DETAILED ACTION
This examination is in response to the communication filed on 08/29/2024. Claims 1-18 are currently pending, where claims 1 and 10 are independent.
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 .
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 10/20/2025, 04/09/2025, and 08/29/2024 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
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 1-18 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. Independent claim 1 and 10 recite “obtaining text samples”, “converting the text samples into text embeddings”, “generating an outlier-inlier ranking for the text samples…”, “selecting partial samples…according to the outlier-inlier ranking”, “receiving a manual input…to assign labels on the partial samples”, “generating a prompt message…” and “providing the prompt message to a generative pre-trained transformer model”. In addition, independent claim 10 further recites “generating a first prompt message comprising…a feature engineering task instruction”, “generating a second prompt message…for generating feature predictions…”, and “performing a classification algorithm based on the feature predictions…to generate inlier-outlier prediction labels about the unlabeled samples”.
The limitations of “obtaining…”, “converting…”, “generating…”, “selecting…”, “receiving…”, “generating…” and “providing…” and further limitations of “generating…”, and “providing…” as drafted, are a process that, under a broadest reasonable interpretation, covers the abstract idea of “mental processes” because they cover concepts performed in the human mind, including observation, evaluation, judgement and opinion. See MPEP 2106.04(a)(2). That is, other than reciting “a generative pre-trained transformer model” (claim 1), nothing in the claimed elements preclude the steps from practically being performed by a person
obtaining text sample from a dataset (e.g., by the person obtaining a plurality of text documents, plots, or paragraphs),
converting the text samples into text embeddings in a semantic space (e.g., by the person using a simple vector representation of the topics related to the obtained text),
generating an outlier-inlier ranking of the text samples based on an outlier detection algorithm according to distances between the text embeddings in the semantic space (e.g., by the person preforming a distance calculation between the embedding vectors),
selecting partial samples from the text samples according to the outlier inlier ranking (e.g., by the person selecting text sample based on their distance to a predefined topic or domain),
receiving a manual input command to assign manual-input labels on the partial samples (e.g., by the person annotating the partial samples as inliers or outliers),
generating a prompt message according to the partial samples with the manual-input labels and unlabeled samples of the text samples, wherein the prompt message comprises a task instruction, unlabeled data and anchor data generated based on the partial samples with the manual-input labels (e.g., by person generating a text prompt), and
providing the prompt message to a generative pre-trained transformer model for generating inlier-outlier predictions labels about the unlabeled samples (e.g., by the person providing the text prompt for processing) or
generating a first prompt message comprising the partial samples with the manual-input labels and a feature engineering task instruction (e.g., by person generating a text prompt),
providing the first prompt message to a generative pre-trained transformer model for generating distinguishable features (e.g., by the person providing the text prompt for processing or creating a list of features common to the text sample classes),
generating a second prompt message comprising the distinguishable features, the text samples and a feature scoring task instruction (e.g., by person generating a text prompt) and
performing a classification algorithm based on the feature predications of the text samples comprising the partial samples and unlabeled samples, so as to generate inlier-outlier prediction labels about the unlabeled samples (e.g., by the person, annotating/labeling the text samples as in or out of domain based on the features).
This judicial exception is not integrated into a practical application because the additional elements of “a generative pre-trained transformer model” is recited at a high-level of generality. Accordingly, this additional element does not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. Thus, the claims as a whole are directed to an abstract idea (Step 2A, prong two).
Claims 1 and 10 do not include any additional elements that are sufficient to amount to significantly more than the judicial exception because, as discussed above with respect to integration of the abstract idea into a practical application, the additional elements of “providing a prompt to a generative pre-trained transformer” amounts to no more than mere instructions to apply the exception using generic computer components. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept (Step 2B).
With respect to dependent claims 2-6 and 11-12, these claims are directed the criteria used to select and/or label the text samples. These limitations also relate to the abstract idea of “mental processes.” That is nothing in the claimed elements preclude the steps from practically being performed by a person using the recited criteria when selecting and labeling the text samples. No additional elements are present.
With respect to dependent claims 7-8 and 13-18, these claims are directed the algorithm used to detect outliers. These limitations also relate to the abstract idea of “mental processes.” That is nothing in the claimed elements preclude the steps from practically being performed by a person using the recited algorithms when performing inlier-outlier detection. No additional elements are present.
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, 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 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 non-obviousness.
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.
Claims 1, 2, 7 and 9 are rejected under 35 U.S.C. 103 as being unpatentable over Renders et al. (US 2008/0249999A1; herein “Renders”) in view Larson et al. “Outlier Detection for Improved Data Quality and Diversity in Dialog Systems” arXiv:1904.03122v1 [cs.CL] 5 April 2019 (herein “Larson”), further in view of Agrawal et al. “Large Language Models are Few-Shot Clinical Information Extractors” arXov:2205.12689v2 [cs.CL] 30 Nov 2022 (herein “Agrawal”).
Regarding claim 1, Renders teaches a data classification method, comprising:
obtaining text samples from a dataset (Fig. 1, training documents 10 and document selection and optional annotation 12; );
generating an outlier-inlier ranking of the text samples based on an outlier detection algorithm according
selecting partial samples from the text samples according to the outlier-inlier ranking (Fig. 1, outlier document identifications; );
receiving a manual input command to assign manual-input labels on the partial samples (¶[0019] teaches “If supervised document categorization is to be performed using documents pre-annotated with class identifications, then the user suitably applies such annotations via the document selection interface 12” and ¶[0027] teaches “…the outlier thresholder 34 operates interactively in that a user can select the outlier threshold…”; and ¶[0031] teaches “the user can select to assign an ambiguous document to another class. Once the use makes the selected changes, the probabilistic classifier or categorizer 20 is re-run respective to the set of training documents…”); and
generating inlier-outlier prediction labels about the unlabeled samples (¶[0031] teaches “The accepted model is thenceforth available for use in retrieving documents, classifying new documents, and so forth” ).
Renders fails to disclose that the converting the text samples into text embeddings in a semantic space; the outlier detection algorithm is based on distances between the text embeddings in the semantic space; generating a prompt message according to the partial samples with the manual-input labels and unlabeled samples of the text samples, wherein the prompt message comprises a task instruction, unlabeled data and anchor data generated based on the partial samples with the manual-input labels; or providing the prompt message to a generative pre-trained transformer model.
Larson teaches a distance based outlier detection method that includes, inter alia, converting the text samples into text embeddings in a semantic space (page 2 Section 3.1 teaches “We detect outliers in a dataset as follows: 1. Generate a vector representation of each instance [i.e., data sample]”); generating an outlier-inlier ranking of the text samples based on an outlier detection algorithm according to distances between the text embeddings in the semantic space (page 2 Section 3.1 teaches “We detect outliers in a dataset as follows…3. Calculate the distance of each instance from the mean. 4. Rank by distance in ascending order….”).
It would have been obvious to one having ordinary skill in the art, before the effective filing date of the invention, to have replace the statistical model taught by Renders with the distance based outlier algorithm taught by Larson as it merely constitutes the substitution of known processes to achieve the predictable result of allowing detection of both artificial and real errors in the dataset (Larson, p.1, second column, second paragraph.)
The combination of Renders and Larson fails to teach generating a prompt message according to the partial samples with the manual-input labels and unlabeled samples of the text samples, wherein the prompt message comprises a task instruction, unlabeled data and anchor data generated based on the partial samples with the manual-input labels and providing the prompt to a generative pre-trained transformer to generate inlier-outlier prediction labels.
Agrawal teaches utilizing large language models, e.g., GPTs, for zero- and few-shot information extractions. More specifically, page 2, section 2.1 “Prompt-based learning” of Agrawal teaches “In prompt-based learning (also known as in-context
learning), a pretrained language model is adapted to different tasks via priming on natural language prompts—pieces of text that are combined with an input and then fed to the language model to produce an output for that task. This paradigm has been successful for few-shot and zero-shot learning at many general-domain tasks.” Agrawal further notes that prompt-based learning can be extended straightforwardly to classification tasks.
It would have been obvious to one having ordinary skill in the art, before the effective filing date of the invention, to have replaced the rerunning/updated of the distance based model as taught by the combination of Renders and Larson with few-shot prompt-based learning using a GPT as taught by Agrawal as it merely constitutes the substitution of known processes to achieve predictable result of reducing the amount of human annotated data while increasing the reliability and accuracy of the GPT model.
Regarding claim 2, the combination of Renders, Larson, and Agrawal teaches all of the elements of claim 1 (see detailed element mapping above). In addition, Agrawal further teaches at least one of the text samples prone to be outlier according to the outlier-inlier ranking are selected as the partial samples, an amount of the partial samples is fewer than an amount of the unlabeled samples (Page 3, second column, 1st paragraph teaches “Prompt-based learning requires the specification of a prompt template to be applied on the input. In this work, we handcraft our prompt templates using a set of 5 validation examples per task” Few shot prompting inherently includes outlier samples which number less than the number of unlabeled samples).
It would have been obvious to one having ordinary skill in the art, before the effective filing date of the invention, to have replaced the rerunning/updated of the distance based model as taught by the combination of Renders and Larson with few-shot prompt-based learning using a GPT as taught by Agrawal as it merely constitutes the substitution of known processes to achieve predictable result of reducing the amount of human annotated data while increasing the reliability and accuracy of the GPT model.
Regarding claim 7, the combination of Renders, Larson, and Agrawal teaches all of the elements of claim 1 (see detailed element mapping above). In addition, Agrawal further teaches the task instruction in the prompt message is configured to inform the generative pre-trained transformer model to identify outliers in a given dataset (Page 3, second column, 1st paragraph teaches “Prompt-based learning requires the specification of a prompt template to be applied on the input. In this work, we handcraft our prompt templates using a set of 5 validation examples per task” Few shot prompting inherently includes the task instructions for LLM or GPT).
It would have been obvious to one having ordinary skill in the art, before the effective filing date of the invention, to have replaced the rerunning/updated of the distance based model as taught by the combination of Renders and Larson with few-shot prompt-based learning using a GPT as taught by Agrawal as it merely constitutes the substitution of known processes to achieve predictable result of reducing the amount of human annotated data while increasing the reliability and accuracy of the GPT model.
Regarding claim 9, the combination of Renders, Larson, and Agrawal teaches all of the elements of claim 1 (see detailed element mapping above). In addition, Renders further teaches each of the text samples comprises a text passage (Fig. 1, training documents 10) or a combination of a question and a response (the “or language makes this element optional).
Claim 8 is rejected under 35 U.S.C. 103 as being unpatentable over the combination of Renders, Larson, and Agrawal as applied to claim 1 above, and further in view of Karale, Ankita “Outlier Detection Methods and the Challenges for their Implementation with Steaming Data” in Journal of Mobile Multimedia, vol. 16, no. 3, pp. 351-388, July 2020 (herein “Karale”).
Regarding claim 8, the combination of Renders, Larson, and Agrawal teaches all of the elements of claim 1 (see detailed element mapping above). However, the combination fails to disclose or suggest the outlier detection algorithm is implemented by a RANSAC-NN (the “or language makes this element optional) algorithm, an Isolation Forest algorithm (the “or language makes this element optional) or a Local Outlier Factor algorithm.
Karale teaches that the Local Outlier Factor algorithm is a well-known algorithm or outlier detection. More specifically, page 9, section i. teaches “Local Outlier Factor (LOF) strategy given by Breunig et al. [12], is one of the principal crucial, approximately relevant density-based clustering anomaly discovery techniques. The system utilizes the k-nearest neighbors (KNN).” Therefore, it would have been obvious to one having ordinary skill in the art, before the effective filing date of the invention, to have modified the system taught by the combination of Renders, Larson, and Agrawal to utilizes a Local Outlier Factor strategy for the initial outlier detection algorithm as it merely constitutes the substitution of known processes to achieve the predictable result of anomaly/outlier identification based on a k-nearest neighbors.
Allowable Subject Matter
Claim 3-6 and 10-18 would be allowable if rewritten or amended to overcome the rejection under 35 U.S.C. 101 set forth in this Office action.
The following is a statement of reasons for the indication of allowable subject matter:
Regarding dependent claim 3, the combination of Renders, Larson and Agrawal teaches all of the elements of claim 1 (see detailed element mapping above). However, the combination fails to disclose or suggest the manual input command is configured to assign keep labels on keep samples among the partial samples and assign remove labels on remove samples among the partial samples, generating the prompt message comprising: selecting first anchor samples from the keep samples according to a clustering distribution of the keep samples; selecting second anchor samples from the remove samples according to a clustering distribution of the remove samples; and combining the first anchor samples and the second anchor samples to form the anchor data in the prompt message.
Dependent claims 4-6 variously depend from claim 3 and therefore are allowable for at least these respective dependence from claim 3.
Regarding independent claim 10, the combination of Renders, Larson, and Agrwal teaches a data classification method comprising the steps of claim 1. However, the combination fails to disclose the additional steps of: generating a first prompt message comprising the partial samples with the manual-input labels and a feature engineering task instruction; providing the first prompt message to a generative pre-trained transformer model for generating distinguishable features;
generating a second prompt message comprising the distinguishable features, the text samples and a feature scoring task instruction; providing the second prompt message to the generative pre-trained transformer model for generating feature predictions of the text samples relative to the distinguishable features; and performing a classification algorithm based on the feature predictions of the text samples comprising the partial samples and unlabeled samples, so as to generate inlier-outlier prediction labels about the unlabeled samples as recited in independent claim 10.
Dependent claims 11-18 variously depend from independent claim 10 and therefore are allowable for at least these respective dependence from claim 10.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Koch et al. (US 2023/0147685 A1) teaches methods and systems for training a system to detect anomalies in image of documents in a class of documents;
Sabeti et al. (US 2023/0074604 A1) teaches systems and method for enhancing anomaly detection using a pattern dictionary; and
Shriram et al. (US 2023/0161838 A1) teaches techniques that facilitate training an AI model to ensure equitable model performance across different sub-groups.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to PENNY L CAUDLE whose telephone number is (703)756-1432. The examiner can normally be reached M-Th 8:00 am to 5:00 pm eastern.
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, Daniel Washburn can be reached at 571-272-5551. 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.
/PENNY L CAUDLE/Examiner, Art Unit 2657
/DANIEL C WASHBURN/Supervisory Patent Examiner, Art Unit 2657
Prosecution Insights