Detailed Action
This Office Action is in response to the remarks entered on 10/01/2025. Claims 3, 11 and 19 have been canceled. Claims 1-2, 4-10, 12-18 and 20 are currently pending.
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 statement (IDS) submitted on 4/3/2025 is 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 1-2, 4-10, 12-18 and 20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
MPEP 2106 (III) sets out steps for evaluating whether a claim is drawn to patent-eligible subject matter. The analysis of claims 1-20, in accordance with these steps, follows.
Step 1 Analysis:
Claims 1-2 and 4-8 are directed to a method, claims 9-10 and 12-16 are directed to a system comprising of hardware processor and memory, claims 17-18 and 20 are directed to computer program products. Therefore, claims 1-2, 4-10, 12-18 and 20 fall into one of four statutory categories (i.e., process, machine, article of manufacture).
Claim 1,
Step 2A Prong 1: This claim recites the following abstract ideas:
parsing the output from a generative model into chunks to be attributed to one or more source passages, wherein the output includes sentences (this limitation is interpreted as a mental process with the aid of pen and paper as the human mind can parse output coming from a received information into chunks and associating the chunks to one or more source passages);
attributing at least a portion of the one or more source passages to the chunks based on a similarity threshold value (this limitation is interpreted as a mental process as the human mind can associate passages to the chunks that are similar based on a value);
combining the chunks with the source passages attributed to those chunks (this limitation is interpreted as a mental process as the human mind can combine chunks with passages together); and
generating a response to the prompt based on the combination, the response including the output with inline source identifiers that identify the attributed source passages (this limitation is interpreted as a mental process as the human mind can generate a response where the output to be identified with the source passage).
Step 2A Prong 2:
The following limitations are additional elements: receiving output from a generative artificial intelligence model processing a prompt, wherein the output includes sentences. This limitation covers receiving data, see MPEP 2106.05(d)(II)(i), which is extra solution activity as disclosed in MPEP 2106.05(g).Furthermore, the limitation retrieving the one or more source passages based on a similarity evaluation between the chunks and the one or more source passages. This limitation covers retrieving information See MPEP 2106.05(d)(II)(iv), which is extra solution activity as disclosed in MPEP 2106.05(g). These additional limitations do no integrate the judicial exception into a practical application.
Step 2B:
The limitation receiving output from a generative artificial intelligence model processing a prompt, wherein the output includes sentences and retrieving the one or more source passages based on a similarity evaluation between the chunks and the one or more source passages do not amount to significantly more than a judicial exception and are well-understood, routine, conventional activities according to MPEP 2106.05(d)(II)(i) and MPEP 2106.05(d)(II)(iv).
Claim 9, this claim is directed to a system claim that corresponds to method claim 1. See rejection for claim 1 above which also applies to claim 10. In addition, claim 9 recites the additional limitation of a system having a memory, instructions, and one or more processors for executing the instructions. This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process the data elements, see MPEP 2106.05(f). The additional limitations do not integrate the judicial exception into a practical application and do not amount to significantly more than the judicial exception.
Claim 17, this claim directed to a program product claim that corresponds to method claim 1. See the rejection for claim 1 above, which also applies to claim 16. In addition, claim 16 recites the additional limitation of a computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by one or more processors to cause the one or more processors to perform operations. This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process the data elements, see MPEP 2106.05(f). The additional limitations do not integrate the judicial exception into a practical application and do not amount to significantly more than the judicial exception.
Claim 2, 10, 18, the limitation, wherein the generative artificial intelligence model employs a generative adversarial network, a variational autoencoder, an autoregressive model, or a recurrent neural network, further limits the additional element, “generative artificial intelligence model”. Therefore it is directed to data transmission and does not integrate the judicial exception into a practical application and does not amount to significantly more than the judicial exception. See MPEP 2106.05(d)(II)(i), which describes data transmission as well-understood, routine, conventional activity. Additionally, the limitation can be directed to mere instructions to apply an exception using generic computer components MPEP 2106.05(f).
Claim 4, 12, 20, the limitation wherein the similarity evaluation includes a similarity evaluation that calculates similarity scores associated with the chunks and the source passages. The limitation further limits the similarity evaluation which is interpreted as a mental process. Claims 4, 12, and 20 do not recite any additional elements.
Claims 5 and 13, the limitation wherein the response is generated by a generative artificial intelligence model is an additional element which is directed to mere instruction to apply the abstract idea on a generic computer to process the data elements, see MPEP 2106.05(f). The additional limitations do not integrate the judicial exception into a practical application and do not amount to significantly more than the judicial exception.
Claims 6 and 14, the limitation, filtering the one or more source passages based on the similarity scores is a mental process. Claims 6 and 14 do not recite any additional elements.
Claims 7 and 15, generating an information graph wherein the information graph describes relationships between source passages and one or more other classes, the other classes including source images, source tables, and source code, is a mental process because it is taking data to generate graph. Furthermore the limitation “for each data record in a set of data records including the one or more source passages” is an additional element which is being applied to the abstract idea. Therefore, the additional element does not integrate the judicial exception into a practical application and does not amount to significantly more than the judicial exception. See MPEP 2106.05(f).
Claims 8 and 16, the limitation, retrieving the one or more source passages is based on the information graph, is an additional element. This limitation covers retrieving information See MPEP 2106.05(d)(II)(iv) which is extra solution activity as disclosed in MPEP 2106.05(g). The additional limitation does not integrate the judicial exception into a practical application.
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.
Claims 1-2, 4-10, 12-18 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Tunstall-Pedoe et al (Pub. No. US20230274094) in view of Peng et al. (“Check Your Facts and Try again Improving Large Language Models with External Knowledge and Automated Feedback”).
Claim 1: Tunstall-Pedoe teaches a method comprising:
receiving output (LLM generated continuation output is provided to a processing system; fig 10) from a generative artificial intelligence model processing a prompt, wherein the output includes sentences (i.e. the processing system is configured to analyze the continuation output (e.g. text output) generated by the LLM in response to a prompt; para. [0023]. The sentences are split into words (or other atomic parts of the language; para. [0497]);
parsing the output from a generative model into chunks (i.e. The continuation data generated by the LLM (i.e., the generative model) is provided to a processing system. [0024] and [0145] break down the structure of the given sentence and compare it to the structure of each of the known ground truth translations to sort by similarity. The sentences are split into words (or other atomic parts of the language) and then re-merged into subparts (sequences of words); para. [0497]. Word embeddings such as word2vec tor GloVe (processing system) is used to analyze the output texts [0500]) to be attributed to one or more source passages, wherein each chunk includes one or more of the sentences (i.e. the entity resolver currently works by comparing large numbers of passages that the two nodes are used in and determining how similarly they are used; para. [0498]; note regarding two node see example in [0497]. The sentences are split into words (or other atomic parts of the language) and then re-merged into a plurality of subparts (sequence of words) prior to the processing; para. [0497]);
Tunstall-Pedoe further teaches a similarity evaluation between the chunks and the one or more source passages (i.e. the entity resolver currently works by comparing large numbers of passages that the two nodes are used in and determining how similarly they are used; para. [0498]; note regarding two node see example in [0497]).
Tunstall-Pedoe does not explicitly teach retrieving the one or more source passages based on a similarity evaluation between the chunks and the one or more source passages attributing at least a portion of the one or more source passages to the chunks based on a similarity threshold value; combining the chunks with the source passages attributed to those chunks; and generating a response to the prompt based on the combination, the response including the output with inline source identifiers that identify the attributed source passages.
However, Peng teaches retrieving the one or more source passages (i.e. BM25-based retriever over the knowledge bases of FAQs and Yelp reviews, see section 3.2) based on a similarity evaluation between the chunks and the one or more source passages (i.e. To evaluate the degree to which the generated responses are grounded in consolidated evidence, we use the utility score, Knowledge F1 to measure the overlap between a prediction and the evidence which is either consolidated by knowledge consolidator. See Section 3.2 Utility); attributing at least a portion of the one or more source passages to the chunks based on a similarity threshold value (i.e. preferring responses with higher token overlap similar to section 3, see section 4.2; only kept examples with an F1 score higher than a certain threshold see section 3.1); combining the chunks with the source passages attributed to those chunks (i.e. DPR both question and passage/able inputs are represented by corresponding special tokens; see section 4.2); and generating a response to the prompt based on the combination, the response including the output with inline source identifiers that identify the attributed source passages (i.e. see fig. 1- consolidating evidences from external knowledge from the LLM to generate responses grounded in evidence and revising LLM’s candidate responses using automated feedback).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the invention of Tunstall-Pedoe to include external knowledge as part of an LLM prompts to help generate more responses that are more grounded in external knowledge relevant to the current conversation (see section 7 conclusion of Peng).
Claim 2: Tunstall-Pedoe and Peng teach the method of claim 1. Tunstall-Pedoe further teaches wherein the generative artificial intelligence model employs a generative adversarial network, a variational autoencoder, an autoregressive model, or a recurrent neural network ( i.e. Large language models (LLMs) are neural architectures; para. [0828]; the neural architecture utilises recurrent neural networks; para [1566]).
Claim 4: Tunstall-Pedoe and Peng teach the method of claim 1. Tunstall-Pedoe further teaches wherein the similarity evaluation includes a similarity evaluation that calculates similarity scores associated with the chunks and the source passages (i.e. When translating from natural language into UL, we break down the structure of the given sentence and compare it to the structure of each of the known ground truth translations to sort by similarity. The sentences are split into words (or other atomic parts of the language) and then re-merged into subparts (sequences of words) that we have an existing translation for such as (ExpressedInEnglish Camembert “Camembert”) and (ExpressedInEnglish IsA “is a”). These two passages would mean that the Camembert node becomes an option for the “Camembert” part of the sentence and IsA becomes an option for the “is a” part. When “Camembert is a creamy French cheese” is matched against GroundTruthTranslation1, the translator will give a high similarity score because most of the sentence is the same and the only part that is different (“Camembert”) has the same part of speech as “Brie” and has an option (Camembert) in the list which is very “similar” to the node used in GroundTruthTranslation1 which was Brie. In a preferred example, the similarity of these two nodes is compared using a component of the UL platform called the entity resolver; para. [0497]).
Claim 5: Tunstall-Pedoe and Peng teach the method of claim 1. Tunstall-Pedoe further teaches wherein the response is generated by a generative artificial intelligence model (i.e. fig. 10; generated by the LLM in response to a prompt to enable an improved version of the continuation output to be provided to a user).
Claim 6: Tunstall-Pedoe and Peng teach the method of claim 1. Peng further teaches further comprising filtering the one or more source passages based on the similarity scores (i.e. preferring responses with higher token overlap similar to section 3, see section 4.2; only kept examples with an F1 score higher than a certain threshold see section 3.1).
Claim 7: Tunstall-Pedoe and Peng teach the method of claim 6. Tunstall further teaches further comprising: generating an information graph for each data record in a set of data records including the one or more source passages (i.e. can create a “semantic template” by mapping a UL passage with a number of variables onto a semantic graph of similar shape; para. [0477], wherein the information graph describes relationships between source passages and one or more other classes, the other classes including source images, source tables, and source code, generating a semantic graph from the natural language [0479] 2. attempting to match it against templates with a graph of similar shape, that would already have associated UL, and finally [0480] 3. applying direct translation to only those nodes in the graph that match variables in the UL para [0477]).
Claim 8: Tunstall-Pedoe and Peng teach the method of claim 7. Tunstall-Pedoe further teaches wherein retrieving the one or more source passages is based on the information graph (i.e. generating a semantic graph from the natural language [0479] 2. attempting to match it against templates with a graph of similar shape, that would already have associated UL; para. [0477]).
Claim 9: Tunstall- Pedoe teaches a system comprising:
one or more processors; and memory storing instructions that, when executed by the one or more processors (i.e. para. 0119]), cause the system to perform:
receiving output (LLM generated continuation output is provided to a processing system; fig 10) from a generative artificial intelligence model processing a prompt, wherein the output includes sentences (i.e. the processing system is configured to analyze the continuation output (e.g. text output) generated by the LLM in response to a prompt; para. [0023], [0060] and [0066]. The sentences are split into words (or other atomic parts of the language; para. [0497]);
parsing the output from a generative model into chunks (i.e. break down the structure of the given sentence and compare it to the structure of each of the known ground truth translations to sort by similarity. The sentences are split into words (or other atomic parts of the language) and then re-merged into subparts (sequences of words); para. [0497]) to be attributed to one or more source passages, wherein each chunk includes one or more of the sentences (i.e. the entity resolver currently works by comparing large numbers of passages that the two nodes are used in and determining how similarly they are used; para. [0498]; note regarding two node see example in [0497]. The sentences are split into words (or other atomic parts of the language) and then re-merged into a plurality of subparts (sequence of words) prior to the processing; para. [0497]);
Tunstall-Pedoe further teaches a similarity evaluation between the chunks and the one or more source passages (i.e. the entity resolver currently works by comparing large numbers of passages that the two nodes are used in and determining how similarly they are used; para. [0498]; note regarding two node see example in [0497]).
Tunstall-Pedoe does not explicitly teach retrieving the one or more source passages based on a similarity evaluation between the chunks and the one or more source passages attributing at least a portion of the one or more source passages to the chunks based on a similarity threshold value; combining the chunks with the source passages attributed to those chunks; and generating a response to the prompt based on the combination, the response including the output with inline source identifiers that identify the attributed source passages.
However, Peng teaches retrieving the one or more source passages (i.e. BM25-based retriever over the knowledge bases of FAQs and Yelp reviews, see section 3.2) based on a similarity evaluation between the chunks and the one or more source passages (i.e. To evaluate the degree to which the generated responses are grounded in consolidated evidence, we use the utility score, Knowledge F1 to measure the overlap between a prediction and the evidence which is either consolidated by knowledge consolidator. See Section 3.2 Utility); attributing at least a portion of the one or more source passages to the chunks based on a similarity threshold value (i.e. preferring responses with higher token overlap similar to section 3, see section 4.2; only kept examples with an F1 score higher than a certain threshold see section 3.1); combining the chunks with the source passages attributed to those chunks (i.e. DPR both question and passage/able inputs are represented by corresponding special tokens; see section 4.2); and generating a response to the prompt based on the combination, the response including the output with inline source identifiers that identify the attributed source passages (i.e. see fig. 1- consolidating evidences from external knowledge from the LLM to generate responses grounded in evidence and revising LLM’s candidate responses using automated feedback).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the invention of Tunstall-Pedoe to include external knowledge as part of an LLM prompts to help generate more responses that are more grounded in external knowledge relevant to the current conversation (see section 7 conclusion of Peng).
Claim 10: Tunstall-Pedoe and Peng teach the system of claim 9. Tunstall-Pedoe further teaches wherein the generative artificial intelligence model employs a generative adversarial network, a variational autoencoder, an autoregressive model, or a recurrent neural network ( i.e. Large language models (LLMs) are neural architectures; para. [0828]; the neural architecture utilises recurrent neural networks; para [1566]).
Claim 12: Tunstall-Pedoe and Peng teach the system of claim 9. Tunstall-Pedoe further teaches wherein the similarity evaluation includes a similarity evaluation that calculates similarity scores associated with the chunks and the source passages (i.e. When translating from natural language into UL, we break down the structure of the given sentence and compare it to the structure of each of the known ground truth translations to sort by similarity. The sentences are split into words (or other atomic parts of the language) and then re-merged into subparts (sequences of words) that we have an existing translation for such as (ExpressedInEnglish Camembert “Camembert”) and (ExpressedInEnglish IsA “is a”). These two passages would mean that the Camembert node becomes an option for the “Camembert” part of the sentence and IsA becomes an option for the “is a” part. When “Camembert is a creamy French cheese” is matched against GroundTruthTranslation1, the translator will give a high similarity score because most of the sentence is the same and the only part that is different (“Camembert”) has the same part of speech as “Brie” and has an option (Camembert) in the list which is very “similar” to the node used in GroundTruthTranslation1 which was Brie. In a preferred example, the similarity of these two nodes is compared using a component of the UL platform called the entity resolver; para. [0497]).
Claim 13: Tunstall-Pedoe and Peng teach the system of claim 9. Tunstall-Pedoe further teaches wherein the response is generated by a generative artificial intelligence model (i.e. fig. 10; generated by the LLM in response to a prompt to enable an improved version of the continuation output to be provided to a user).
Claim 14: Tunstall-Pedoe and Peng teach the system of claim 12. Peng further teaches further comprising filtering the one or more source passages based on the similarity scores (i.e. preferring responses with higher token overlap similar to section 3, see section 4.2; only kept examples with an F1 score higher than a certain threshold see section 3.1).
Claim 15: Tunstall-Pedoe and Peng teach the system of claim 14. Tunstall further teaches further comprising: generating an information graph for each data record in a set of data records including the one or more source passages (i.e. can create a “semantic template” by mapping a UL passage with a number of variables onto a semantic graph of similar shape; para. [0477], wherein the information graph describes relationships between source passages and one or more other classes, the other classes including source images, source tables, and source code, generating a semantic graph from the natural language [0479] 2. attempting to match it against templates with a graph of similar shape, that would already have associated UL, and finally [0480] 3. applying direct translation to only those nodes in the graph that match variables in the UL para [0477]).
Claim 16: Tunstall-Pedoe and Peng teach the system of claim 15. Tunstall-Pedoe further teaches wherein retrieving the one or more source passages is based on the information graph (i.e. generating a semantic graph from the natural language [0479] 2. attempting to match it against templates with a graph of similar shape, that would already have associated UL; para. [0477]).
Claim 17: Tunstall-Pedoe teaches a non-transitory computer readable medium comprising instructions that, when executed, cause one or more processors to perform:
receiving output (LLM generated continuation output is provided to a processing system; fig 10) from a generative artificial intelligence model processing a prompt, wherein the output includes sentences (i.e. the processing system is configured to analyze the continuation output (e.g. text output) generated by the LLM in response to a prompt; para. [0023]. The sentences are split into words (or other atomic parts of the language; para. [0497]);
parsing the output from a generative model into chunks (i.e. break down the structure of the given sentence and compare it to the structure of each of the known ground truth translations to sort by similarity. The sentences are split into words (or other atomic parts of the language) and then re-merged into subparts (sequences of words); para. [0497]) to be attributed to one or more source passages, wherein each chunk includes one or more of the sentences (i.e. the entity resolver currently works by comparing large numbers of passages that the two nodes are used in and determining how similarly they are used; para. [0498]; note regarding two node see example in [0497]. The sentences are split into words (or other atomic parts of the language) and then re-merged into a plurality of subparts (sequence of words) prior to the processing; para. [0497]);
Tunstall-Pedoe further teaches a similarity evaluation between the chunks and the one or more source passages (i.e. the entity resolver currently works by comparing large numbers of passages that the two nodes are used in and determining how similarly they are used; para. [0498]; note regarding two node see example in [0497]).
Tunstall-Pedoe does not explicitly teach retrieving the one or more source passages based on a similarity evaluation between the chunks and the one or more source passages attributing at least a portion of the one or more source passages to the chunks based on a similarity threshold value; combining the chunks with the source passages attributed to those chunks; and generating a response to the prompt based on the combination, the response including the output with inline source identifiers that identify the attributed source passages.
However, Peng teaches retrieving the one or more source passages (i.e. BM25-based retriever over the knowledge bases of FAQs and Yelp reviews, see section 3.2) based on a similarity evaluation between the chunks and the one or more source passages (i.e. To evaluate the degree to which the generated responses are grounded in consolidated evidence, we use the utility score, Knowledge F1 to measure the overlap between a prediction and the evidence which is either consolidated by knowledge consolidator. See Section 3.2 Utility); attributing at least a portion of the one or more source passages to the chunks based on a similarity threshold value (i.e. preferring responses with higher token overlap similar to section 3, see section 4.2; only kept examples with an F1 score higher than a certain threshold see section 3.1); combining the chunks with the source passages attributed to those chunks (i.e. DPR both question and passage/able inputs are represented by corresponding special tokens; see section 4.2); and generating a response to the prompt based on the combination, the response including the output with inline source identifiers that identify the attributed source passages (i.e. see fig. 1- consolidating evidences from external knowledge from the LLM to generate responses grounded in evidence and revising LLM’s candidate responses using automated feedback).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the invention of Tunstall-Pedoe to include external knowledge as part of an LLM prompts to help generate more responses that are more grounded in external knowledge relevant to the current conversation (see section 7 conclusion of Peng).
Claim 18: Tunstall-Pedoe and Peng teach the non-transitory computer readable medium of claim 17. Tunstall-Pedoe further teaches wherein the generative artificial intelligence model employs a generative adversarial network, a variational autoencoder, an autoregressive model, or a recurrent neural network ( i.e. Large language models (LLMs) are neural architectures; para. [0828]; the neural architecture utilises recurrent neural networks; para [1566]).
Claim 20: Tunstall-Pedoe and Peng teach the non-transitory computer readable medium of claim 17. Tunstall-Pedoe further teaches wherein the similarity evaluation includes a similarity evaluation that calculates similarity scores associated with the chunks and the source passages (i.e. When translating from natural language into UL, we break down the structure of the given sentence and compare it to the structure of each of the known ground truth translations to sort by similarity. The sentences are split into words (or other atomic parts of the language) and then re-merged into subparts (sequences of words) that we have an existing translation for such as (ExpressedInEnglish Camembert “Camembert”) and (ExpressedInEnglish IsA “is a”). These two passages would mean that the Camembert node becomes an option for the “Camembert” part of the sentence and IsA becomes an option for the “is a” part. When “Camembert is a creamy French cheese” is matched against GroundTruthTranslation1, the translator will give a high similarity score because most of the sentence is the same and the only part that is different (“Camembert”) has the same part of speech as “Brie” and has an option (Camembert) in the list which is very “similar” to the node used in GroundTruthTranslation1 which was Brie. In a preferred example, the similarity of these two nodes is compared using a component of the UL platform called the entity resolver; para. [0497]).
Response to Arguments
Response to Arguments under 35 U.S.C. 101
Arguments: Applicant asserts that performing the parsing, attributing, combining, and generating as required by claim 1, cannot be performed in the human mind in the time available between receiving output from a generative artificial intelligence model processing a prompt and generating a response to the prompt.
Examiner’s Response: Examiner respectfully disagrees. First, ‘time available between receiving output from a generative artificial intelligence model’ is not claimed. Second, parsing the output from a model into chunks, attributing at least a portion of the source passages to the chunks based on the similarity threshold value (mental process of evaluation), combining the chunks, and generating a response to the prompt, do not require a computer component and can be done in one’s mind with the aid of pen and paper.
Accordingly, arguments regarding claim 1 are not persuasive. Similarly, arguments regarding claims 9 and 17 are not persuasive. Claims 2, 4-8, 10, 12-16, 18 and 20 depend from claims 1, 9 and 17 and rejected under the same rationale as of claims 1, 9 and 17.
Response to Arguments under 35 U.S.C. 103
Arguments: Applicant asserts that claim 1 as amended requires that each chunk includes one or more sentences, however Tunstall-Pedoe [0497-0498] describes breaking single sentences into words.
Examiner’s Response: Examiner respectfully disagrees. Tunstall-Pedoe [0497] discloses splitting the structure of the given sentence into words and then re-merges into subparts (sequences of words). Each subpart corresponds to each chunk and the sequence of words can be interpreted as the ‘sentence’. Additionally, [0060] and [0066] “(i) Receiving output from a natural language processing computer process, the output including an answer to a question;” further indicates that the LLM outputs a sentence.
Accordingly, arguments regarding claim 1 are not persuasive. Similarly, arguments regarding claims 9 and 17 are not persuasive. Claims 2, 4-8, 10, 12-16, 18 and 20 depend from claims 1, 9 and 17 and rejected under the same rationale as of claims 1, 9 and 17.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Liu et al., “A Token-level Reference-free Hallucination Detection Benchmark for Free-form Text Generation”, 2022 (This prior art is pertinent because it discloses prevention of hallucinations in Language Models)
THIS ACTION IS MADE FINAL. 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 JUN KWON whose telephone number is (571)272-2072. The examiner can normally be reached Monday – Friday 7:30AM – 4:30PM ET.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Abdullah Kawsar can be reached at (571)270-3169. 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.
/JUN KWON/Examiner, Art Unit 2127
/ABDULLAH AL KAWSAR/Supervisory Patent Examiner, Art Unit 2127
Prosecution Insights