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 .
Status of Claims
Applicant's “Amendment” filed on 8/11/2026 has been considered.
Rejection to Claims 1-20 under 35 USC 101 have not been overcome.
Claims 1-6, 10-20 are amended.
Claims 1-20 are currently pending and have been examined.
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.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Under Step 1 of the Subject Matter Eligibility Test for Products and Processes, the claims must be directed to one of the four statutory categories. All the claims are directed to one of the four statutory categories (YES).
Under Step 2A of the Subject Matter Eligibility Test for Products and Processes (see MPEP § 2106 Subsection III), it is determined whether the claims are directed to a judicially recognized exception. Step 2A is a two-prong inquiry.
Under Prong 1, it is determined whether the claim recites a judicial exception (YES). Taking Claim 20 as representative, the claim recites limitations that fall within the certain methods of organizing human activity grouping of abstract ideas, including:
A system, comprising:
one or more memories storing instructions; and
one or more processors that are coupled to the one or more memories and,
when executing the instructions, are configured to:
receiving one or more user inputs;
determining one or more context features based on one or more context inputs;
determining one or more user features based on the one or more user inputs;
determining, by a first machine learning model, one or more user embedding vectors based on the one or more user features;
determining, by a second machine learning model, one or more context embedding vectors based on the one or more context features;
merging the one or more user embedding vectors and the one or more context embedding vectors to generate a merged embedding vector; and
generating, by a recommendation machine learning model, recommendations based on the vector of merged embeddings.
Certain methods of organizing human activity include:
fundamental economic principles or practices (including hedging, insurance, and mitigating risk)
commercial or legal interactions (including agreements in the form of contracts; legal obligations; advertising, marketing or sales activities or behaviors; and business relations)
managing personal behavior or relationships or interactions between people (including social activities, teaching, and following rules or instructions)
The limitations as emphasized, are a process that, under its broadest reasonable interpretation, covers a commercial interaction. That is, other than reciting that a user interface is generated from the list and products are displayed on the user interface, nothing in the claim element precludes the step from practically being performed by people. For example, but for the “generating a user interface from the second list” and “on the user interface” language, “receiving” and “show” in the context of this claim encompasses advertising, and marketing or sales activities.
If a claim limitation, under its broadest reasonable interpretation, covers a commercial interaction but for the recitation of generic computer components, then it falls within the “certain methods of organizing human activity” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
Under Prong 2, it is determined whether the claim recites additional elements that integrate the exception into a practical application of the exception. This judicial exception is not integrated into a practical application (NO).
The claim recites additional elements beyond the judicial exception(s), including:
A system, comprising:
one or more memories storing instructions; and
one or more processors that are coupled to the one or more memories and,
when executing the instructions, are configured to:
receiving one or more user inputs;
determining one or more context features based on one or more context inputs;
determining one or more user features based on the one or more user inputs;
determining, by a first machine learning model, one or more user embedding vectors based on the one or more user features;
determining, by a second machine learning model, one or more context embedding vectors based on the one or more context features;
merging the one or more user embedding vectors and the one or more context embedding vectors to generate a merged embedding vector; and
generating, by a recommendation machine learning model, recommendations based on the vector of merged embeddings.
These limitations are not indicative of integration into a practical application because:
The additional elements of claim 1 are recited at a high level of generality (i.e. as generic computing hardware) such that they amount to nothing more than mere instructions to implement or apply the abstract idea on a generic computing hardware (or, merely use a computer as a tool to perform an abstract idea.) Specifically, the additional element of a first machine learning model, a second machine learning model, and a recommendation machine learning model, is recited at a high-level of generality (i.e., as a generic processor performing a generic computer function of connecting to a platform on a network) such that it amounts no more than mere instructions to apply the exception using a generic computer component. Accordingly, this additional element does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea. Further, the additional elements to no more than generally link the use of the judicial exception to a particular technological environment or field of use (such as computers or computing networks). For example, stating that the determining of vectors is performed by machine learning models, and the recommendations are generated by a recommendation machine learning model, only generally links the commercial interactions and management of relationships or interactions between people to a computer environment. Employing well-known computer functions to execute an abstract idea, even when limiting the use of the idea to one particular environment, does not integrate the exception into a practical application.
Additionally, the additional elements are insufficient to integrate the abstract idea into a practical application because the claim fails to i) reflect an improvement in the functioning of a computer or an improvement to another technology or technical field, ii) apply the judicial exception with, or use the judicial exception in conjunction with, a particular machine or manufacture that is integral to the claim, iii) effect a transformation or reduction of a particular article to a different state or thing, or iv) apply or use the judicial exception in some other meaningful way beyond generally linking the use of the judicial exception to a particular technological environment.
Accordingly, the judicial exception is not integrated into a practical application.
Under Step 2B, it is determined whether the claims recite additional elements that amount to significantly more than the judicial exception. The claims of the present application do not include additional elements that are sufficient to amount to significantly more than the judicial exception (NO).
In the case of system claim 20 taken individually or as a whole, the additional elements of claim 20 do not provide an inventive concept. As discussed above under step 2A (prong 2) with respect to the integration of the abstract idea into a practical application, the additional elements used to perform the claimed functions amount to no more than a general link to a technological environment.
Even considered as an ordered combination (as a whole), the additional elements do not add anything significantly more than when considered individually.
Therefore, claim 20 does not provide an inventive concept and does not qualify as eligible subject matter.
Claim 14 is a Computer readable medium reciting similar functions as claim 20, and does not qualify as eligible subject matter for similar reasons.
Claim 1 is a method reciting similar functions as claim 1, and does not qualify as eligible subject matter for similar reasons.
Claims 2-13, 15-19 are dependencies of claims 1 and 14. The dependent claims do not add “significantly more” to the abstract idea. They recite additional functions that describe the abstract idea and only generally link the abstract idea to a particular technological environment.
Accordingly, the Examiner concludes that there are no meaningful limitations in the claim that transform the judicial exception into a patent eligible application such that the claim amounts to significantly more than the judicial exception itself. The analysis above applies to all statutory categories of invention.
Claim Rejections - 35 USC § 103
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 of this title, 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.
Claims 1-5, 7-20 are rejected under 35 U.S.C. 103 as being unpatentable over US 2024/0354556 A1 to Rao in view of US 2024/0345551 A1 to Ramanasankaran in view of US 20210004682 A1 to Gong.
Regarding Claim 1, RAO discloses A computer-implemented method for generating recommendations using a recommendation model, the method comprising:
receiving one or more user inputs; ([0097] Given a sequence of actions for a new session of a user, the recommendation module 235 converts the actions into a set of input tokens [cereal]>[milk]>[granola].)
determining one or more context features based on one or more context inputs; ([0094] the list of items that are popular may change over time depending on seasonality, month of the year, and the like. In this manner, the training module 230 can update the token space to reflect the popular items as well as an updated recommendation model that reflects this change of user demand in items of the online system 140.)
determining one or more features based on the one or more user inputs; ([0095] the recommendation module 235 obtains from a feature store 245, a history of user actions for the user. The recommendation module 235 may perform one or more parsing operations, such as filtering the features to user actions for the current browsing session. The recommendation module 235 tokenizes the user actions to a sequence of token ID's.)
determining one or more user embeddings based on the one or more user features; ([0097] In the example of FIG. 3A, the embedding E_1 is generated for the first position (which was masked), embedding E_brown sugar is generated for the second position, embedding E_3 is generated for the third position (which was masked), and embedding E_bacon is generated for the fourth position.)
determining one or more context embeddings based on the one or more context features; ([0097] The embeddings are used to generate a contextual embedding. )
merging the one or more user embeddings and the one or more context embeddings to generate one or more merged embeddings; and ([0088] The training module 230 applies the pre-trained generator model 310 to the sequence of input tokens and generates a set of embeddings E_pasta, E_Alfredo sauce, E_Olive oil. The embeddings are combined to generate a contextual embedding for the sequence.)
generating recommendations based on the one or more merged embeddings.. ([0097] The recommendation module 235 applies the classification layer 430 to the contextual embedding and generates a set of predictions for the items in the vocabulary. As shown in FIG. 4, the items associated with the first and second elements have high likelihoods. [0076] The classification layer is coupled to receive a contextual embedding (e.g., combination of embeddings generated for an input sequence) and generate a set of predictions for the set of items, where each prediction indicates a likelihood the user will interact with a content item or recommendation associated with the respective item given the sequence of actions encoded for the user.)
But does not explicitly disclose user features; by a first machine learning model, by a second machine learning model; merging the one or more user embedding vectors and the one or more context embedding vectors to generate a merged embedding vector.
RAMANASANKARAN, on the other hand, teaches user features. ([0105] feedback received as user input associated with outputs presented by the application sessions 308. The feedback system 400 can incorporate features of the feedback repository 124 and/or feedback trainer 128 described with reference to FIG. 1.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of RAO to include the features as taught by RAMANASANKARAN. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify RAO in order to provide real-time or near real-time data so that the outputs (described below) are in line with both historical data and any recent differential data. (RAMANASANKARAN, [0186]).
Gong, on the other hand, teaches by a first machine learning model, by a second machine learning model; merging the one or more user embedding vectors and the one or more context embedding vectors to generate a merged embedding vector. ([0032] Prediction model 166 of computing system 160 combines the output of sequence model 165 (e.g., a sequence embedding or latent crosses) with other feature embeddings obtained from feature module 163 (e.g., user feature embeddings indicative of other observed characteristics of the user, application feature embeddings indicative of application characteristics, etc.). The combined inputs enable prediction model 166 to model and then predict future behaviors of the user, during particular future engagements with the application provider service or for specific contexts, that are more likely to result in a conversion from the application provider service. [0073] The sequence embedding (e.g., a sequential feature vector) may be encoded into a sing, average embedding. [0101] machine-learned model 300 can be or include one or more of various different types of machine-learned models. Examples of such different types of machine-learned models are provided below for illustration. One or more of the example models described below can be used (e.g., combined) to provide the output data in response to the input data. Additional models beyond the example models provided below can be used as well.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system, as taught by RAO and RAMANASANKARAN to include the features as taught by Gong. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify the combination in order to improve performance (Gong, [0003]).
Regarding Claim 2, RAO in view of RAMANASANKARAN and Gong teaches the method of claim 1.
RAO discloses wherein the recommendation model further comprises: a first model, which processes the one or more user features and generates the one or more user embeddings; ([0094] the list of items that are popular may change over time depending on seasonality, month of the year, and the like. In this manner, the training module 230 can update the token space to reflect the popular items as well as an updated recommendation model that reflects this change of user demand in items of the online system 140.) a second model, which processes the one or more context inputs and generates the one or more context embeddings; ([0097] In the example of FIG. 3A, the embedding E_1 is generated for the first position (which was masked), embedding E_brown sugar is generated for the second position, embedding E_3 is generated for the third position (which was masked), and embedding E_bacon is generated for the fourth position.) a merge layer, which processes the one or more user embeddings and the one or more context embeddings and generates the one or more merged embeddings; ([0088] The training module 230 applies the pre-trained generator model 310 to the sequence of input tokens and generates a set of embeddings E_pasta, E_Alfredo sauce, E_Olive oil. The embeddings are combined to generate a contextual embedding for the sequence.) and a dense layer, which processes the one or more merged embeddings and generates the recommendations. ([0097] The recommendation module 235 applies the classification layer 430 to the contextual embedding and generates a set of predictions for the items in the vocabulary. As shown in FIG. 4, the items associated with the first and second elements have high likelihoods. [0076] The classification layer is coupled to receive a contextual embedding (e.g., combination of embeddings generated for an input sequence) and generate a set of predictions for the set of items, where each prediction indicates a likelihood the user will interact with a content item or recommendation associated with the respective item given the sequence of actions encoded for the user.)
Gong, on the other hand, teaches the first machine learning model; vectors; the second machine learning model; merged embedding vector. ([0032] Prediction model 166 of computing system 160 combines the output of sequence model 165 (e.g., a sequence embedding or latent crosses) with other feature embeddings obtained from feature module 163 (e.g., user feature embeddings indicative of other observed characteristics of the user, application feature embeddings indicative of application characteristics, etc.). The combined inputs enable prediction model 166 to model and then predict future behaviors of the user, during particular future engagements with the application provider service or for specific contexts, that are more likely to result in a conversion from the application provider service. [0073] The sequence embedding (e.g., a sequential feature vector) may be encoded into a sing, average embedding. [0101] machine-learned model 300 can be or include one or more of various different types of machine-learned models. Examples of such different types of machine-learned models are provided below for illustration. One or more of the example models described below can be used (e.g., combined) to provide the output data in response to the input data. Additional models beyond the example models provided below can be used as well.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system, as taught by RAO and RAMANASANKARAN to include the features as taught by Gong. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify the combination in order to improve performance (Gong, [0003]).
Regarding Claim 3, RAO in view of RAMANASANKARAN and Gong teaches the method of claim 2.
RAO does not explicitly teach wherein the first model is a user foundation model.
RAMANASANKARAN, on the other hand, teaches wherein the first model is a user foundation model. ([0178] the model system 1232 can configure a foundational LLM (e.g., GPT model, etc.) using the metadata, the sample data, the contextual data, and the usage data discussed above, with respect to the data manager 1224, to create a fine-tuned LLM)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of RAO to include the features as taught by RAMANASANKARAN. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify RAO in order to provide real-time or near real-time data so that the outputs (described below) are in line with both historical data and any recent differential data. (RAMANASANKARAN, [0186]).
Gong, on the other hand, teaches the first machine learning model. ([0032] Prediction model 166 of computing system 160 combines the output of sequence model 165 (e.g., a sequence embedding or latent crosses) with other feature embeddings obtained from feature module 163 (e.g., user feature embeddings indicative of other observed characteristics of the user, application feature embeddings indicative of application characteristics, etc.). The combined inputs enable prediction model 166 to model and then predict future behaviors of the user, during particular future engagements with the application provider service or for specific contexts, that are more likely to result in a conversion from the application provider service. [0073] The sequence embedding (e.g., a sequential feature vector) may be encoded into a sing, average embedding. [0101] machine-learned model 300 can be or include one or more of various different types of machine-learned models. Examples of such different types of machine-learned models are provided below for illustration. One or more of the example models described below can be used (e.g., combined) to provide the output data in response to the input data. Additional models beyond the example models provided below can be used as well.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system, as taught by RAO and RAMANASANKARAN to include the features as taught by Gong. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify the combination in order to improve performance (Gong, [0003]).
Regarding Claim 4, RAO in view of RAMANASANKARAN and Gong teaches the method of claim 2.
RAO discloses wherein the recommendation model is trained by: performing one or more operations to train the first model of the recommendation model based on user interaction data; and training the second model, the merge layer, and the dense layer of the recommendation model based on contextual task-specific data. ([0090] The training module 230 computes a loss function between the next token vector and the estimated prediction vector that indicates a difference between the two vectors. The training module 230 updates the parameters of the classification layer 330 to reduce the loss function. This process is repeated until a convergence criterion is reached. In this manner, given a sequence of input tokens, the classification layer 330 is incrementally trained to predict the item a user is likely to interact with next. In one or more embodiments, the parameters of the generator model 310 are frozen (unchanged), while the parameters of the classification layer 330 are updated.)
RAMANASANKARAN, on the other hand, teaches freezing parameters of the first model after training the first model; ([0070] the feedback trainer 128 identifies one or more first parameters of the second model 116 to maintain as having predetermined values (e.g., freeze the weights and/or biases of one or more first layers of the second model 116), and performs a training process, such as a fine tuning process, to configure parameters of one or more second parameters of the second model 116 using the feedback)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of RAO to include the features as taught by RAMANASANKARAN. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify RAO in order to provide real-time or near real-time data so that the outputs (described below) are in line with both historical data and any recent differential data. (RAMANASANKARAN, [0186]).
Gong, on the other hand, teaches the first machine learning model; vectors; the second machine learning model; merged embedding vector. ([0032] Prediction model 166 of computing system 160 combines the output of sequence model 165 (e.g., a sequence embedding or latent crosses) with other feature embeddings obtained from feature module 163 (e.g., user feature embeddings indicative of other observed characteristics of the user, application feature embeddings indicative of application characteristics, etc.). The combined inputs enable prediction model 166 to model and then predict future behaviors of the user, during particular future engagements with the application provider service or for specific contexts, that are more likely to result in a conversion from the application provider service. [0073] The sequence embedding (e.g., a sequential feature vector) may be encoded into a sing, average embedding. [0101] machine-learned model 300 can be or include one or more of various different types of machine-learned models. Examples of such different types of machine-learned models are provided below for illustration. One or more of the example models described below can be used (e.g., combined) to provide the output data in response to the input data. Additional models beyond the example models provided below can be used as well.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system, as taught by RAO and RAMANASANKARAN to include the features as taught by Gong. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify the combination in order to improve performance (Gong, [0003]).
Regarding Claim 5, RAO in view of RAMANASANKARAN and Gong teaches the method of claim 4.
RAO discloses wherein training the second model, the merge layer, and the dense layer of the recommendation model based on conceptual task-specific data further comprises: and evaluating the recommendation model using one or more ranking metrics.. ([0073] the training module 230 ranks the items of the retailer with respect to a performance metric, such as conversion, of the item. The training module 230 tokenizes a first subset of items (e.g., N) for the retailer that has a highest performance metric (e.g., highest conversions).)
RAMANASANKARAN, on the other hand, teaches optimizing the parameters of the second model, the merge layer, and the dense layer using an Adam optimizer and a binary cross-entropy loss function; ([0054] For example, the model updater 108 can evaluate an objective function of the convergence condition, such as a loss function (e.g., L1 loss, L2 loss, root mean square error, cross-entropy or log loss, etc.) based on the one or more candidate outputs and the training data; this evaluation can indicate how closely the candidate outputs generated by the candidate second model 116 correspond to the ground truth represented by the training data. The model updater 108 can use any of a variety of optimization algorithms (e.g., gradient descent, stochastic descent, Adam optimization, etc.) to modify one or more parameters (e.g., weights or biases of the layer(s) of the candidate second model 116 that are not frozen) of the candidate second model 116 according to the evaluation of the objective function.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of RAO to include the features as taught by RAMANASANKARAN. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify RAO in order to provide real-time or near real-time data so that the outputs (described below) are in line with both historical data and any recent differential data. (RAMANASANKARAN, [0186]).
Gong, on the other hand, teaches the first machine learning model; vectors; the second machine learning model; merged embedding vector. ([0032] Prediction model 166 of computing system 160 combines the output of sequence model 165 (e.g., a sequence embedding or latent crosses) with other feature embeddings obtained from feature module 163 (e.g., user feature embeddings indicative of other observed characteristics of the user, application feature embeddings indicative of application characteristics, etc.). The combined inputs enable prediction model 166 to model and then predict future behaviors of the user, during particular future engagements with the application provider service or for specific contexts, that are more likely to result in a conversion from the application provider service. [0073] The sequence embedding (e.g., a sequential feature vector) may be encoded into a sing, average embedding. [0101] machine-learned model 300 can be or include one or more of various different types of machine-learned models. Examples of such different types of machine-learned models are provided below for illustration. One or more of the example models described below can be used (e.g., combined) to provide the output data in response to the input data. Additional models beyond the example models provided below can be used as well.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system, as taught by RAO and RAMANASANKARAN to include the features as taught by Gong. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify the combination in order to improve performance (Gong, [0003]).
Regarding Claim 7, RAO in view of RAMANASANKARAN and Gong teaches the method of claim 1.
RAMANASANKARAN, on the other hand, teaches wherein determining the one or more context features based on the one or more context inputs further comprises in response to determining that the one or more context inputs are not available, imputing the one or more context inputs. ([0032] the first model 104 can predict or generate new data (e.g., artificial data; synthetic data; data not explicitly represented in data used for configuring the first model 104)..)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of RAO to include the features as taught by RAMANASANKARAN. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify RAO in order to provide real-time or near real-time data so that the outputs (described below) are in line with both historical data and any recent differential data. (RAMANASANKARAN, [0186]).
Regarding Claim 8, RAO in view of RAMANASANKARAN and Gong teaches the method of claim 7.
RAMANASANKARAN, on the other hand, teaches wherein imputing the one or more context inputs comprises applying a plurality of heuristics. ([0032] the first model 104 can predict or generate new data (e.g., artificial data; synthetic data; data not explicitly represented in data used for configuring the first model 104). [0087] The model system 260 can include one or more rules, heuristics, logic, policies, algorithms, functions, machine learning models, neural networks, scripts, or various combinations thereof to perform operations including configuring one or more machine learning models 268 based on instructions from the training management system 240..)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of RAO to include the features as taught by RAMANASANKARAN. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify RAO in order to provide real-time or near real-time data so that the outputs (described below) are in line with both historical data and any recent differential data. (RAMANASANKARAN, [0186]).
Regarding Claim 9, RAO in view of RAMANASANKARAN and Gong teaches the method of claim 8.
RAMANASANKARAN, on the other hand, teaches wherein the plurality of heuristics comprises at least one of: assigning a null value to title context input when a source title is missing; enriching the missing context by extracting relevant information from a source title; or extracting context input based on typical user engagement determined from historical user interaction data.. ([0174] the system 1200 can include or access various usage data 1220. The usage data 1220 can provide additional context regarding the manner in which data from the data sources 1208 (or from similar data sources, such as used by other building management systems) is accessed, retrieved, processed, loaded, or otherwise used. For example, the usage data 1220 can include a historical access log relating to the data. The historical access log can include a log of who accessed or retrieved various data, as well as when the data was accessed or retrieved.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of RAO to include the features as taught by RAMANASANKARAN. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify RAO in order to provide real-time or near real-time data so that the outputs (described below) are in line with both historical data and any recent differential data. (RAMANASANKARAN, [0186]).
Regarding Claim 10, RAO in view of RAMANASANKARAN and Gong teaches the method of claim 1.
RAMANASANKARAN, on the other hand, teaches wherein determining the one or more user embeddings based on the one or more user features further comprises in response to determining that the one or more user embeddings are not cached: generating, using a user foundation machine learning model, the one or more user embedding vectors based on the one or more user features; and caching the one or more user embedding vectors.. ([0031] The memories can include random access memory (RAM), read-only memory (ROM), hard drive storage, temporary storage, [0178] the model system 1232 can configure a foundational LLM (e.g., GPT model, etc.) using the metadata, the sample data, the contextual data, and the usage data discussed above, with respect to the data manager 1224, to create a fine-tuned LLM that, when utilized with a conversational interface, acts as a data catalog that allows users to obtain data availability and quality information for buildings, building subsystems, or other components or equipment of buildings, as described herein. In some instances, the model system 1232 can configured the foundational LLM a single time to create the fine-tuned LLM. In some other instances, the model system 1232 configure and reconfigure the foundational LLM to create and then periodically update the fine-tuned LLM to ensure that the fine-tuned LLM is up to date.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of RAO to include the features as taught by RAMANASANKARAN. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify RAO in order to provide real-time or near real-time data so that the outputs (described below) are in line with both historical data and any recent differential data. (RAMANASANKARAN, [0186]).
Regarding Claim 11, RAO in view of RAMANASANKARAN and Gong teaches the method of claim 1.
RAMANASANKARAN, on the other hand, teaches The computer-implemented method of claim 1, wherein determining the one or more user embeddings based on the one or more user features further comprises in response to determining that the one or more user embeddings are cached in a cache, retrieving one or more user embeddings from the cache. ([0144] With the knowledge graph nodes and edges translated into natural text prompts and converted into corresponding vectors stored within the embedded vector storage and similarity search database, the system can then perform the user prompt completion generation process to generate one or more user prompt completions. [0031] The memories can include random access memory (RAM), read-only memory (ROM), hard drive storage, temporary storage, [0178] the model system 1232 can configure a foundational LLM (e.g., GPT model, etc.) using the metadata, the sample data, the contextual data, and the usage data discussed above, with respect to the data manager 1224, to create a fine-tuned LLM that, when utilized with a conversational interface, acts as a data catalog that allows users to obtain data availability and quality information for buildings, building subsystems, or other components or equipment of buildings, as described herein. In some instances, the model system 1232 can configured the foundational LLM a single time to create the fine-tuned LLM. In some other instances, the model system 1232 configure and reconfigure the foundational LLM to create and then periodically update the fine-tuned LLM to ensure that the fine-tuned LLM is up to date.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of RAO to include the features as taught by RAMANASANKARAN. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify RAO in order to provide real-time or near real-time data so that the outputs (described below) are in line with both historical data and any recent differential data. (RAMANASANKARAN, [0186]).
Gong, on the other hand, teaches the first machine learning model; embedding vectors; the second machine learning model; merged embedding vector. ([0032] Prediction model 166 of computing system 160 combines the output of sequence model 165 (e.g., a sequence embedding or latent crosses) with other feature embeddings obtained from feature module 163 (e.g., user feature embeddings indicative of other observed characteristics of the user, application feature embeddings indicative of application characteristics, etc.). The combined inputs enable prediction model 166 to model and then predict future behaviors of the user, during particular future engagements with the application provider service or for specific contexts, that are more likely to result in a conversion from the application provider service. [0073] The sequence embedding (e.g., a sequential feature vector) may be encoded into a sing, average embedding. [0101] machine-learned model 300 can be or include one or more of various different types of machine-learned models. Examples of such different types of machine-learned models are provided below for illustration. One or more of the example models described below can be used (e.g., combined) to provide the output data in response to the input data. Additional models beyond the example models provided below can be used as well.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system, as taught by RAO and RAMANASANKARAN to include the features as taught by Gong. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify the combination in order to improve performance (Gong, [0003]).
Regarding Claim 12, RAO in view of RAMANASANKARAN and Gong teaches the method of claim 1.
RAMANASANKARAN, on the other hand, teaches where merging the one or more user embedding vectors and the one or more context embeddings to generate the merged embedding vectors further comprises at least one of concatenation, element-wise multiplication, or feature crossing. ([0056] For example, the system 100 combine (e.g., concatenate) various such classifiers with the data for inputting to the second model 116 during training, for at least a subset of the data used to configure the second model 116, which can enable the second model 116 to be responsive to analogous information for runtime/inference time operations..)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of RAO to include the features as taught by RAMANASANKARAN. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify RAO in order to provide real-time or near real-time data so that the outputs (described below) are in line with both historical data and any recent differential data. (RAMANASANKARAN, [0186]).
Regarding Claim 13, RAO in view of RAMANASANKARAN and Gong teaches the method of claim 1.
However the combination of RAO and RAMANASANKARAN does not explicitly teach where generating recommendations based on the merged embedding vectors further comprises using residual connections in a dense layer of the recommendation machine learning model..
Gong, on the other hand, teaches where generating recommendations based on the merged embedding vectors further comprises using residual connections in a dense layer of the recommendation machine learning model.. ([0084] To ensure that sequence embeddings are properly emphasized and not looked in view of other feature embeddings, prediction model 266 may include skip connections between each hidden layer. In transitioning from one layer to the next, prediction model 266 may concatenate a sequence embedding to the input of each of the hidden layers.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system, as taught by RAO and RAMANASANKARAN to include the features as taught by Gong. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify the combination in order to improve performance (Gong, [0003]).
Claim 14 recites a non-transitory computer-readable medium comprising substantially similar limitations as claim 1. The claim is rejected under substantially similar grounds as claim 1.
Claim 15 recites a non-transitory computer-readable medium comprising substantially similar limitations as claim 2. The claim is rejected under substantially similar grounds as claim 2.
Claim 16 recites a non-transitory computer-readable medium comprising substantially similar limitations as claim 3. The claim is rejected under substantially similar grounds as claim 3.
Claim 17 recites a non-transitory computer-readable medium comprising substantially similar limitations as claim 7. The claim is rejected under substantially similar grounds as claim 7.
Claim 18 recites a non-transitory computer-readable medium comprising substantially similar limitations as claim 9. The claim is rejected under substantially similar grounds as claim 9.
Claim 19 recites a non-transitory computer-readable medium comprising substantially similar limitations as claim 11. The claim is rejected under substantially similar grounds as claim 11.
Claim 20 recites a system comprising substantially similar limitations as claim 1. The claim is rejected under substantially similar grounds as claim 1.
Claims 6 are rejected under 35 U.S.C. 103 as being unpatentable over US 2024/0354556 A1 to Rao in view of US 2024/0345551 A1 to Ramanasankaran in view of US 2025/0245720 A1 to Jadhav.
Regarding Claim 6, RAO in view of RAMANASANKARAN and Gong teaches the method of claim 5.
However the combination of RAO and RAMANASANKARAN does not explicitly teach wherein evaluating the recommendation machine learning model using of the one or more ranking metrics comprises using at least one of a normalized mean reciprocal rank or a normalized discounted cumulative gain.
JADHAV, on the other hand, teaches wherein evaluating the recommendation machine learning model using of the one or more ranking metrics comprises using at least one of a normalized mean reciprocal rank or a normalized discounted cumulative gain. ([0112] In some embodiments, affinity engine 304 is configured to evaluate the one or more models used to generate the affinity score. In assessing the effectiveness of the one or more models, a comprehensive evaluation strategy is employed. The evaluation focuses on the refinement and personalization of recommendations based on customer journey-centric cross-pollination intent. Key metrics such as Normalized Discounted Cumulative Gain (NDCG), Mean Average Precision (MAP), and Mean Reciprocal Rank (MRR) are shown below and are employed to gauge the model's ability to optimize XP product rankings, ensuring that personalized recommendations are not only contextually relevant but also strategically positioned within the customer's journey.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system, as taught by RAO and RAMANASANKARAN to include the features as taught by JADHAV. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify the combination in order to provide relevant and desirable recommendations to the customer (JADHAV, [0003]).
Response to Arguments
Applicant’s arguments filed with respect to the rejection of claims under 35 USC 101 have been fully considered but they are not persuasive.
Applicant argues that “determining, by a first machine learning model, one or more user embedding vectors based on the one or more user features,” “determining, by a second machine learning model, one or more context embedding vectors based on the one or more context features,” “merging the one or more user embedding vectors and the one or more context embedding vectors to generate a merged embedding vector,” and “generating, by the recommendation model, recommendations based on the vector of merged embeddings” are specific to imparting a technological improvement.
Examiner disagrees. The problem of conventional recommendation systems deploying separate machine learning models for performing specific tasks, with each model having a specific data pipeline, processing framework, and set of algorithms tailored to a specific task, and unifying tasks into a single recommendation machine learning model is not reflected in the claim language. No details are recited regarding the specific manner this improvement is performed. The claims reflect a recommendation system that assist users in discovering content, products, or services that are relevant to the users’ interests, as argued at p. 11. This is an improvement to a business method and does not integrate the abstract idea into a practical application.
Applicant’s arguments with respect to rejection of the claim under 35 USC 103 have been considered but are moot in view of new grounds of rejection, necessitated by Applicant’s amendment.
Applicant argues that none or the cited references teaches or suggests merging the one or more user embedding vectors and the one or more context embedding vectors to generate a merged embedding vector.
However, neither Rao nor Ramanasankaran is relied upon to teach the limitation as amended. Examiner directs Applicant’s attention to the office action, above.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Michelle T. Kringen whose telephone number is (571)270-0159. The examiner can normally be reached M-F: 11am-7pm.
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, Marissa Thein can be reached at (571)272-6764. 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.
/MICHELLE T KRINGEN/Primary Examiner, Art Unit 3689