DETAILED ACTION
This action is responsive to communications filed 11 July 2025.
Claims 1-20 are subject to examination.
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 11 July 2025 and 07 May 2026. The submission 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 1-20 rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Claims 1, 8 and 15 recite receive a conversation history from a messaging interface; extract relationship information involving one or more users based on the conversation history; generate, using a first large language model (LLM), a personalized prompt based on the relationship information, wherein, to generate the personalized prompt, the at least one processing device is configured to predict user interactions based on real-time chat context and historical interaction patterns using a dynamic pre-fetching model; and generate, using a second LLM, an initial reply suggestion based on the personalized prompt, wherein, to generate the initial reply suggestion, the at least one processing device is configured to: asynchronously transmit partial pre-processed data from the first LLM to the second LLM using a streaming interface; and generate the initial reply suggestion using a lightweight language model (LM) of a latency-aware fallback model in response to detecting of a processing delay of the second LLM.
The limitations of “receive a conversation history from a messaging interface; extract relationship information involving one or more users based on the conversation history; generate, using a first large language model (LLM), a personalized prompt based on the relationship information, wherein, to generate the personalized prompt, the at least one processing device is configured to predict user interactions based on real-time chat context and historical interaction patterns using a dynamic pre-fetching model; and generate, using a second LLM, an initial reply suggestion based on the personalized prompt, wherein, to generate the initial reply suggestion”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of certain methods of organizing human activity but for the recitation of generic computer components. That is, other than reciting “by a processor,” nothing in the claim element precludes the step from being certain methods of organizing human activity. For example, the process above where “receive” is to obtain messages, “extract” is to observe relationships between users, “generate” is to organize responses of users and “predict” is to estimate the responses of the users in dialogue. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Certain Methods of Organizing Human Activity” grouping of abstract ideas. Accordingly, the claim recites an abstract idea. See MPEP 2106.04(a)(2) II.
This judicial exception is not integrated into a practical application because in particular, the claims only recite the additional elements of using generic computing components and implemented in a generic computing environment, such as using large language models to perform the certain methods of organizing human activity. The computing components and computing environment are recited at a high-level of generality such that they amount to no more than mere instructions to apply the exception using generic computing components in a generic computing environment. Accordingly, the additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The claims are directed to an abstract idea.
The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements of having LLMs asynchronously transmit data between and using fallback models amounts to no more than mere instructions to apply the exception and adding insignificant extra-solution activity to the judicial exception. Mere instructions to apply an exception using generic computing components for certain methods of organizing human activity cannot provide an inventive concept. The claims are not patent eligible.
Claims 2-7, 9-14 and 16-20 recite anticipate user interaction based on contextual patterns in the conversation history; pre-process multimodal chat data into hierarchical structured representations; and generate partial structured data based on the anticipated user interaction and the hierarchical structured representations; and the second LLM is configured to: incrementally generate response actions based on the partial structured data received from the first LLM; and adapt generated response actions using additional structured data; generate, using a real-time feedback model, real-time feedback to adjust message priority based on an emotional tone of the response actions using an urgent score model; adjust a pre-fetching policy of the dynamic pre-fetching model based on an uncertainty score of a current response trajectory of the response actions from the second LLM; combine a placeholder suggestion with the response actions from the second LLM using an attention-weighted semantic alignment model; perform context-aware analysis, using an adaptive context-aware response timing model coupled to the first LLM, for each user of the one or more users based on the conversation history; and analysis of conversational flow, language nuances, and user intent.
The limitations of “anticipate user interaction based on contextual patterns in the conversation history; pre-process multimodal chat data into hierarchical structured representations; and generate partial structured data based on the anticipated user interaction and the hierarchical structured representations; and the second LLM is configured to: incrementally generate response actions based on the partial structured data received from the first LLM; and adapt generated response actions using additional structured data; generate, using a real-time feedback model, real-time feedback to adjust message priority based on an emotional tone of the response actions using an urgent score model; adjust a pre-fetching policy of the dynamic pre-fetching model based on an uncertainty score of a current response trajectory of the response actions from the second LLM; perform context-aware analysis, using an adaptive context-aware response timing model coupled to the first LLM, for each user of the one or more users based on the conversation history; and analysis of conversational flow, language nuances, and user intent”, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of certain methods of organizing human activity but for the recitation of generic computer components. That is, other than reciting “by a processor,” nothing in the claim element precludes the step from being certain methods of organizing human activity. For example, the process above where “anticipate”, “pro-process” and “generate” is to organize responses of users, “generate” and “adapt” is to estimate the responses of the users in dialogue in continual organization, “adjust” is to organize interactions based on priority and uncertainty, and “analysis” is the evaluation of the activity to result in organizing the activity. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Certain Methods of Organizing Human Activity” grouping of abstract ideas. Accordingly, the claim recites an abstract idea. See MPEP 2106.04(a)(2) II.
This judicial exception is not integrated into a practical application because in particular, the claims only recite the additional elements of using generic computing components and implemented in a generic computing environment, such as using large language models to perform the certain methods of organizing human activity and using an attention-weighted semantic alignment model. The computing components and computing environment are recited at a high-level of generality such that they amount to no more than mere instructions to apply the exception using generic computing components in a generic computing environment. Accordingly, the additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The claims are directed to an abstract idea.
The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements of having the latency-aware fallback model combine a placeholder with the response actions using an attention-weighted semantic alignment model amounts to no more than mere instructions to apply the exception and adding insignificant extra-solution activity to the judicial exception. Mere instructions to apply an exception using generic computing components for certain methods of organizing human activity cannot provide an inventive concept. The claims are not patent eligible.
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.
Claim(s) 1-2, 4, 6, 8-9, 11, 13, 15-16 and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Gelfenbeyn et al. (US-11960983-B1) hereinafter Gelfenbeyn in view of Sahu et al. (US-12475178-B2) hereinafter Sahu.
Regarding claim 1, Gelfenbeyn discloses:
An electronic device ([16:40-53] computer) comprising:
at least one processing device ([16:40-17:3] processor) configured to:
receive a conversation history from a messaging interface ([13:65-14:13] acquiring … a context of an interaction between a user and an AI character … based on historical data (i.e. must receive historical data to acquire the context thereafter));
extract relationship information involving one or more users based on the conversation history ([13:65-14:13] context of an interaction between a user and an AI character … based on historical data concerning interactions between the user and the AI character (i.e. relationship information from the historical information));
generate, using a first large language model (LLM) ([6:65-7:16] language model 304 can be based on an LLM … form a query for the LLM (i.e. language model 304 forms a query for the LLM)), a personalized prompt based on the relationship information ([14:14-28] anticipated words … based on the context (i.e. as above, e.g. relationship information) … generating … at least one query (i.e. prompt) … based on the one or more anticipated words), wherein, to generate the personalized prompt ([14:14-28] at least one query), the at least one processing device is configured to predict user interactions based on real-time chat context and historical interaction patterns using a dynamic pre-fetching model ([14:14-28] predicting … one or more anticipated words to be uttered by the user … based on the context … using a pretrained neural network (i.e. dynamic pre-fetching model)); and
generate, using a second LLM ([14:14-49] at least one query to an LLM (i.e. query sent to second LLM, e.g. using second LLM)), an initial reply suggestion based on the personalized prompt ([14:14-49] at least one query may be generated based on the one or more anticipated words … at least one query to the LLM (e.g. second LLM as above) to obtain at least one response from the LLM (i.e. response to query, an initial reply suggestion)), wherein, to generate the initial reply suggestion ([14:14-49] at least one response from the LLM (i.e. response to query)), the at least one processing device is configured to:
asynchronously transmit partial pre-processed data from the first LLM to the second LLM using a streaming interface ([5:3-11] allow responses from LLMs to be retrieved ahead of time and replies to a user to be prepared while the user is still speaking … start processing potential questions of the user even before the user starts speaking and any time before the user finishes speaking … allow pre-processing of data needed for generation of replies and provide lower latency of generation of replies to the user, see [7:45-59] pre-processing and transforming an incoming data stream (i.e. incoming stream requires a streaming interface, e.g. from model 304 above to the LLM)); and
Gelfenbeyn does not explicitly disclose:
generate the initial reply suggestion using a lightweight language model (LM) of a latency-aware fallback model in response to detecting of a processing delay of the second LLM.
However, Sahu discloses:
generate the initial reply suggestion using a lightweight language model (LM) of a latency-aware fallback model in response to detecting of a processing delay of the second LLM ([14:26-64] if a match is not found in the reformulated query cache … utilizing a lightweight language model to determine the reformulated autosuggest query on-the-fly … within a low-latency time threshold … by utilizing the reformulated query cache to identify reformulated autosuggest queries, the query gateway system does not have to reprocess the same queries again … utilizing an online lightweight language model … without sacrificing latency (i.e. latency-aware fallback by using a lightweight model)).
It would have been obvious to one of ordinary skill in the pertinent art before the effective filing date of the claimed invention to modify the invention of Gelfenbeyn in view of Sahu to have generated the initial reply suggestion using a lightweight language model (LM) of a latency-aware fallback model in response to detecting of a processing delay of the second LLM. One of ordinary skill in the art would have been motivated to do so to reduce wasted processing and improving efficiency without sacrificing latency (Sahu, [14:26-64]).
Regarding claim 2, Gelfenbeyn-Sahu disclose:
The electronic device of Claim 1, set forth above, wherein:
Gelfenbeyn discloses:
the first LLM ([6:65-7:16] language model 304 can be based on an LLM) is configured to:
anticipate user interaction based on contextual patterns in the conversation history ([14:14-28] context of an interaction between a user and an AI character … based on historical data concerning interactions between the user and the AI character (i.e. relationship information from the historical information … anticipated words … based on the context (i.e. as above, e.g. relationship information) … generating … at least one query (i.e. prompt) … based on the one or more anticipated words);
pre-process multimodal chat data into hierarchical structured representations ([8:20-33] pre-processing the incoming streams of data … differentially based on the specific modality … singular format … speech to text … object classification); and
generate partial structured data based on the anticipated user interaction and the hierarchical structured representations ([8:34-49] transformed into text and/or embeddings [15:57-62] one or more words received from the user may be used to predict the rest of the phrase of the user … phrase received from the user and the predicted remaining portion of the phrase (i.e. partial structured data)); and
the second LLM ([14:14-49] at least one query to an LLM (i.e. query sent to second LLM, e.g. using second LLM)) is configured to:
incrementally generate response actions based on the partial structured data received from the first LLM ([14:14-49] at least one query may be generated based on the one or more anticipated words … at least one query to the LLM (e.g. second LLM as above) to obtain at least one response from the LLM (i.e. response to query, an initial reply suggestion)); and
adapt generated response actions using additional structured data ([14:66-15:43] rest of the one or more words may be used to evaluate the relevance of the anticipated reply to the phrase uttered by the user … query may be sent to the LLM to receive at least one updated response from the LLM … updated reply to the user … at least one updated query may be sent to the LLM before the user has finished pronouncing the rest of the one or more words (i.e. incrementally generating responses and adapting based on additional structured data up to the full pronunciation)).
Regarding claim 4, Gelfenbeyn-Sahu disclose:
The electronic device of Claim 2, set forth above,
Gelfenbeyn discloses:
wherein the dynamic pre-fetching model comprises a reinforcement learning model configured to adjust a pre-fetching policy of the dynamic pre-fetching model based on an uncertainty score of a current response trajectory of the response actions from the second LLM ([14:66-15:43] rest of the one or more words may be used to evaluate the relevance of the anticipated reply to the phrase uttered by the user … determined that the relevance is below a predetermined threshold (i.e. current response trajectory by decreasing relevance) … query may be sent to the LLM to receive at least one updated response from the LLM … updated reply to the user … at least one updated query may be sent to the LLM before the user has finished pronouncing the rest of the one or more words (i.e. incrementally generating responses and adapting based on additional structured data up to the full pronunciation is adjusting the pre-fetching policy)).
Regarding claim 6, Gelfenbeyn-Sahu disclose:
The electronic device of Claim 2, wherein the at least one processing device, set forth above, is further configured to:
Gelfenbeyn discloses:
perform context-aware analysis, using an adaptive context-aware response timing model coupled to the first LLM ([15:26-37] one or more words may be analyzed to update the context (i.e. adaptive context-aware by updating) … updated context, at least one updated query to the LLM (i.e. by the first LLM to the second LLM)), for each user of the one or more users based on the conversation history ([14:14-28] context of an interaction between a user and an AI character … based on historical data concerning interactions between the user and the AI character (i.e. relationship information from the historical information … anticipated words … based on the context (i.e. as above, e.g. relationship information) … generating … at least one query (i.e. prompt) … based on the one or more anticipated words).
Regarding claims (8-9, 11 and 13); and (15-16 and 18), they do not further define nor teach over the limitations of claims 1-2, 4 and 6, therefore, claims (8-9, 11 and 13); and (15-16 and 18) are rejected for at least the same reasons set forth above as in claims 1-2, 4 and 6.
Claim(s) 3, 7, 10, 14, 17 and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Gelfenbeyn et al. (US-11960983-B1) hereinafter Gelfenbeyn in view of Sahu et al. (US-12475178-B2) hereinafter Sahu further in view of Jain et al. (US-20260023932-A1) hereinafter Jain.
Regarding claim 3, Gelfenbeyn-Sahu disclose:
The electronic device of Claim 2, wherein the at least one processing device, set forth above, is further configured to:
Gelfenbeyn-Sahu do not explicitly disclose:
generate, using a real-time feedback model, real-time feedback to adjust message priority based on an emotional tone of the response actions using an urgent score model.
However, Jain discloses:
generate, using a real-time feedback model, real-time feedback to adjust message priority based on an emotional tone of the response actions using an urgent score model ([0056] priority request response program determines that the manager prefers short, succinct responses … priority request response program may account for writing style, sentiment, tone … or any other stylistic or tactful factor (i.e. urgent score model) in assisting a user, see [0041] high priority, low priority … complex schedule that needs to be addressed in particular ways (i.e. real-time) … identify instances of users not responding to a request at all (i.e. feedback)).
It would have been obvious to one of ordinary skill in the pertinent art before the effective filing date of the claimed invention to modify the invention of Gelfenbeyn-Sahu in view of Jain to have generated real-time feedback to adjust message priority based on an emotional tone of the response actions using an urgent score model. One of ordinary skill in the art would have been motivated to do so to combine use of a machine learning model trained with long short-term memory and other techniques in order to assist users in responding to nuanced requests (Jain, [0011]).
Regarding claim 7, Gelfenbeyn-Sahu disclose:
The electronic device of Claim 6, set forth above,
Gelfenbeyn discloses:
wherein the context-aware analysis includes analysis of conversational flow and user intent ([13:65-14:13] context of an interaction between a user and an AI character … historical data concerning interactions between the user and the AI character (i.e. conversational flow) behavioral pattern of the user, and so forth … gesture of the user, an action performed by the user, a movement performed by the user, and the like (i.e. user intent) … based on current parameters of the virtual environment).
Gelfenbeyn does not explicitly disclose:
analysis of language nuances,
However, Jain discloses:
analysis of language nuances ([0011] combine use of a machine learning (“ML”) model trained with Long Short-Term Memory (“LSTM”) and other techniques, such as Layer-wise Relevance Propagation (“LRP”), in order to assist users in responding to nuanced requests, see [0033-0034] various types of affirmative or negative response with conditions or context … analyzing … contextual information … artificial intelligence-based techniques, such as sentiment analysis),
It would have been obvious to one of ordinary skill in the pertinent art before the effective filing date of the claimed invention to modify the invention of Gelfenbeyn in view of Jain to have utilized analysis of language nuances. One of ordinary skill in the art would have been motivated to do so to combine use of a machine learning model trained with long short-term memory and other techniques in order to assist users in responding to nuanced requests (Jain, [0011]).
Regarding claims 10, 14, 17 and 20, they do not further define nor teach over the limitations of claims 3 and 7, therefore, claims 10, 14, 17 and 20 are rejected for at least the same reasons set forth above as in claims 3 and 7.
Claim(s) 5, 12 and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Gelfenbeyn et al. (US-11960983-B1) hereinafter Gelfenbeyn in view of Sahu et al. (US-12475178-B2) hereinafter Sahu further in view of Zhao et al. (US-12223284-B2) hereinafter Zhao.
Regarding claim 5, Gelfenbeyn-Sahu disclose:
The electronic device of Claim 2, set forth above,
Gelfenbeyn-Sahu do not explicitly disclose:
wherein the latency-aware fallback model is configured to combine a placeholder suggestion with the response actions from the second LLM using an attention-weighted semantic alignment model.
However, Zhao discloses:
wherein the latency-aware fallback model is configured to combine a placeholder suggestion with the response actions from the second LLM using an attention-weighted semantic alignment model ([7:42-8:63] knowledge-guided attention mechanism is used to complete semantic alignment between an image and potential text knowledge … attention-based fusion mechanism is also applied to effectively integrate various features (i.e. various actions and placeholder suggestion)).
It would have been obvious to one of ordinary skill in the pertinent art before the effective filing date of the claimed invention to modify the invention of Gelfenbeyn-Sahu in view of Zhao to have the latency-aware fallback model configured to combine a placeholder suggestion with the response actions from the second LLM using an attention-weighted semantic alignment model. One of ordinary skill in the art would have been motivated to do so to effectively integrate various features (Zhao, [7:42-8:63]).
Regarding claims 12 and 19, they do not further define nor teach over the limitations of claim 5, therefore, claims 12 and 19 are rejected for at least the same reasons set forth above as in claim 5.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Cai et al. (US-12099975-B1) SYSTEM FOR ANALYZING LEARNERS;
Brewer et al. (US-20240356873-A1) PERSONAL AI INTENT UNDERSTANDING;
Wang et al. (US-20250094814-A1) FINE-TUNING A LARGE LANGUAGE MODEL (LLM) TO REDUCE THE INSTABILITY OF LLM OUTPUTS TO VARIATIONS IN PROMPTS;
Jensen et al. (US-20250133037-A1) GUIDED CONVERSATION CONTEXT COMPRESSION WITH ADVERSARIAL HYPOTHETICAL QUESTIONS AND EVALUATING RELEVANCE OF CONTEXTUAL INFORMATION FOR LLMS;
Lewis (US-20250201362-A1) NAVIGATING HISTORICAL HEALTHCARE INFORMATION;
Cantu et al. (US-20250371417-A1) METRICS FOR EVALUATING AI AGENT PERFORMANCE;
Nakayama et al. (US-20250363309-A1) DIALOGUE SUPPORT SYSTEM, DIALOGUE SUPPORT METHOD, AND STORAGE MEDIUM.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Alex Tran whose telephone number is (571)272-8173. The examiner can normally be reached Monday-Friday 10AM-6PM 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, Kamal Divecha can be reached at (571)272-5863. 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.
/Alex Tran/Primary Examiner, Art Unit 2453
