DETAILED ACTION
1. This office action is in response to Application No. 18821351 filed on 08/30/2024. Claims 1-20 are presented for examination and are currently pending.
Notice of Pre-AIA or AIA Status
2. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
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.
3. Claims 1-20 are rejected under 35 U.S.C 101 because the claimed invention is directed towards an abstract idea without significantly more.
Step 1
Independent claim 1 is directed to a method and falls into one of the four statutory categories.
Step 2A, Prong 1
Claim 1 recites the following abstract ideas:
generating at least one regular expression … based on context description of the at least one instruction and the information about the plurality of enterprise network assets and the configuration of the enterprise network (Mental process directed to generating at least one regular expression based on context description of instruction and the information about enterprise network assets which can be done observing the context description of instruction and information about enterprise network assets and making a judgement on the generation of regular expression);
generating at least one solution for configuring at least one network asset of the plurality of enterprise network assets based on the at least one regular expression (Mental process directed to generating at least one solution for configuring at least one network asset which can be done by observing the enterprise network assets and making a judgement on the solution based on the regular expression); and
Step 2A, Prong 2
Claim 1 recites the following additional elements:
obtaining at least one instruction and information about a plurality of enterprise network assets and configuration of an enterprise network that includes the plurality of enterprise network assets (this limitation is directed to insignificant extra solution activity of data transmission. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(g));
using an artificial intelligence model (this limitation is directed to mere instructions to apply a judicial exception. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(f))
providing the at least one solution to cause a configuration change in the at least one network asset (this limitation is directed to insignificant extra solution activity of data transmission. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(g)).
Step 2B
Claim 1 recites the following additional elements:
obtaining at least one instruction and information about a plurality of enterprise network assets and configuration of an enterprise network that includes the plurality of enterprise network assets (this limitation is directed to insignificant extra solution activity of data transmission and it is well understood routine and conventional. This does not amount to significantly more than judicial exception. See MPEP 2106.05(d)(II), example i);
using an artificial intelligence model (this limitation is directed to mere instructions to apply a judicial exception. This does not amount to significantly more than judicial exception. See MPEP 2106.05(f))
providing the at least one solution to cause a configuration change in the at least one network asset (this limitation is directed to insignificant extra solution activity of data transmission and it is well understood routine and conventional. This does not amount to significantly more than judicial exception. See MPEP 2106.05(d)(II), example i).
4. Dependent claim 2 is directed to a method and falls into one of the four statutory categories.
Claim 2 recites the following abstract ideas:
to generate the at least one regular expression by learning a plurality of regular expressions and corresponding plurality of solutions as ground truths (Mental process directed to generate the at least one regular expression which can be done by observing the plurality of regular expressions and corresponding plurality of solutions as ground truths and making a judgement on the generation of regular expression)
Claim 2 recites the following additional elements:
training the artificial intelligence model (this limitation is directed to mere instructions to apply a judicial exception. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(f))
using a reinforcement learning edit distance score feedback loop (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not integrate the abstract idea into a practical application. see MPEP 2106.05 (h)).
Claim 2 recites the following additional elements:
training the artificial intelligence model (this limitation is directed to mere instructions to apply a judicial exception. This does not amount to significantly more than judicial exception. See MPEP 2106.05(f))
using a reinforcement learning edit distance score feedback loop (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This does not amount to significantly more than judicial exception, see MPEP 2106.05 (h)).
5. Dependent claim 3 is directed to a method and falls into one of the four statutory categories.
Claim 3 do not recite any abstract ideas.
Claim 3 recites the following additional elements:
configuring the at least one network asset based on the at least one solution, wherein training the artificial intelligence model includes (this limitation is directed to mere instructions to apply a judicial exception. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(f)):
tuning the artificial intelligence model using a reinforcement learning feedback loop based on configuring the at least one network asset (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05 (h)).
Claim 3 recites the following additional elements:
configuring the at least one network asset based on the at least one solution, wherein training the artificial intelligence model includes (this limitation is directed to mere instructions to apply a judicial exception. This does not amount to significantly more than judicial exception. See MPEP 2106.05(f):
tuning the artificial intelligence model using a reinforcement learning feedback loop based on configuring the at least one network asset (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This does not amount to significantly more than judicial exception. see MPEP 2106.05 (h)).
6. Dependent claim 4 is directed to a method and falls into one of the four statutory categories.
Claim 4 recites the following abstract ideas:
validating the at least one solution based on whether a network issue is resolved (Mental process directed to validating the at least one solution based on whether a network issue is resolved which can be done by observing the solution and making a judgement whether the network is resolved);
generating a feedback score for the at least one solution, wherein the feedback score is positive based on the at least one solution being validated and is negative based on the at least one solution not being validated (Mental process directed to generating a feedback score for the at least one solution which can be done by observing the feedback score and making a judgment on the core based the solution being validated or not); and
Claim 4 recites the following additional elements:
providing the feedback score to the artificial intelligence model (this limitation is directed to insignificant extra solution activity of data transmission. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(g)).
Claim 4 recites the following additional elements:
providing the feedback score to the artificial intelligence model (this limitation is directed to insignificant extra solution activity of data transmission and it is well understood routine and conventional. This does not amount to significantly more than judicial exception. See MPEP 2106.05(d)(II), example i).
7. Dependent claim 5 is directed to a method and falls into one of the four statutory categories.
Claim 5 do not recite any abstract ideas.
Claim 5 recites the following additional elements:
wherein the artificial intelligence model is a generative large language model (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not integrate the abstract idea into a practical application, see MPEP 2106.05 (h)), and
generating the at least one solution includes: generating one or more code snippets by inputting the information and the at least one instruction into the generative large language model (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not integrate the abstract idea into a practical application, see MPEP 2106.05 (h)); and
executing the one or more code snippets to configure the at least one network asset (this limitation is directed to mere instructions to apply a judicial exception. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(f)).
Claim 5 recites the following additional elements:
wherein the artificial intelligence model is a generative large language model (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This does not amount to significantly more than judicial exception, see MPEP 2106.05 (h)), and
generating the at least one solution includes: generating one or more code snippets by inputting the information and the at least one instruction into the generative large language model (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This does not amount to significantly more than judicial exception, see MPEP 2106.05 (h)); and
executing the one or more code snippets to configure the at least one network asset (this limitation is directed to mere instructions to apply a judicial exception. This does not amount to significantly more than judicial exception. See MPEP 2106.05(f)).
8. Dependent claim 6 is directed to a method and falls into one of the four statutory categories.
Claim 6 recites the following abstract ideas:
mapping the user input to a plurality of feature embeddings (Mental process directed to mapping the user input to a plurality of feature embeddings which can be done by observing the feature embeddings).
Claim 6 recites the following additional elements:
wherein the at least one instruction is a user input in a natural language format, and further comprising (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(h)):
using the artificial intelligence model and based on a plurality of regular expression features (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not integrate the abstract idea into a practical application, see MPEP 2106.05 (h)),
wherein the plurality of regular expression features are indicative of the information about the plurality of enterprise network assets and the configuration of the enterprise network (This limitation is directed to a particular type or source of data, which is field of use. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(h)).
Claim 6 recites the following additional elements:
wherein the at least one instruction is a user input in a natural language format, and further comprising (This limitation is directed to a particular type or source of data, which is field of use. This does not amount to significantly more than judicial exception. See MPEP 2106.05(h)):
using the artificial intelligence model and based on a plurality of regular expression features (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This does not amount to significantly more than judicial exception, see MPEP 2106.05 (h)),
wherein the plurality of regular expression features are indicative of the information about the plurality of enterprise network assets and the configuration of the enterprise network (This limitation is directed to a particular type or source of data, which is field of use. This does not amount to significantly more than judicial exception. See MPEP 2106.05(h)).
9. Dependent claim 7 is directed to a method and falls into one of the four statutory categories.
Claim 7 recites the following abstract ideas:
wherein generating the at least one regular expression includes generating a sequence of a plurality of regular expression signatures … and based on a mapping of the user input to the plurality of feature embeddings and ordering the plurality of feature embeddings (Mental process directed to generating a sequence of a plurality of regular expression signatures which can be done by making a judgment based on mapping of the user to the feature embedding to generate regular expression).
Claim 7 recites the following additional elements:
using the artificial intelligence model (this limitation is directed to mere instructions to apply a judicial exception. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(f))
Claim 7 recites the following additional elements:
using the artificial intelligence model (this limitation is directed to mere instructions to apply a judicial exception. This does not amount to significantly more than judicial exception. See MPEP 2106.05(f))
10. Dependent claim 8 is directed to a method and falls into one of the four statutory categories.
Claim 8 recites the following abstract ideas:
generating the one or more code snippets based on the raw context description (Mental process directed to generating the one or more code snippets which can be done by observing the raw context description and making a judgement on the generation of the code snippets).
Claim 8 recites the following additional elements:
wherein generating the one or more code snippets includes: obtaining a raw context description for each of the plurality of regular expression signatures in the sequence (this limitation is directed to insignificant extra solution activity of data transmission. This limitation does not integrate the abstract idea into a practical application. See MPEP 2106.05(g)); and
Claim 8 recites the following additional elements:
wherein generating the one or more code snippets includes: obtaining a raw context description for each of the plurality of regular expression signatures in the sequence (this limitation is directed to insignificant extra solution activity of data transmission and it is well understood routine and conventional. This does not amount to significantly more than judicial exception. See MPEP 2106.05(d)(II), example i); and
11. Dependent claim 9 is directed to a method and falls into one of the four statutory categories.
Claim 9 recites the following abstract ideas:
generating the at least one regular expression includes: generating a plurality of regular expressions based on a network issue (Mental process directed to generating a plurality of regular expressions based on a network issue which can be done by observing the network issue and making a judgement on the generation of the regular expression)
generating the at least one solution including a set of configuration actions to perform to fix the network issue and a corresponding code snippet for a configuration action in the set of configuration actions based on the plurality of regular expressions (Mental process directed to generating the at least one solution including a set of configuration actions to perform to fix the network issue and a corresponding code snippet for a configuration action which can be done by making a judgment on the generation of a solution that includes of configuration actions and a corresponding code snippet).
Claim 9 recites the following additional elements:
wherein the artificial intelligence model is a multi-task generative large language model (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not integrate the abstract idea into a practical application, see MPEP 2106.05 (h)), and
using the multi-task generative large language model (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This limitation does not integrate the abstract idea into a practical application, see MPEP 2106.05 (h)); and
Claim 9 recites the following additional elements:
wherein the artificial intelligence model is a multi-task generative large language model (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This does not amount to significantly more than judicial exception. see MPEP 2106.05 (h)), and
using the multi-task generative large language model (this limitation is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. This does not amount to significantly more than judicial exception. see MPEP 2106.05 (h)); and
12. Independent claim 10 is directed to an apparatus and falls into one of the four statutory categories.
With regards to claim 10, it is substantially similar to claim 1, and is rejected in the
same manner and reasoning applying.
Further, claim 10 recites “An apparatus comprising: a memory; a network interface configured to enable network communications; and a processor, wherein the processor is configured to perform operations comprising:” this limitation is directed to recitation of a high level generic computer component. This does not integrate the abstract idea into a practical application, nor amount to significantly more than judicial exception. See MPEP 2106.05(f)
13. Dependent claim 11 is directed to an apparatus and falls into one of the four statutory categories.
With regards to claim 11, it is substantially similar to claim 2, and is rejected in the
same manner and reasoning applying.
14. Dependent claim 12 is directed to an apparatus and falls into one of the four statutory categories.
With regards to claim 12, it is substantially similar to claim 3, and is rejected in the
same manner and reasoning applying.
15. Dependent claim 13 is directed to an apparatus and falls into one of the four statutory categories.
With regards to claim 13, it is substantially similar to claim 4, and is rejected in the
same manner and reasoning applying.
16. Dependent claim 14 is directed to an apparatus and falls into one of the four statutory categories.
With regards to claim 14, it is substantially similar to claim 5, and is rejected in the
same manner and reasoning applying.
17. Dependent claim 15 is directed to an apparatus and falls into one of the four statutory categories.
With regards to claim 15, it is substantially similar to claim 6, and is rejected in the
same manner and reasoning applying.
18. Dependent claim 16 is directed to an apparatus and falls into one of the four statutory categories.
With regards to claim 16, it is substantially similar to claim 7, and is rejected in the
same manner and reasoning applying.
19. Dependent claim 17 is directed to an apparatus and falls into one of the four statutory categories.
With regards to claim 17, it is substantially similar to claim 8, and is rejected in the
same manner and reasoning applying.
20. Independent claim 18 is directed to a machine and falls into one of the four statutory categories.
With regards to claim 18, it is substantially similar to claim 1, and is rejected in the
same manner and reasoning applying.
Further, claim 18 recites “One or more non-transitory computer readable storage media encoded with software comprising computer executable instructions that, when executed by a processor, cause the processor to perform a method including:” this limitation is directed to recitation of a high level generic computer component. This does not integrate the abstract idea into a practical application, nor amount to significantly more than judicial exception. See MPEP 2106.05(f)
21. Dependent claim 19 is directed to a machine and falls into one of the four statutory categories.
With regards to claim 19, it is substantially similar to claim 2, and is rejected in the
same manner and reasoning applying.
22. Dependent claim 20 is directed to a machine and falls into one of the four statutory categories.
With regards to claim 20, it is substantially similar to claim 3, and is rejected in the
same manner and reasoning applying.
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.
23. Claims 1, 5-9 are rejected under 35 U.S.C. 103 as being unpatentable over Fakhoury et al. ("Towards generating functionally correct code edits from natural language issue descriptions." arXiv:2304.03816v1 [cs.SE] 7 Apr 2023) in view of Chen et al. (“SEED: Domain-Specific Data Curation With Large Language Models”, arXiv:2310.00749v2 [cs.DB] 2 Dec 2023)
Regarding claim 1, Fakhoury teaches a computer-implemented method (More interestingly, the ChatGPT model gpt-3.5-turbo outperforms other models in terms of both the pass@1, pass@5, and pass@100 accuracy metrics, pg. 2, left col., third para.) comprising:
obtaining at least one instruction and information (The following code is buggy. The issue is: … Please provide a fixed version with minimal changes, Fig. 1, pg. 2; Given a code and instruction written in NL such as "Improve the runtime complexity of this function", pg. 3 right col., last sentence to pg. 4, left col., first sentence; Looking at the information contained in the issue descriptions for each of these examples, pg. 9, left col., last sentence) about a plurality of enterprise network assets (When asked to identify the lines of code where the bug
exists, pg. 9, left col., first para. The Examiner notes that according to the instant specification: “software 102(1) – 102(N) are resources or assets of an enterprise”[0040]) and configuration (Each bug in the Defects4J dataset contains a PRE_FIX_REVISION and POST_FIX_REVISION version that represents the buggy/fixed versions of the code respectively, pg. 3, left col., third para.) of an enterprise network (software development with cloud-hosted continuous
integration (CI) pipelines, pg. 1, right col., second para.) that includes the plurality of enterprise network assets (When asked to identify the lines of code where the bug exists, pg. 9, left col., first para. The Examiner notes that according to the instant specification: “software 102(1) – 102(N) are resources or assets of an enterprise” [0040]);
generating … an expression using an artificial intelligence model (LLM, Fig. 1, pg. 2) based on context description of the at least one instruction and the information (Chat-GPT’s conversational nature allows the natural the use of advanced prompt structures that maintain conversational context, like chain of thought reasoning and reasoning extraction. Therefore, we only use this prompt strategy with gpt-3.5-turbo, pg. 4, right col., first para.; The Examiner notes gpt-3.5-turbo is an artificial intelligence) about the plurality of enterprise network assets and the configuration (Each bug in the Defects4J dataset contains a PRE_FIX_REVISION and POST_FIX_REVISION version that represents the buggy/fixed versions of the code respectively, pg. 3, left col., third para.) of the enterprise network (software development with cloud-hosted continuous integration (CI) pipelines, pg. 1, right col., second para.);
generating at least one solution (candidate solutions, Fig. 1, pg. 2) for configuring at least one network asset of the plurality of enterprise network assets (When asked to identify the lines of code where the bug exists, pg. 9, left col., first para.)
and providing the at least one solution (candidate solutions, Fig. 1, pg. 2) to cause a configuration change in the at least one network asset (For each setting we generate 100 candidate fixes for each of the 283 bugs and evaluate the correctness of each candidate against the trigger and relevant tests, pg. 5, left col., first para., Fig. 1; translating natural language descriptions of code changes (namely bug fixes described in Issue reports in repositories) into correct code fixes, abstract).
Fakhoury is silent about generating at least one regular expression and generating at least one solution for configuring at least one network asset of the plurality of enterprise network assets based on the at least one regular expression;
Chen teaches generating at least one regular expression using artificial intelligence (For example, code or small model modules are synthesized by the LLM to provide a domain-specific solution (e.g., a regular expression for extracting monetary amounts), pg. 1, right col., last para.) based on context description of the at least one instruction and the information (For each request, the LLM module may further employ a data access module that retrieves relevant information from a database or other user-supplied data to assist the LLM in solving the task, pg. 1, right col., last para.).
generating at least one solution for configuring at least one network asset of the plurality of enterprise network assets (SEED leverages LLMs’ synthesis, reasoning, semantics understanding abilities as well as the encoded common knowledge to construct a domain-specific solution, (pg. 2, left col., first para.) based on the at least one regular expression (… such as domain-specific code (like the regex…), pg. 1, left col., last sentence to pg. 1 right col., first sentence);
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Fakhoury to incorporate the teachings of Chen for the benefit of generic LLM tool to automatically synthesize domain-specific data curation solutions (Chen, pg. 12, left col., second to the last para.)
Regarding claim 5, Fakhoury and Chen teaches the computer-implemented method of claim 1, Fakhoury teaches wherein the artificial intelligence model is a generative large language model (LLM, Fig. 1, pg. 2), and
generating the at least one solution includes: generating one or more code snippets (GPT-3.5 model, which is further finetuned using Reinforcement Learning with Human Feedback (RLHF) … early evaluation has demonstrated strong capabilities … including understanding and generating code snippets, pg. 4, left col., second para.) by inputting the information and the at least one instruction (The following code is buggy. The issue is: … Please provide a fixed version with minimal changes, Fig. 1, pg. 2; Given a code and instruction written in NL such as "Improve the runtime complexity of this function", pg. 3, right col., last sentence in last para. to pg. 4, left col.; Looking at the information contained in the issue descriptions for each of these examples, pg. 9, left col., last sentence in last para. to pg. 9, right col.,) into the generative large language model (LLM, Fig. 1); and
Chen teaches executing the one or more code snippets (This prompt is then sent to an LLM to generate a series of code snippets, which are automatically evaluated and refined, pg. 3, left col., last sentence) to configure the at least one network asset (Code 1 and Code 2 are configured, see Fig. 3, pg. 5).
The same motivation to combine independent claim 1 applies here.
Regarding claim 6, Fakhoury and Chen teaches the computer-implemented method of claim 5, Chen wherein the at least one instruction is a user input in a natural language format (describing the task, inputs, … in natural language, pg. 2, right col., second para.), and further comprising:
mapping the user input to a plurality of feature embeddings using the artificial intelligence model (Given a set of inputs 𝑋 with corresponding vector embeddings 𝑉 = {𝑣𝑥 |𝑥 ∈ 𝑋} where each vector 𝑣𝑥 = C𝜃 (𝑥) ∈ R𝑑 , and a distance threshold 𝑑, pg. 7, left col., second to the last para.) and based on a plurality of regular expression features (For example, code or small model modules are synthesized by the LLM to provide a domain-specific solution (e.g., a regular expression for extracting monetary amounts), pg. 1, right col., last para.),
wherein the plurality of regular expression features are indicative of the information about the plurality of enterprise network assets (For example, for the task of data extraction, extracting monetary amounts can be effectively done by a regular expression such as that searches for a dollar sign followed by digits separated by commas and periods, i.e., "$\d[\d|,|.]*", while extracting human names requires a totally different method such as searching for capitalized words near salutations like “Mr.”or“Ms.”, pg. 1, left col., last para.) and the configuration of the enterprise network (JSON config, Fig. 1, pg. 3).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Fakhoury to incorporate the teachings of Chen for the benefit of generic LLM tool to automatically synthesize domain-specific data curation solutions (Chen, pg. 12, left col., second to the last para.)
Regarding claim 7, Fakhoury and Chen teaches the computer-implemented method of claim 6, Chen teaches wherein generating the at least one regular expression (For example, for the task of data extraction, extracting monetary amounts can be effectively done by a regular expression such as that searches for a dollar sign followed by digits separated by commas and periods, pg. 1, left col., last para.) includes generating a sequence of a plurality of regular expression signatures ("$\d[\d|,|.]*", pg. 1, left col., last para.) using the artificial intelligence model and based on a mapping of the user input to the plurality of feature embeddings ((Given a set of inputs 𝑋 with corresponding vector embeddings 𝑉 = {𝑣𝑥 |𝑥 ∈ 𝑋} where each vector 𝑣𝑥 = C𝜃 (𝑥) ∈ R𝑑 , and a distance threshold 𝑑, pg. 7, left col., second to the last para.) and ordering the plurality of feature embeddings (Next, we introduce our embedding-based index, pg. 6, right col., last para.).
The same motivation to combine dependent claim 6 applies here.
Regarding claim 8, Fakhoury and Chen teaches the computer-implemented method of claim 7, Chen teaches wherein generating the one or more code snippets includes: obtaining a raw context description for each of the plurality of regular expression signatures in the sequence (searches for a dollar sign followed by digits separated by commas and periods, i.e., "$\d[\d|,|.]*", pg. 1, left col., last para.); and
generating the one or more code snippets based on the raw context description (Example #1: a code snippet generated for the Buy dataset with LLM-generated advice, pg. 17, right col., last para.).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Fakhoury to incorporate the teachings of Chen for the benefit of generic LLM tool to automatically synthesize domain-specific data curation solutions (Chen, pg. 12, left col., second to the last para.)
Regarding claim 9, Fakhoury and Chen teaches the computer-implemented method of claim 1, Fakhoury teaches wherein the artificial intelligence model is a multi-task generative large language model (ChatGPT’s conversational nature allows it to excel in tasks that require both code generation and human-like interactions, allowing the use of advanced prompt structures that involve chain of thought[57] and reasoning extraction, pg. 4, left col., second para.), and
Chen teaches generating the at least one regular expression (For example, for the task of data extraction, extracting monetary amounts can be
effectively done by a regular expression such as that searches for a dollar sign followed by digits separated by commas and periods, i.e., "$\d[\d|,|.]*", while extracting human names requires a totally different method such as searching for capitalized words near salutations like “Mr.” or “Ms.”, pg. 1, left col., last para.) includes: generating a plurality of regular expressions based on a network issue using the multi-task generative large language model (In this way, SEED leverages LLMs’ synthesis, reasoning, semantics understanding abilities as well as the encoded common knowledge to construct a domain-specific solution, pg. 2, left col., first para.); and
generating the at least one solution including a set of configuration actions to perform to fix the network issue (SEED is able to produce efficient and effective solutions for multiple data curation problems, demonstrating its potential in practice, pg. 2, left col., fifth para.; JSON config, pg. 3, Fig. 1) and
a corresponding code snippet for a configuration action in the set of configuration actions based on the plurality of regular expressions (For plans that use code module, SEED first translates the config file into a task description prompt using a pre-defined template. This prompt is then sent to an LLM to generate a series of code snippets, pg. 3, left col., last para.).
The same motivation to combine independent claim 1 applies here.
24. Claims 2-4 are rejected under 35 U.S.C. 103 as being unpatentable over Fakhoury et al. ("Towards generating functionally correct code edits from natural language issue descriptions." arXiv:2304.03816v1 [cs.SE] 7 Apr 2023) in view of Chen et al. (“SEED: Domain-Specific Data Curation With Large Language Models”, arXiv:2310.00749v2 [cs.DB] 2 Dec 2023) and further in view of Singh et al. (US12222911 filed 09/28/2023)
Regarding claim 2, Fakhoury and Chen teaches the computer-implemented method of claim 1, Fakhoury teaches further comprising: training the artificial intelligence model (LLM, Fig. 1, pg. 2; Generative Pre-trained Transformers (GPT) are large-scale autoregressive … generation models trained to predict the next token given a natural language prefix (prompt) context, pg. 3, right col., second para.) to generate … the expression and corresponding plurality of solutions (candidate solutions, Fig. 1, pg. 2) as ground truths (Ground truth fix, Fig. 1, pg. 2) using a reinforcement learning edit distance score feedback loop (ChatGPT model (gpt-3.5-turbo) is based on the pretrained GPT-3.5 model, which is further finetuned using Reinforcement Learning with Human Feedback (RLHF), pg. 4, left col., second para.; We select the example to be an issue where the example buggy code is closest to the target buggy code, using standard edit distance metric, pg. 4, left col., last para.).
Chen teaches generating at least one regular expression (For example, code or small model modules are synthesized by the LLM to provide a domain-specific solution (e.g., a regular expression for extracting monetary amounts), pg. 1, right col., last para.).
The same motivation to combine independent claim 1 applies here.
They are silent about learning a plurality of regular expressions.
Singh teaches training the artificial intelligence model (set of values that are used to train the pattern learner 304 (col., 13, line 13); the pattern learner includes a language model interface 324 which upon execution by the processor set accesses a language model 208, col., 10, lines 36-38)
learning a plurality of regular expressions (uses an LLM 208 to identify 406 and mask 404 semantic substrings, allowing a regular-expression-based pattern learner 304 to capture strings having both syntactic and semantic substrings, col., 17, lines 19-22).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Fakhoury and Chen to incorporate the teachings of Singh for the benefit of learning patterns for detecting syntactic errors in strings that contain semantic substrings (Singh, col. 6, lines 45-46)
Regarding claim 3, Fakhoury, Chen and Singh computer-implemented method of claim 2, Fakhoury teaches further comprising: configuring the at least one network asset based on the at least one solution (Each bug in the Defects4J dataset contains a PRE_FIX_REVISION and POST_FIX_REVISION version that represents … reflect the actual state of the project when the bug was discovered/-fixed, pg. 4, right col., third para.),
wherein training the artificial intelligence model includes: tuning the artificial intelligence model using a reinforcement learning feedback loop based on configuring the at least one network asset (ChatGPT model (gpt-3.5-turbo) is based on the pretrained GPT-3.5 model, which is further finetuned using Reinforcement Learning with Human Feedback (RLHF), pg. 4, left col., second para.).
Regarding claim 4, Fakhoury, Chen and Singh the computer-implemented method of claim 3, Fakhoury teaches wherein tuning the artificial intelligence model (ChatGPT model (gpt-3.5-turbo) is based on the pretrained GPT-3.5 model, which is further finetuned using Reinforcement Learning with Human Feedback (RLHF), pg. 4, left col., second para.) includes: validating the at least one solution based on whether a network issue is resolved (Experiments are run in two phases: fix generation and fix validation, pg. 4, right col., second para.);
Singh teaches generating a feedback score for the at least one solution (some pattern learners utilize feedback received via a user interface 306 repair engine (col., 42, lines 31-33); This functionality has the technical benefit of learning patterns based on correct values, and of identifying incorrect values, col. 6, lines 52-54),
wherein the feedback score is positive based on the at least one solution being validated (learning 402 at least one non-negative-outcome-based pattern 206
which is based solely on string data values which are not associated with any negative outcome, col. 13, lines 66-67 to col. 14, line 1) and is negative based on the at least one solution not being validated (FIG. 22 shows that while removing either learned concretization or ranking had a negative impact on performance, col. 31, lines 22-24); and
providing the feedback score to the artificial intelligence model (In a variation, a language model is trained on … repair accuracies (based, e.g., on feedback from user interaction with nominally repaired values), col. 24, lines 45-48).
The same motivation to combine dependent claim 2 applies here.
25. Claims 10, 14-18 are rejected under 35 U.S.C. 103 as being unpatentable over Fakhoury et al. ("Towards generating functionally correct code edits from natural language issue descriptions." arXiv:2304.03816v1 [cs.SE] 7 Apr 2023) in view of Chen et al. (“SEED: Domain-Specific Data Curation With Large Language Models”, arXiv:2310.00749v2 [cs.DB] 2 Dec 2023) and further in view of
Groenewegen et al (US20240069907 filed 08/24/2022)
Regarding claim 10, claim 10 is similar to claim 1. It is rejected in the same manner and reasoning applying.
Fakhoury and Chen does not explicitly teach an apparatus comprising: a memory; a network interface configured to enable network communications; and a processor, wherein the processor is configured to perform operations comprising:
Groenewegen teaches an apparatus comprising: a memory; a network interface configured to enable network communications; and a processor, wherein the processor is configured to perform operations comprising (In addition to processors 110 (e.g., CPUs, ALUs, FPUs, TPUs, GPUs, and/or quantum processors), memory/storage media 112, …, wired and wireless network interface cards [0040]; the system includes multiple computers connected by a wired and/or wireless network 108. Networking interface equipment 128 can provide access to networks 108, using network components such as a packet-switched network interface card [0041]):
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Fakhoury and Chen to incorporate the teachings of Groenewegen for the benefit of helping a developer assess the likely correctness of the generated code (Groenewegen [0085])
Regarding claim 14, claim 14 is similar to claim 5. It is rejected in the same manner and reasoning applying.
Regarding claim 15, claim 15 is similar to claim 6. It is rejected in the same manner and reasoning applying.
Regarding claim 16, claim 16 is similar to claim 7. It is rejected in the same manner and reasoning applying.
Regarding claim 17, claim 17 is similar to claim 8. It is rejected in the same manner and reasoning applying.
Regarding claim 18, claim 18 is similar to claim 1. It is rejected in the same manner and reasoning applying.
Fakhoury and Chen does not explicitly teach one or more non-transitory computer readable storage media encoded with software comprising computer executable instructions that, when executed by a processor, cause the processor to perform a method including:
Groenewegen teaches one or more non-transitory computer readable storage media encoded with software comprising computer executable instructions that, when executed by a processor, cause the processor to perform a method including (a memory or other computer-readable storage medium is not a propagating signal [0200]; in the claim, “computer readable medium” means a computer readable storage medium, not a propagating signal per se [0201]):
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Fakhoury and Chen to incorporate the teachings of Groenewegen for the benefit of helping a developer assess the likely correctness of the generated code (Groenewegen [0085])
26. Claims 11-13, 19 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Fakhoury et al. ("Towards generating functionally correct code edits from natural language issue descriptions." arXiv:2304.03816v1 [cs.SE] 7 Apr 2023) in view of Chen et al. (“SEED: Domain-Specific Data Curation With Large Language Models”, arXiv:2310.00749v2 [cs.DB] 2 Dec 2023) in view of Groenewegen et al (US20240069907 filed 08/24/2022) and further in view of Singh et al. (US12222911 filed 09/28/2023)
Regarding claim 11, claim 11 is similar to claim 2. It is rejected in the same manner and reasoning applying.
Regarding claim 12, claim 12 is similar to claim 3. It is rejected in the same manner and reasoning applying.
Regarding claim 13, claim 13 is similar to claim 4. It is rejected in the same manner and reasoning applying.
Regarding claim 19, claim 19 is similar to claim 2. It is rejected in the same manner and reasoning applying.
Regarding claim 20, claim 20 is similar to claim 3. It is rejected in the same manner and reasoning applying.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MORIAM MOSUNMOLA GODO whose telephone number is (571)272-8670. The examiner can normally be reached Monday-Friday 8:00am-5:00pm EST.
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 T. Bechtold can be reached on (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.
/M.G./Examiner, Art Unit 2148
Prosecution Insights