Notice of Pre-AIA or AIA Status
1. 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 § 102
2. In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
3. Claims 1-2, 10-11 and 16-17 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Srinivasan (US 2024/0135187).
Regarding Claim 1:
Srinivasan discloses a method for generating a file (Srinivasan: ¶19 discloses a multi stage distillation method for training query processing models, including large language models, using training examples and model generated labels), comprising:
inputting M1 first-type files into a first model, and outputting second-type files corresponding to the respective first-type files from the first model (Srinivasan: ¶63 discloses a first plurality of training examples comprising respective search queries. ¶64 further discloses that the queries are included in augmented training examples used as inputs to a query-processing model, ¶65 discloses that first query processing model processes the augmented training examples to generate respective inferred labels 16 for those examples. The query containing training examples correspond to the M1 first type files and the respective inferred labels correspond to the second type files output by the first model);
determining a plurality of file pairs according to an output result, wherein each file pair comprises a first-type file and a second-type file corresponding to the first-type file (Srinivasan: ¶65 discloses generated inferred labels for the respective augmented training examples, ¶66 discloses using the first training examples and their inferred labels to train the second query processing model. Each corresponding training example and inferred label association is the claimed file pair and using the examples and their respective target labels necessarily identifies those associations. ¶74 clearly states the first model respectively generates a first plurality of inferred labels for the first plurality of augmented training examples. “Respectively” establishes a one-to-one correspondence);
adjusting a second model by using the plurality of file pairs (Srinivasan: ¶66 discloses training second query processing model 18 using the first set of training examples and corresponding inferred labels 16 so that second query processing model 18 learns to predict the inferred labels. Training the second model using the corresponding input examples and target labels is adjusting the second model using the plurality of file pairs); and
inputting M2 first-type files into an adjusted second model, and outputting the second-type files corresponding to the respective first-type files from the adjusted second model, wherein M1 and M2 are positive integers (Srinivasan: ¶67 discloses a second plurality of unlabeled training examples 20 comprising respective search queries. ¶68 further discloses that trained second query processing model 18 processes the second plurality of training examples 20 and generates respective inferred labels 22 for those training examples. The second plurality of query containing examples corresponds to the M2 first type files and the respective inferred labels correspond to the second type files output by the adjusted second model. Because the first and second sets each contain a plurality of training examples, their respective quantities M1 and M2 are necessarily training examples).
Regarding Claim 2:
Srinivasan further discloses the method of claim 1, wherein M2 is greater than M1 (Srinivasan: ¶67 discloses the second training data set is larger than the first set).
Regarding Claim 10:
Claim 10 has been analyzed with regard to claim 1 and is rejected for the same reasons of anticipation set forth above.
It is noted that Claim 10 recited the operations of claim 1 in electronic device form and additionally recited at least one process, a memory connected in communication with the processor and instructions stored in the memory and executable by the processor. Srinivasan discloses a computing system having one or more processors and non-transitory computer readable media storing instructions that cause the system to perform the previously mapped operations at least at ¶8 and ¶31.
Regarding Claim 11:
Claim 11 has been analyzed with regard to claim 2 and is rejected for the same reasons of anticipation set forth above.
Regarding Claim 16:
Claim 16 has been analyzed with regard to claim 1 and is rejected for the same reasons of anticipation set forth above.
Regarding Claim 17:
Claim 17 has been analyzed with regard to claim 2 and is rejected for the same reasons of anticipation set forth above.
Claim Rejections - 35 USC § 103
4. In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
5. Claims 3, 12 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Srinivasan in view of Esra (US 2024/0303347).
Regarding Claim 3:
Srinivasan further discloses the method of claim 1, except wherein the M2 first-type files comprise a file to be secured.
However, Esra discloses wherein the M2 first-type files comprise a file to be secured (¶53 discloses it protects sensitive information during data analysis and model training, ¶55 discloses protecting user data on the user device through encryption against unauthorized access or interception, ¶75 discloses that the trained local model analyzes newly obtained user data corresponding to the M2 first type files input after model adjustment).
Srinivasan and Esra are combinable because the references are in pertinent fields of endeavor, i.e., both disclose systems or methods for training machine-learning models and applying trained models to additional input data. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Srinivasan to configure the M2 first type field to include sensitive user-data files requiring security protection and to process those files locally using privacy preserving techniques as taught by Esra. This modification would have protected the privacy of the sensitive input files while permitting the adjusted second model to derive useful results from their contents because Esra explains that its privacy preserving techniques “protect user privacy while enabling meaningful insights to be derived from the data” in ¶53.
Regarding Claim 12:
Claim 12 has been analyzed with regard to claim 3 (see rejection above) and is rejected for the same reasons of obviousness set forth above.
Regarding Claim 18:
Claim 18 has been analyzed with regard to claim 3 (see rejection above) and is rejected for the same reasons of obviousness set forth above.
6. Claims 4-5, 13-14 and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Srinivasan and Wang (US 2022/0358851).
Regarding Claim 4:
Srinivasan further discloses the method of claim 1, except wherein the M1 first-type files comprise a chapter corpus and the second-type files comprise a dialogue corpus
However, Wang discloses:
wherein the M1 first-type files comprise a chapter corpus and the second-type files comprise a dialogue corpus (Wang: ¶27 discloses an uploaded text corpus and explicitly parses that corpus into chapters, ¶29 discloses using a question generation model to generate corresponding questions and answers for the complete corpus, ¶30 further discloses feeding those question answer pairs to a chatbot or conversational agent. This collection of conversation question answer pairs are a dialogue corpus).
Srinivasan and Wang are combinable because they are from the same field of endeavor, i.e., both disclose systems or methods for using machine-learning models to process text corpora and generate corresponding model outputs. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Srinivasan to configure the M1 first type as field chapter corpora obtained from book text and the corresponding second type files as dialogue corpora comprising generated question answer pairs, as taught by Wang. This modification would have enabled the efficient conversion of book or chapter content into conversational question answer material usable by a conversational agent because Wang explicitly states that “efficiency may be gained by providing a system that can parse the text of a storybook, or other document, and generate question answer pairs” in ¶14.
Regarding Claim 5:
The proposed combination of Srinivasan in view of Wang further discloses the method of claim 4, wherein inputting the M1 first-type files into the first model, and outputting the second-type files corresponding to the respective first-type files from the first model comprises:
inputting the chapter corpus into the first model, and outputting a plurality of dialogue questions associated with the chapter corpus from the first model (Wang: ¶26-29 discloses a deep learning architecture that receives book text or an e book, parses it into chapters, inputs the parsed corpus into its models and generates multiple questions associated with the corpus); and
outputting, from the first model, a corresponding dialogue corpus based on the chapter corpus and the plurality of dialogue questions, wherein the dialogue corpus comprises the plurality of dialogue questions and answers to the dialogue questions (Wang: ¶29 discloses using concepts extracted from the chapter corpus as answers , generating corresponding questions and combining both into a question answer pairs for the complete corpus, ¶30 outputs those pairs to a chatbot or conversational agent. The complete collection of conversational question answer pairs corresponds to the claimed dialogue corpus).
Srinivasan and Wang are combinable because they are from the same field of endeavor, i.e., both disclose systems or methods for using machine-learning models to process text corpora and generate corresponding model outputs. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Srinivasan to configure the M1 first type as field chapter corpora obtained from book text and the corresponding second type files as dialogue corpora comprising generated question answer pairs, as taught by Wang. This modification would have enabled the efficient conversion of book or chapter content into conversational question answer material usable by a conversational agent because Wang explicitly states that “efficiency may be gained by providing a system that can parse the text of a storybook, or other document, and generate question answer pairs” in ¶14.
Regarding Claim 13:
Claim 13 has been analyzed with regard to claim 4 (see rejection above) and is rejected for the same reasons of obviousness set forth above.
Regarding Claim 14:
Claim 14 has been analyzed with regard to claim 5 (see rejection above) and is rejected for the same reasons of obviousness set forth above.
Regarding Claim 19:
Claim 19 has been analyzed with regard to claim 4 (see rejection above) and is rejected for the same reasons of obviousness set forth above.
Regarding Claim 20:
Claim 20 has been analyzed with regard to claim 5 (see rejection above) and is rejected for the same reasons of obviousness set forth above.
7. Claims 6-8 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Srinivasan in view of Wang and further in view of Gelfenbeyn (US 2023/0351118).
Regarding Claim 6:
The proposed combination of Srinivasan in view of Wang further discloses the method of claim 4, except wherein inputting the M1 first-type files into the first model comprises:
generating a prompt for each chapter corpus and inputting the prompt into the first model,
wherein the prompt carries a content of the chapter corpus and identity characteristics of dialogue participants; and
the identity characteristics of the dialogue participants are configured to enable the first model to output a dialogue corpus that satisfies the identity characteristics of the dialogue participants.
Srinivasan and Wang do not explicitly disclose:
generating a prompt for each chapter corpus and inputting the prompt into the first model,
wherein the prompt carries a content of the chapter corpus and identity characteristics of dialogue participants; and
the identity characteristics of the dialogue participants are configured to enable the first model to output a dialogue corpus that satisfies the identity characteristics of the dialogue participants.
However Gelfenbeyn discloses:
generating a prompt for each chapter corpus and inputting the prompt into the first model (Gelfenbeyn: ¶68 discloses generating and forming dialogue prompts and providing those prompts to an LLM),
wherein the prompt carries a content of the chapter corpus and identity characteristics of dialogue participants (Gelfenbeyn: ¶31 discloses providing conversational context to the LLM as a natural language description identifying the participants such as “this is a conversation between Darth Vader and Luke Skywalker, ¶40 and 42 disclose character descriptions, motivations, memory and personality features and ¶50 discloses forming an LLM request based on the scene, environmental parameters, participants’ emotional states and conversational context. Combined with Wang’s chapter content, these teachings corresponding to a prompt carrying both the chapter corpus content and identity characteristics of the dialogue participants); and
the identity characteristics of the dialogue participants are configured to enable the first model to output a dialogue corpus that satisfies the identity characteristics of the dialogue participants (Gelfenbeyn: ¶28 explicitly discloses modifying requests and generated responses based on a character’s personality, role and emotional state, ¶78 further discloses that a character’s identity profile, including role and interests, influences how the character behaves).
Srinivasan, Wang and Gelfenbeyn are analogous art because they are in the same field of endeavor, i.e., both disclose systems or methods for using language models to generate conversational content. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Wang to generate, for each chapter corpus, a prompt carrying the chapter content and identity characteristics of the dialogue participants and to provide that prompt to the language model so that the resulting dialogue corpus reflects those identity characteristics, as taught by Gelfenbeyn. The modification would have produced more effective and appropriate dialogue responses consistent with the participant’s personalities and roles in the conversational context. Gelfenbeyn motivates this in ¶28, stating: “In order to obtain more effective and appropriate responses to user questions and messages, the platform may apply various restrictions, classifications, shortcuts, and filters in response to user questions. These targeted requests to the LLMs will result in optimized performance.”
Regarding Claim 7:
The proposed combination of Srinivasan, Wang and Gelfenbeyn further discloses the method of claim 6, wherein inputting the prompt into the first model comprises: optimizing the prompt by adopting a prompt optimization method (Gelfenbeyn: ¶28 discloses modifying a request before it is sent to the LLM by classifying and filtering the request and changing its words according to the character personalities and conversational context, explicitly stating that these target requests produce optimized performance); and
inputting an optimized prompt into the first model (Gelfenbeyn: ¶50 discloses forming the LLM request by classifying and adjusting its text according to character, scene, emotional state and conversational context parameters).
Srinivasan, Wang and Gelfenbeyn are analogous art because they are in the same field of endeavor, i.e., both disclose systems or methods for using language models to generate conversational content. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Wang to generate, for each chapter corpus, a prompt carrying the chapter content and identity characteristics of the dialogue participants and to provide that prompt to the language model so that the resulting dialogue corpus reflects those identity characteristics, as taught by Gelfenbeyn. The modification would have produced more effective and appropriate dialogue responses consistent with the participant’s personalities and roles in the conversational context. Gelfenbeyn motivates this in ¶28, stating: “In order to obtain more effective and appropriate responses to user questions and messages, the platform may apply various restrictions, classifications, shortcuts, and filters in response to user questions. These targeted requests to the LLMs will result in optimized performance.”
Regarding Claim 8:
The proposed combination of Srinivasan, Wang and Gelfenbeyn further discloses the method of claim 6, wherein the chapter corpus has a predefined expression style (Wang: ¶26 discloses receiving a completed book or e book as the input text corpus and ¶27 discloses parsing that corpus into chapters. A completed storybook necessarily has a preexisting narrative manner of expression before it is received and divided into chapters. Accordingly, each resulting chapter corpus has the predefined narrative expression of the input storybook).
Regarding Claim 15:
Claim 15 has been analyzed with regard to claim 6 (see rejection above) and is rejected for the same reasons of obviousness set forth above.
8. Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over Srinivasan in view of Wang, further in view of Gelfenbeyn and further in view of Gutierrez (US 2024/0420688).
Regarding Claim 9:
The proposed combination of Srinivasan in view of Wang and Gelfenbeyn further discloses the method of claim 8, except wherein determining the plurality of file pairs according to the output result comprises: removing any dialogue corpus which does not conform to the predefined expression style, and removing any dialogue corpus which is inconsistent with the identity characteristics of the dialogue participants from an output result of the first model to obtain a remaining output result; and determining the plurality of file pairs according to the remaining output result.
However, Gutierrez discloses:
wherein determining the plurality of file pairs according to the output result comprises:
removing any dialogue corpus which does not conform to the predefined expression style, and removing any dialogue corpus which is inconsistent with the identity characteristics of the dialogue participants from an output result of the first model to obtain a remaining output result (Gutierrez: ¶39 discloses evaluating linguistic data using tone classifications such as formal, formal, optimistic or grave and personality classifications. ¶45-46 disclose generating multiple natural language dialogue scripts according to linguistic characteristics associated with a particular personality type. ¶47 discloses validating each generated script against personality associated sample text and rejecting a script when its alignment score falls below a threshold. ¶48 discloses scoring multiple generated scripts and selecting the highest scoring script); and
determining the plurality of file pairs according to the remaining output result (Gutierrez: ¶47-48 discloses retaining the generated scripts that satisfy the alignment requirement).
Srinivasan, Wang, Gelfenbeyn and Gutierrez are combinable because they are from the same field of endeavor, i.e., both disclose systems or methods for generating conversational text through computational dialogue systems. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Srinivasan, Wang and Gelfenbeyn, to evaluate the generated dialogue corpora for conformity with the predefined expression style and dialogue-participant identity characteristics, remove non-conforming dialogue corpora and determine the file pairs from the remaining conforming output as taught by Gutierrez. The modification would have produced dialogue training pairs that more accurately and consistently reflect the intended style and participant identities because Gutierrez explains that personality tailored responses “results in increased harmony, realism, immersion, accuracy, and fidelity in communication, while reducing frustration” in ¶13.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to IAN SCOTT MCLEAN whose telephone number is (703)756-4599. The examiner can normally be reached "Monday - Friday 8:00-5:00 EST, off Every 2nd Friday".
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, Hai Phan can be reached at (571) 272-6338. 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.
/IAN SCOTT MCLEAN/Examiner, Art Unit 2654
/HAI PHAN/Supervisory Patent Examiner, Art Unit 2654
Prosecution Insights