the meaning of the word such that the words that are closer in the vector space are expected to be similar in meaning. FIG. 4 illustrates methodologies whereby positive and negative samples of merchant associates are used as input to a neural network to generate an output vector graph.” Martins, paragraph 0062, “According to some aspects of this disclosure, the ML model 122 may be trained in a dataset that includes labels for fraudulent merchants, potentially risky (e.g. high risk) merchants, non-fraudulent merchants, and/or any other type of classification based on aggregations of historical data, such as historical transactions and fraud reported thereupon.” Examiner notes the plurality of training samples are the positive and negative samples of merchant associates.)
a training sample comprises a positive data instance belonging to a target category from the list of custom-defined categories and a negative data instance outside of the target category (Martins, paragraph 0056, “FIG. 4 illustrates methodologies whereby positive and negative samples of merchant associates are used as input to a neural network to generate an output vector graph. A positive sample may be one where merchants are deemed sufficiently proximate in a transaction space, thereby requiring further analysis. A negative sample may be one where merchants are deemed insufficiently proximate in a transaction space, thereby not requiring further analysis. For example, an association between Merchant 1 and Merchant 2 may be detected. If a consumer transacts at Merchant 1 and within a short period of time transacts at Merchant 2, it can be determined that Merchant 1 and Merchant 2 have a positive association (e.g., deemed sufficiently proximate for further analysis).”)
train the machine-learned encoder model that generates embeddings of data instances (Martins, paragraph 0059, “According to some aspects, the model may be trained on a link prediction task. That is, given a merchant pair (e.g., Merchant 1 and Merchant 2), the model may be trained to predict whether this pair has at least one account in common. In this regard, the model produces embeddings that encode accounts in common between merchants—thereby allowing organization system 108 to link merchants to detect fraudulent patterns.”)
wherein the machine-learned encoder model is trained to separate a plurality of embeddings of positive data instances belong to the target category from a plurality of embeddings of negative data instances (Martins, paragraph 0058, “One example of training is the SVM, where features having the smallest weights are removed and the algorithm is re-trained with the remaining weights, wherein the process is repeated until features remain that are able to accurately separate the data into different patterns or classes…Such trends may correspond to increasing similarities between close merchants (e.g., Merchant 1 and Merchant 2) and decreasing similarities between distant merchants (e.g., Merchant 1 and Merchant 3). A vector graph may be generated as an output representing a mapping of the merchants whereby Merchant 1 and Merchant 2 are closely related (due to similarities in transactions, users, etc.), and Merchant 1 and Merchant 3 are distantly related.” Examiner notes the positive data instances are the close merchants and negative instances are the distant merchants.)
receive a target data instance that is to be imported to the third-party data platform (Martins, FIG. 4, 702. Martins, FIG. 1. Martins, paragraph 0047, “According to some aspects of this disclosure, application server 120 may comprise one or more computer systems configured to compile data from a plurality of sources, such as location server 110, communication server 112, and transaction server 114, correlate and analyze the compiled data in real-time (continuously), arrange the compiled data, generate derived data based on the compiled data, and store the compiled and derived in a database such as database 118. According to some aspects of this disclosure, application server 120 may perform merchant detection and prediction operations, for example, via a trained machine learning (ML) model 122 included with the application server 120. Example merchant detection and prediction operations performed by application server 120 (e.g., the ML model 122, etc.) are further described herein with reference to FIGS. 2-6.” Examiner notes the target data instance is the retrieved labeled entity in step 702, which may be stored (imported) into a database, as described in paragraph 0047 and FIG. 1.)
generate features of the target data instance to prepare the target data instance for further processing (Martins, FIG. 4, 708. Martins, paragraph 0070, “At 708, method 700 may include joining (or assigning) entities to their associated embeddings. Martins, paragraph 0064, “According to some aspects of this disclosure, information such as embeddings describing fraudulent merchants may be used to seed a nearest neighbor search.” Examiner notes the features of the target data instance are the embeddings generated by the model for the entities, which is used for classifying the entities.)
apply the machine-learned encoder model to the target data instance to determine an assignment of a category from the list of custom-defined categories (Martins, FIG. 4, step 712.)
and export the target data instance including the assignment of the category to the third-party data platform (Martins, FIG. 4, step 712. Martins, FIG. 1. Martins, paragraph 0047, “According to some aspects of this disclosure, application server 120 may comprise one or more computer systems configured to compile data from a plurality of sources, such as location server 110, communication server 112, and transaction server 114, correlate and analyze the compiled data in real-time (continuously), arrange the compiled data, generate derived data based on the compiled data, and store the compiled and derived in a database such as database 118. According to some aspects of this disclosure, application server 120 may perform merchant detection and prediction operations, for example, via a trained machine learning (ML) model 122 included with the application server 120. Example merchant detection and prediction operations performed by application server 120 (e.g., the ML model 122, etc.) are further described herein with reference to FIGS. 2-6.” Examiner notes the classified (labeled) target data instance is the entity in step 712, which is assigned a category of either fraudulent or non-fraudulent. Examiner further notes the entity than has been assigned a category may be stored (imported) into a database, as described in paragraph 0047 and FIG. 1.)
Regarding claim 17, Martins teaches
The non-transitory computer-readable storage medium of claim 15, wherein retrieving the list of custom-defined categories of the database maintained by the third-party platform comprises: creating the list of custom-defined categories of data instances (Martins, paragraph 0037, “According to some aspects of this disclosure, web server 110 may generate and/or store a list of nefarious actors, for example, such as identified malicious and/or fraudulent merchant websites. Examiner notes that the data instances are the actors and the custom-defined categories are whether the actors are fraudulent or not.)
and maintaining the list of the custom-defined categories of data instances in a database of the third-party platform (Martins, paragraph 0037, “According to some aspects of this disclosure, web server 110 may generate and/or store a list of nefarious actors, for example, such as identified malicious and/or fraudulent merchant websites. According to some aspects of this disclosure, web server 110 may enable application server 120, for example, to access the detected list and utilize it as an initial training data set. According to some aspects of this disclosure, detected transactions that are deemed to be in close proximity to the detected fraudulent merchants may then be flagged as potential fraudulent merchants. It can be appreciated that the list of detected merchants may include merchants previously flagged as being fraudulent.” Martins, paragraph 0036, “According to some aspects of this disclosure, the web server 110 may include one or more processors 132 and one or more databases 134 (e.g., any suitable repository of website data, etc.).” Examiner notes that the list of custom-defined categories is the list of nefarious actors which is stored in databases 134 by the web server 110.)
Regarding claim 19, Martins teaches
The non-transitory computer-readable storage medium of claim 15, wherein accessing the plurality of training samples for training the machine-learned encoder model comprises: retrieving the training samples from a database of a third-party platform; or storing the training samples in the database of the third-party platform (Martins, paragraph 0037, “According to some aspects of this disclosure, web server 110 may generate and/or store a list of nefarious actors, for example, such as identified malicious and/or fraudulent merchant websites. According to some aspects of this disclosure, web server 110 may enable application server 120, for example, to access the detected list and utilize it as an initial training data set.” Martins, paragraph 0036, “According to some aspects of this disclosure, the web server 110 may include one or more processors 132 and one or more databases 134 (e.g., any suitable repository of website data, etc.).” Examiner notes the training samples are the detected list, which can be used as a training data set. Examiner further notes the list is accessed (retrieving) by the application server and stored by the web server using database 134.)
Regarding claim 20, Martins teaches
A system, comprising: one or more processors; and memory configured to store instructions, the instructions, when executed by the one or more processors, cause the one or more processors to (Martins, paragraph 0115, “In some aspects of this disclosure, a tangible, non-transitory apparatus or article of manufacture comprising a tangible, non-transitory computer usable or readable medium having control logic (software) stored thereon may also be referred to herein as a computer program product or program storage device. This includes, but is not limited to, computer system 1100, main memory 1108, secondary memory 1110, and removable storage units 1118 and 1122, as well as tangible articles of manufacture embodying any combination of the foregoing. Such control logic, when executed by one or more data processing devices (such as computer system 1100), may cause such data processing devices to operate as described herein.”)
retrieve a list of custom-defined categories of a database maintained by a third-party platform (Martins, paragraph 0067, “At 702, method 400 may include querying or retrieving a database of entities with their associated fraudulent labels. For example, method 700 carried out by application server 120 may query a labeled data set that has been processed and stored in database 118. The data set may consist of multiple populations of entities, for example, fraudulent, high-risk, and not fraudulent merchants.” Examiner notes the custom-defined categories of the database are the labels of either fraudulent or non-fraudulent for each entity.)
the list of custom-defined categories defined by an entity who uses the third-party data platform (Martins, paragraph 0068, “According to some aspects, at 704, method 700 may include identifying a first group of entities (e.g., merchants) that have a high fraud risk level. This may entail having application server 120 search and detect the respective label of each entity and identify the entities that are labeled as high-risk or risky entities.” Examiner notes the entity is the application server 120, which defines and labels whether the entities are fraudulent or not.)
access a plurality of training samples for training a machine-learned encoder model (Martins, paragraph 0008, “FIG. 4 illustrates an example process flow for a predictive model trained for embedding analysis for entity classification, according to some aspects of this disclosure.” Martins, paragraph 0056, “FIG. 4 illustrates an input and output representation for leveraging embeddings (e.g., embedding analysis for entity classification), according to some aspects of this disclosure. According to some aspects, in reference to natural language processing (NLP), a word embedding is a term used for the representation of words for text analysis, typically in the form of a real-valued vector that encodes the meaning of the word such that the words that are closer in the vector space are expected to be similar in meaning. FIG. 4 illustrates methodologies whereby positive and negative samples of merchant associates are used as input to a neural network to generate an output vector graph.” Martins, paragraph 0062, “According to some aspects of this disclosure, the ML model 122 may be trained in a dataset that includes labels for fraudulent merchants, potentially risky (e.g. high risk) merchants, non-fraudulent merchants, and/or any other type of classification based on aggregations of historical data, such as historical transactions and fraud reported thereupon.” Examiner notes the plurality of training samples are the positive and negative samples of merchant associates.)
a training sample comprises a positive data instance belonging to a target category from the list of custom-defined categories and a negative data instance outside of the target category (Martins, paragraph 0056, “FIG. 4 illustrates methodologies whereby positive and negative samples of merchant associates are used as input to a neural network to generate an output vector graph. A positive sample may be one where merchants are deemed sufficiently proximate in a transaction space, thereby requiring further analysis. A negative sample may be one where merchants are deemed insufficiently proximate in a transaction space, thereby not requiring further analysis. For example, an association between Merchant 1 and Merchant 2 may be detected. If a consumer transacts at Merchant 1 and within a short period of time transacts at Merchant 2, it can be determined that Merchant 1 and Merchant 2 have a positive association (e.g., deemed sufficiently proximate for further analysis).”)
train the machine-learned encoder model that generates embeddings of data instances (Martins, paragraph 0059, “According to some aspects, the model may be trained on a link prediction task. That is, given a merchant pair (e.g., Merchant 1 and Merchant 2), the model may be trained to predict whether this pair has at least one account in common. In this regard, the model produces embeddings that encode accounts in common between merchants—thereby allowing organization system 108 to link merchants to detect fraudulent patterns.”)
wherein the machine-learned encoder model is trained to separate a plurality of embeddings of positive data instances belong to the target category from a plurality of embeddings of negative data instances (Martins, paragraph 0058, “One example of training is the SVM, where features having the smallest weights are removed and the algorithm is re-trained with the remaining weights, wherein the process is repeated until features remain that are able to accurately separate the data into different patterns or classes…Such trends may correspond to increasing similarities between close merchants (e.g., Merchant 1 and Merchant 2) and decreasing similarities between distant merchants (e.g., Merchant 1 and Merchant 3). A vector graph may be generated as an output representing a mapping of the merchants whereby Merchant 1 and Merchant 2 are closely related (due to similarities in transactions, users, etc.), and Merchant 1 and Merchant 3 are distantly related.” Examiner notes the positive data instances are the close merchants and negative instances are the distant merchants.)
receive a target data instance that is to be imported to the third-party data platform (Martins, FIG. 4, 702. Martins, FIG. 1. Martins, paragraph 0047, “According to some aspects of this disclosure, application server 120 may comprise one or more computer systems configured to compile data from a plurality of sources, such as location server 110, communication server 112, and transaction server 114, correlate and analyze the compiled data in real-time (continuously), arrange the compiled data, generate derived data based on the compiled data, and store the compiled and derived in a database such as database 118. According to some aspects of this disclosure, application server 120 may perform merchant detection and prediction operations, for example, via a trained machine learning (ML) model 122 included with the application server 120. Example merchant detection and prediction operations performed by application server 120 (e.g., the ML model 122, etc.) are further described herein with reference to FIGS. 2-6.” Examiner notes the target data instance is the retrieved labeled entity in step 702, which may be stored (imported) into a database, as described in paragraph 0047 and FIG. 1.)
generate features of the target data instance to prepare the target data instance for further processing (Martins, FIG. 4, 708. Martins, paragraph 0070, “At 708, method 700 may include joining (or assigning) entities to their associated embeddings. Martins, paragraph 0064, “According to some aspects of this disclosure, information such as embeddings describing fraudulent merchants may be used to seed a nearest neighbor search.” Examiner notes the features of the target data instance are the embeddings generated by the model for the entities, which is used for classifying the entities.)
apply the machine-learned encoder model to the target data instance to determine an assignment of a category from the list of custom-defined categories (Martins, FIG. 4, step 712.)
and export the target data instance including the assignment of the category to the third-party data platform (Martins, FIG. 4, step 712. Martins, FIG. 1. Martins, paragraph 0047, “According to some aspects of this disclosure, application server 120 may comprise one or more computer systems configured to compile data from a plurality of sources, such as location server 110, communication server 112, and transaction server 114, correlate and analyze the compiled data in real-time (continuously), arrange the compiled data, generate derived data based on the compiled data, and store the compiled and derived in a database such as database 118. According to some aspects of this disclosure, application server 120 may perform merchant detection and prediction operations, for example, via a trained machine learning (ML) model 122 included with the application server 120. Example merchant detection and prediction operations performed by application server 120 (e.g., the ML model 122, etc.) are further described herein with reference to FIGS. 2-6.” Examiner notes the classified (labeled) target data instance is the entity in step 712, which is assigned a category of either fraudulent or non-fraudulent. Examiner further notes the entity than has been assigned a category may be stored (imported) into a database, as described in paragraph 0047 and FIG. 1.)
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claim(s) 2, 8, and 16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Martins in view of Kumar et al. (“Vehicle Re-identification: an Efficient Baseline Using Triplet Embedding”).
Regarding claim 2, Martins teaches
The computer-implemented method of claim 1, wherein training of the machine-learned encoder model comprises: receiving the training samples comprising positive data instances belonging to the target category and negative data instances outside of the target category (Martins, paragraph 0008, “FIG. 4 illustrates an example process flow for a predictive model trained for embedding analysis for entity classification, according to some aspects of this disclosure.” Martins, paragraph 0056, “FIG. 4 illustrates methodologies whereby positive and negative samples of merchant associates are used as input to a neural network to generate an output vector graph. A positive sample may be one where merchants are deemed sufficiently proximate in a transaction space, thereby requiring further analysis. A negative sample may be one where merchants are deemed insufficiently proximate in a transaction space, thereby not requiring further analysis.”)
generating a plurality of positive embeddings corresponding to the positive data instances and a plurality of negative embeddings corresponding to the negative data instances (Martins, paragraph 0057, “According to some aspects, when a table of positive and negative associations of merchants is generated, the table may be used as an input to a machine learning model using a neural network.” Examiner notes the embedding model generates a table of positive and negative associations of merchants, wherein the positive and negative associations of merchants are the positive and negative embeddings, respectively.)
backpropagating the loss function through the machine-learned encoder model (Martins, paragraph 0058, “Alternately, a neural network type algorithm is used, such as a back-propagation neural network, where there may be a weight matrix for each layer of the neural network, wherein for each layer, a bias vector is defined. The model may first undergo forward propagation. In forward propagation, the input signals may be multiplied by the weights in the weight matrices for each layer, and activation functions may transform the output at each layer, wherein the end output may be calculated. Back propagation aids in computing the error of partial derivatives, which can then be minimized across layers and can form the central mechanism by which the neural network learns.” Examiner notes the loss function is the error of partial derivatives, which is minimized.)
and adjusting one or more parameters of the machine-learned encoder model through the backpropagation (Martins, paragraph 0057, “For example, a support vector machine (SVM), random-forest, K means clustering, a multi-layer neural network with back-propagation, or other algorithms may be used with several associating factors to adjust weights and learn target data within the tables.”)
Martins does not, but Kumar teaches
determining a loss function that measure distances for a plurality of embedding pairs, each embedding pair comprising at least one of the positive embeddings and one of the negative embeddings (Kumar, page 3, “Let D(xi,xj) : RF ×RF → R be a metric measuring distance of images xi and xj in embedding space.” Kumar, page 3, “Triplet loss forces the data points from the same class to be closer to each other than a data point form any other class. Notice that contrary to contrastive loss in (2), triplet loss adds context to the loss function by considering both a positive and negative pair distances from the same point. ltriplet(a, p, n) = [Dap − Dan + α]+.” Examiner notes a pair consists of Dap and Dan, which are the distances between the anchor and the positive and negative embeddings.)
and the distance for each embedding pair measuring a distance between the one of the positive embeddings and one of the negative embeddings (Kumar, page 3, ltriplet(a, p, n) = [Dap − Dan + α]+.” Examiner notes the distance is between positive and negative embeddings.)
Martins and Kumar utilize embedding machine learning models and thus are analogous to the claimed invention. It would have been obvious to one having ordinary skill in the art prior to the effective filing date of the application to have modified Martins to utilize a loss function based on distances between positive and negative embeddings like Kumar. Doing so would have been advantageous because it is utilizing a triplet loss function and “several approaches [2], [9], [11], [13], [30], [35] have reported state of-the-art performances using triplet loss. This superiority of triplet loss is attributed to the additional context using three samples” (Kumar, page 3).
Regarding claim 8, Martin teaches The computer-implemented method of claim 6. Martins does not, but Kumar teaches
wherein defining the loss function that calculates a relationship between embeddings of anchor, positive and negative data instances comprises: defining a triplet loss function to minimize the relative distance between embeddings of positive data instances and maximize the relative distance between embeddings of negative data instances (Kumar, page 3, “Triplet loss forces the data points from the same class to be closer to each other than a data point form any other class. Notice that contrary to contrastive loss in (2), triplet loss adds context to the loss function by considering both a positive and negative pair distances from the same point.”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine the teachings of Martins with the teachings of Kumar for the same reasons disclosed in claim 2.
Regarding claim 16, Martins teaches
The non-transitory computer-readable storage medium of claim 15, wherein training of the machine-learned encoder model comprises: receiving the training samples comprising positive data instances belonging to the target category and negative data instances outside of the target category (Martins, paragraph 0056, “FIG. 4 illustrates methodologies whereby positive and negative samples of merchant associates are used as input to a neural network to generate an output vector graph. A positive sample may be one where merchants are deemed sufficiently proximate in a transaction space, thereby requiring further analysis. A negative sample may be one where merchants are deemed insufficiently proximate in a transaction space, thereby not requiring further analysis. For example, an association between Merchant 1 and Merchant 2 may be detected. If a consumer transacts at Merchant 1 and within a short period of time transacts at Merchant 2, it can be determined that Merchant 1 and Merchant 2 have a positive association (e.g., deemed sufficiently proximate for further analysis).”)
generating a plurality of positive embeddings corresponding to the positive data instances and a plurality of negative embeddings corresponding to the negative data instances (Martins, paragraph 0057, “According to some aspects, when a table of positive and negative associations of merchants is generated, the table may be used as an input to a machine learning model using a neural network.” Examiner notes the embedding model generates a table of positive and negative associations of merchants, wherein the positive and negative associations of merchants are the positive and negative embeddings, respectively.)
backpropagating the loss function through the machine-learned encoder model (Martins, paragraph 0058, “Alternately, a neural network type algorithm is used, such as a back-propagation neural network, where there may be a weight matrix for each layer of the neural network, wherein for each layer, a bias vector is defined. The model may first undergo forward propagation. In forward propagation, the input signals may be multiplied by the weights in the weight matrices for each layer, and activation functions may transform the output at each layer, wherein the end output may be calculated. Back propagation aids in computing the error of partial derivatives, which can then be minimized across layers and can form the central mechanism by which the neural network learns.” Examiner notes the loss function is the error of partial derivatives, which is minimized.)
and adjusting one or more parameters of the machine-learned encoder model through the backpropagation (Martins, paragraph 0057, “For example, a support vector machine (SVM), random-forest, K means clustering, a multi-layer neural network with back-propagation, or other algorithms may be used with several associating factors to adjust weights and learn target data within the tables.”)
Martins does not, but Kumar teaches
determining a loss function that measure distances for a plurality of embedding pairs, each embedding pair comprising at least one of the positive embeddings and one of the negative embeddings (Kumar, page 3, “Let D(xi,xj) : RF ×RF → R be a metric measuring distance of images xi and xj in embedding space.” Kumar, page 3, “Triplet loss forces the data points from the same class to be closer to each other than a data point form any other class. Notice that contrary to contrastive loss in (2), triplet loss adds context to the loss function by considering both a positive and negative pair distances from the same point. ltriplet(a, p, n) = [Dap − Dan + α]+.” Examiner notes a pair consists of Dap and Dan, which are the distances between the anchor and the positive and negative embeddings.)
and the distance for each embedding pair measuring a distance between the one of the positive embeddings and one of the negative embeddings (Kumar, page 3, ltriplet(a, p, n) = [Dap − Dan + α]+.” Examiner notes the distance is between positive and negative embeddings.)
Martins and Kumar utilize embedding machine learning models and thus are analogous to the claimed invention. It would have been obvious to one having ordinary skill in the art prior to the effective filing date of the application to have modified Martins to utilize a loss function based on distances between positive and negative embeddings like Kumar. Doing so would have been advantageous because it is utilizing a triplet loss function and “several approaches [2], [9], [11], [13], [30], [35] have reported state of-the-art performances using triplet loss. This superiority of triplet loss is attributed to the additional context using three samples” (Kumar, page 3).
Claim(s) 4, 14, and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Martins in view of Dasgupta et al. (US20200334282).
Regarding claim 4, Martins teaches The computer-implemented method of claim 3. Martins does not, but Dasgupta teaches
wherein creating the list of custom-defined categories of data instances comprises: providing a category creation tool on a user interface of the third-party data platform (Dasgupta, paragraph 0021, “A domain 110 may be an environment in which a system operates and/or an environment for a group of units and individuals to use common domain knowledge to organize activities, information and entities related to the domain 110 in a specific way. An example of a domain 110 may be an organization, such as a business, an institute, or a subpart thereof and the data within it. A domain 110 may be associated with a specific domain knowledge ontology, which could include representations, naming, definitions of categories, properties, logics, and relationships among various concepts, data, transactions, and entities that are related to the domain 110.” Dasgupta, paragraph 0027, “A client device 130 includes one or more applications 132 and user interfaces 114 that may display visual elements of the applications 132. In one embodiment, a resource planning application 115 allows a client to input master data 214 and to define domain knowledge ontology 226.” Examiner notes that a client device includes one or more user interfaces that contain a resource planning application, which allows a client to input data and define domain knowledge ontology. Examiner further notes the categories are the domains and the resource planning application allows the user to modify and define domains.)
defining custom-defined categories of data instances on the category creation tool, wherein defining the custom-defined categories of data instances on the category creation tool comprises: selecting, on the category creation tool, one or more data parameters that determine the category of a data instance (Dasgupta, paragraph 0021, “A domain 110 may be associated with a specific domain knowledge ontology, which could include representations, naming, definitions of categories, properties, logics, and relationships among various concepts, data, transactions, and entities that are related to the domain 110.” Dasgupta, paragraph 0027, “In one embodiment, a resource planning application 115 allows a client to input master data 214 and to define domain knowledge ontology 226.” Examiner notes that the domain knowledge ontology (representations, naming, properties, logics, and relationships) includes data parameters that can determine the domain (category) of a data instance, and the category creation tool is the resource planning application.)
wherein the data parameters comprise any one of: an origin of the data instance; an amount associated with the data instance; a type associated with the data instance; a user associated with the data instance; and contextual information associated with the data instance (Dasgupta, paragraph 0021, “A domain 110 may be associated with a specific domain knowledge ontology, which could include representations, naming, definitions of categories, properties, logics, and relationships among various concepts, data, transactions, and entities that are related to the domain 110.” Examiner notes that relationships among various concepts, data, transactions, and entities are contextual information associated with the data.)
and assigning, on the category creation tool, the one or more data parameters to the defined custom-defined categories of data instances (Dasgupta, paragraph 0021, “A domain 110 may be associated with a specific domain knowledge ontology, which could include representations, naming, definitions of categories, properties, logics, and relationships among various concepts, data, transactions, and entities that are related to the domain 110.” Dasgupta, paragraph 0027, “In one embodiment, a resource planning application 115 allows a client to input master data 214 and to define domain knowledge ontology 226.” Examiner notes the user can use the resource planning application (category creation tool) to input data and define ontology, which includes the data parameters.)
Martins and Dasgupta utilize embedding models to categorize large amounts of data to be stored in a database and thus are analogous to the claimed invention. It would have been obvious to one having ordinary skill in the art prior to the effective filing date of the application to have modified Martins to implement a category creation tool than can define, update, and modify categories and data parameters like Dasgupta. Doing so would have been advantageous because “it allows the improved process model 340 may be passed back into a resource planning program 115 through the API of the program 115. This allows the computing server 120 to automatically optimize a process model 222 and feed the improved process model 340 back to the domain 110” (Dasgupta, paragraph 0091).
Regarding claim 14, Martins teaches
The computer-implemented method of claim 1, further comprising: generating a message to a user, wherein generating the message to the user comprises: applying a natural language generation process to the target data instance and its category assignment to provide one or more sentences that present them to the user (Martins, FIG. 4, step 714. Martins, paragraph 0072, “According to some aspects of this disclosure, application server 120 may also output a notification to user device 102. The notification may include, for example, a notification that the transaction was not authorized due to merchant classification (e.g., fraudulent/high risk), a warning notification informing the user that the merchant previously transacted with is fraudulent, and that future transaction will not be authorized. By doing so, organization system 108 can maintain a high degree of satisfaction from the user experience standpoint because the user can be warned of potential threats, and be promptly notified that future transactions with this merchant will not be authorized.” Examiner notes that after the target data instance (merchant) has been assigned a category (fraudulent or non-fraudulent), the application server provides one or more sentences (warning notification) and presents them to the user.)
transmitting the message to the user through a communication channel comprising a short message service (SMS) message, email, or a software as a service (SaaS) platform (Martins, paragraph 0036, “According to some aspects of this disclosure, the web server 110 may include a computer system and/or communication module that enables/facilitates communications between the user device 102, for example, via an application (e.g., a mobile application, etc.), a user interface, a chat program, an instant messaging program, a voice-to-text program, an SMS message, email, or any other type or format of written or electronic communication.”)
Martins does not, but Dasgupta teaches
in response to transmitting the message to the user, receiving feedback from the user (Dasgupta, paragraph 0056, “The interface 262 may provide an interface to transmit and display results and data generated from the analysis performed by the computing server 120. For example, a process model 222 may be visualized as a process map. The interface 262 may be in the form of a GUI to display the process map and allow users to provide inputs via the GUI. A user may input manual actions to the process models 222. For example, a user may specify that a certain step is automated, eliminate a certain step in a process model 222 that is deemed unnecessary, and change one or more steps. The user may also provide feedback and corrections to a process model 222.” Examiner notes the message being transmitted to the user is the interface displaying results and data generated by the server.)
and updating the category assignment for the target data instance based on the feedback received from the user (Dasgupta, paragraph 0056, “For example, a process model 222 that is generated by automatic process mining engine 254 or optimized by the process optimization engine 256 may be corrected and further defined by a user input.” Dasgupta, abstract, “The computing server may aggregate, based on domain knowledge ontology of the domain, the master data and the transaction data to generate a fact table.” Examiner notes the model can be corrected by user input, and thus the model data classification can be updated based on user feedback.)
Martins and Dasgupta utilize embedding models to categorize large amounts of data to be stored in a database and thus are analogous to the claimed invention. It would have been obvious to one having ordinary skill in the art prior to the effective filing date of the application to have modified Martins to utilize user feedback to update data like Dasgupta. Doing so would have been advantageous because “The optimization and feedback process may be performed automatically to form a self-optimized closed-loop system” (Dasgupta, paragraph 0061), which would ensure “the process model is automatically optimized (self-optimized)” (Dasgupta, paragraph 0061). User feedback would also allow users to correct errors unknown to the computer, further increasing the accuracy of category assignment and classification by the model.
Regarding claim 18, Martins teaches The non-transitory computer-readable storage medium of claim 17. Martins does not, but Dasgupta teaches
wherein creating the list of custom-defined categories of data instances comprises: providing a category creation tool on a user interface of the third-party data platform (Dasgupta, paragraph 0021, “A domain 110 may be an environment in which a system operates and/or an environment for a group of units and individuals to use common domain knowledge to organize activities, information and entities related to the domain 110 in a specific way. An example of a domain 110 may be an organization, such as a business, an institute, or a subpart thereof and the data within it. A domain 110 may be associated with a specific domain knowledge ontology, which could include representations, naming, definitions of categories, properties, logics, and relationships among various concepts, data, transactions, and entities that are related to the domain 110.” Dasgupta, paragraph 0027, “A client device 130 includes one or more applications 132 and user interfaces 114 that may display visual elements of the applications 132. In one embodiment, a resource planning application 115 allows a client to input master data 214 and to define domain knowledge ontology 226.” Examiner notes that a client device includes one or more user interfaces that contain a resource planning application, which allows a client to input data and define domain knowledge ontology. Examiner further notes the categories are the domains and the resource planning application allows the user to modify and define domains.)
defining custom-defined categories of data instances on the category creation tool, wherein defining the custom-defined categories of data instances on the category creation tool comprises: selecting, on the category creation tool, one or more data parameters that determine the category of a data instance (Dasgupta, paragraph 0021, “A domain 110 may be associated with a specific domain knowledge ontology, which could include representations, naming, definitions of categories, properties, logics, and relationships among various concepts, data, transactions, and entities that are related to the domain 110.” Dasgupta, paragraph 0027, “In one embodiment, a resource planning application 115 allows a client to input master data 214 and to define domain knowledge ontology 226.” Examiner notes that the domain knowledge ontology (representations, naming, properties, logics, and relationships) includes data parameters that can determine the domain (category) of a data instance, and the category creation tool is the resource planning application.)
wherein the data parameters comprise any one of: an origin of the data instance; an amount associated with the data instance; a type associated with the data instance; a user associated with the data instance; and contextual information associated with the data instance (Dasgupta, paragraph 0021, “A domain 110 may be associated with a specific domain knowledge ontology, which could include representations, naming, definitions of categories, properties, logics, and relationships among various concepts, data, transactions, and entities that are related to the domain 110.” Examiner notes that relationships among various concepts, data, transactions, and entities are contextual information associated with the data.)
and assigning, on the category creation tool, the one or more data parameters to the defined custom-defined categories of data instances (Dasgupta, paragraph 0021, “A domain 110 may be associated with a specific domain knowledge ontology, which could include representations, naming, definitions of categories, properties, logics, and relationships among various concepts, data, transactions, and entities that are related to the domain 110.” Dasgupta, paragraph 0027, “In one embodiment, a resource planning application 115 allows a client to input master data 214 and to define domain knowledge ontology 226.” Examiner notes the user can use the resource planning application (category creation tool) to input data and define ontology, which includes the data parameters.)
Martins and Dasgupta utilize embedding models to categorize large amounts of data to be stored in a database and thus are analogous to the claimed invention. It would have been obvious to one having ordinary skill in the art prior to the effective filing date of the application to have modified Martins to implement a category creation tool than can define, update, and modify categories and data parameters like Dasgupta. Doing so would have been advantageous because “it allows the improved process model 340 may be passed back into a resource planning program 115 through the API of the program 115. This allows the computing server 120 to automatically optimize a process model 222 and feed the improved process model 340 back to the domain 110” (Dasgupta, paragraph 0091).
Claim(s) 6 and 9 is/are rejected under 35 U.S.C. 103 as being unpatentable over Martins in view of Kumar and further in view of Rusu et al. (“META-LEARNING WITH LATENT EMBEDDING OPTIMIZATION”).
Regarding claim 6, Martins teaches The computer-implemented method of claim 1. Martins does not, but Rusu teaches
wherein training the machine-learned encoder model that generates embeddings of data instances comprises: initializing the machine-learned encoder model with predetermined parameters (Rusu, page 1, “First, the initial parameters for a new task are conditioned on the training data, which enables a task-specific starting point for adaptation.” Examiner notes initial parameters are used to initialize the encoder model, which are predetermined by the training data.)
and evaluating the training of the machine-learned encoder (Rusu, page 5, “For each task instance Ti, the initialization and adaptation procedure produce a new classifier fθi tailored to the training set Dtr of the instance, which we can then evaluate on the validation set of that instance Dval.”)
Martins and Rusu teach an embedding model and thus are analogous to the claimed invention. It would have been obvious to one having ordinary skill in the art prior to the effective filing date of the application to have modified Martins to initialize the embedding model with predetermined parameters and evaluate the training of the encoder model like Rusu. This is because initializing and validating a machine learning model are established standard practices in the art. This would be advantageous because “By incorporating a relation network into the encoder, this initialization can better consider the joint relationship between all of the input data. Second, by optimizing in the lower-dimensional latent space, the approach can adapt the behaviour of the model more effectively. Further, by allowing this process to be stochastic, the ambiguities present in the few-shot data regime can be expressed” (Rusu, page 1).
Martins and Rusu do not, but Kumar teaches
defining a loss function that calculates a relationship between embeddings of anchor, positive and negative data instances (Kumar, page 3, “Let D(xi,xj) : RF ×RF → R be a metric measuring distance of images xi and xj in embedding space.” Kumar, page 3, “Triplet loss forces the data points from the same class to be closer to each other than a data point form any other class. Notice that contrary to contrastive loss in (2), triplet loss adds context to the loss function by considering both a positive and negative pair distances from the same point. ltriplet(a, p, n) = [Dap − Dan + α]+.” Examiner notes a pair consists of Dap and Dan, which are the distances between the anchor and the positive and negative embeddings.)
training the machine-learned encoder by processing each training sample to generate embeddings using the loss function (Kumar, page 3, “Consider a dataset X = {(xi,yi)}N i=1 of N training images xi ∈ RD and their corresponding class labels yi ∈ {1···C}. Re-identification approaches aim to learn an embedding f(x;θ) : RD → RF to map images in RD onto a feature (embedding) space in RF such that images of similar identity are metrically close in this feature space. θ corresponds to the parameters of the learning function.” Kumar, page 3, “Triplet loss forces the data points from the same class to be closer to each other than a data point form any other class. Notice that contrary to contrastive loss in (2), triplet loss adds context to the loss function by considering both a positive and negative pair distances from the same point. ltriplet(a, p, n) = [Dap − Dan + α]+.” Kumar, pages 3-4, “When training from scratch, we use standard learning rate of 0.001. We reduce this rate to 0.0003 when using an imagenet based pre-trained model. For online data augmentation a standard image-flip operation is used. We use Nvidia’s Volta GPU for hardware and Tensorflow [1] as the software platform. We replace the margin α in triplet loss (4) by softplus function: ln(1 + exp(·)) which avoids the need of tuning this margin [11]. “ Examiner notes the encoder model is trained using triplet loss to learn an embedding and maps input images onto an embedding space, therefore generating embeddings during training.)
Martins and Kumar utilize embedding machine learning models and thus are analogous to the claimed invention. It would have been obvious to one having ordinary skill in the art prior to the effective filing date of the application to have modified Martins to utilize a loss function based on distances between positive and negative embeddings like Kumar. Doing so would have been advantageous because it is utilizing a triplet loss function for training and “several approaches [2], [9], [11], [13], [30], [35] have reported state of-the-art performances using triplet loss. This superiority of triplet loss is attributed to the additional context using three samples” (Kumar, page 3).
Regarding claim 9, Martins, Rusu, and Kumar teach The computer-implemented method of claim 6. Rusu further teaches
wherein evaluating the training of the machine-learned encoder comprises: applying a validation dataset to the machine-learned encoder model (Rusu, page 7, “The miniImageNet dataset (Vinyals et al., 2016) is a subset of 100 classes selected randomly from the ILSVRC-12 dataset (Russakovsky et al., 2014) with 600 images sampled from each class. Following the split proposed by Ravi & Larochelle (2017), the dataset is divided into training, validation, and test meta-sets, with 64, 16, and 20 classes respectively.”)
the validation dataset comprising a plurality of data instances representing diverse categories from a list of custom categories in a database (Rusu, page 7, “In order to answer the final question we scale up our approach to 1-shot and 5-shot classification problems defined using two commonly used ImageNet subsets. The miniImageNet dataset (Vinyals et al., 2016) is a subset of 100 classes selected randomly from the ILSVRC-12 dataset (Russakovsky et al., 2014) with 600 images sampled from each class. Following the split proposed by Ravi & Larochelle (2017), the dataset is divided into training, validation, and test meta-sets, with 64, 16, and 20 classes respectively.” Rusu, page 9, “Imagenet: A large-scale hierarchical image database.” Examiner notes the miniImageNet dataset comprises data instances (images) representing diverse categories (classes), and is split into a validation set to evaluate the model. Examiner further notes that miniImageNet is a subset of ImageNet, and Imagenet is a large-scale image database.)
determining a metric for the machine-learned encoder model (Rusu, page 8, Table 1. Examiner notes the metric is the accuracy for each model.)
the metric measuring a performance of the model in minimizing the calculated loss function and categorizing new data instances correctly through a comparison of model-predicted categories and actual categories (Rusu, page 2, “The validation set Dval can contain several other samples from the same classes, providing an estimate of generalization performance on the N classes for this problem instance. We note that the validation set of a problem instance Dval (used to optimize a meta-learning objective) should not be confused with the held-out validation meta-set Sval (used for model selection).” Rusu, page 5, “For each task instance Ti, the initialization and adaptation procedure produce a new classifier fθi tailored to the training set Dtr of the instance, which we can then evaluate on the validation set of that instance Dval. During meta-training we use that evaluation to differentiate through the “inner loop” and update the encoder, relation, and decoder network parameters: φe, φr, and φd. Meta-training is performed by minimizing the following objective:
PNG
media_image1.png
39
534
media_image1.png
Greyscale
.” Examiner notes the validation set provides an estimate of the performance of the model, the validation is used to optimize an learning objective, and optimizing the objective is performed by minimizing the objective function, which includes the loss function being aggregated across the entire dataset and thus is minimizing the loss function.)
Claim(s) 12 is/are rejected under 35 U.S.C. 103 as being unpatentable over Martins in view of Rusu.
Regarding claim 12, Martins teaches
The computer-implemented method of claim 1, wherein applying the machine-learned encoder model to the target data instance to determine an assignment of a category from the list of custom-defined categories comprises: accessing, by the machine-learned encoder model, features of the target data instance (Martins, paragraph 0061, “According to some aspects of this disclosure, application server 120 may collect any/all historical data associated with the merchant(s) that have been reported by consumers or flagged by agents of the organization.” Examiner notes the historical data are the features and the target data instance is the entity (merchant).)
generating, by the machine-learned encoder model, embeddings from the features of the target data instance (Martins, paragraph 0082, “According to some aspects of this disclosure, method 800 may further include determining, based on at least one of the historical data or the current data input to a predictive model, the embedding for the first stored entity.”)
and comparing, by the machine-learned encoder model, the target data instance embeddings with embeddings that the model has learned for each data category during training (Martins, paragraph 0037, “According to some aspects of this disclosure, web server 110 may generate and/or store a list of nefarious actors, for example, such as identified malicious and/or fraudulent merchant websites. According to some aspects of this disclosure, web server 110 may enable application server 120, for example, to access the detected list and utilize it as an initial training data set. According to some aspects of this disclosure, detected transactions that are deemed to be in close proximity to the detected fraudulent merchants may then be flagged as potential fraudulent merchants. It can be appreciated that the list of detected merchants may include merchants previously flagged as being fraudulent.” Martins, FIG. 4, step 710. Martins, paragraph 0071, “At 710, method 700 carried out by application server 120 may perform a nearest neighbor search of fraudulent and high-risk entity embeddings. More specifically, taking a fraudulent merchant and a high-risk merchant into consideration, application server 120 may generate a similarity score between the fraudulent merchant and high-risk merchant. Similarity can be defined by, for example, a cosine similarity or other distance measures.” Examiner notes the model compares merchants to previous fraudulent merchant training samples by using similarity metrics such as distance between embeddings.)
assigning, by the machine-learned encoder model, a category to the target data instance based at least on the comparing (Martins, FIG. 4, step 712.)
Martins does not, but Rusu teaches
the embeddings being in multiple layers in a latent space of the machine-learned encoder model (Rusu, page 5, “we use a weighted KL-divergence term to regularize the latent space and encourage the generative model to learn a disentangled embedding” Rusu, page 7, “We used a 3-layer MLP as the underlying model architecture of fθ, and we produced the entire parameter tensor θ with the LEO generator.” Examiner notes the model architecture consists of multiple layers, and the embeddings use a latent space of the encoder.)
Martins and Rusu teach an embedding model and thus are analogous to the claimed invention. It would have been obvious to one having ordinary skill in the art prior to the effective filing date of the application to have modified Martins to place embeddings in a latent space in multiple layers like Rusu. This would be advantageous because “performing adaptation in latent space to generate a simple linear layer can lead to superior generalization” (Rusu, page 6).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Zared O. Cohen whose telephone number is (571)270-0531. The examiner can normally be reached M-Th, 8am to 5pm ET.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Michelle Bechtold can be reached at (571) 431-0762. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/Z.O.C./ Examiner, Art Unit 2148 /MICHELLE T BECHTOLD/Supervisory Patent Examiner, Art Unit 2148
Prosecution Insights