DETAILED ACTION
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 05/16/2025 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.
Claim(s) 1-20 rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1: Independent claims 1, 10, and 17 recite a method, system, and computer program product, respectively. These claims therefore invoke a statutory category (process and machine) in Step 1 of the Subject Matter Eligibility Test.
Step 2A, Prong One: Independent claim 1, under its broadest reasonable interpretation, recites writing out diverse prompts and changing the perplexity level associated with those prompts. Independent claim 10, under its broadest reasonable interpretation, recites writing out diverse prompts, changing the perplexity level associated with those prompts, and filtering the prompts by comparing them to observed lingual characteristics. Independent claim 17, under its broadest reasonable interpretation, recites getting a target value of perplexity, writing out diverse prompts with a perplexity value that matches the target, and analyzing the prompts against datasets and linguistic information. These are abstract ideas in the form of certain methods of organizing human activity (i.e. mental processes such as observation, evaluation, judgement, and opinion). The steps of generating data (i.e. prompts), analyzing the data, and modifying the data could be performed by a human using pen and paper or by purely mental reasoning, save for the recitation of generic computer components.
Step 2A, Prong Two: The claims do not integrate the judicial exception into a practical application. The recitation of “large-language models (LLMs)” is a generic instruction to perform the abstract idea on/using a computer and does not impose a meaningful limit on the judicial exception. The LLMs are recited at such a high-level of generality and are merely used as a tool to perform the abstract idea faster and more efficiently. The data receiving is a pre-solution activity required to perform the method and does not add a meaningful limitation. Mere data gathering, and analysis do not provide an inventive concept. There is no improvement to the data obtaining, analyzing, or modifying, the functioning of computers, or to any other technology or technical field.
Step 2B: The claims do not include any additional elements that amount to significantly more
than the judicial exception. The only additional element beyond the abstract idea is the LLMs, which perform generic computational functions such as receiving, analyzing, and outputting data. Such elements are well-understood, routine, and conventional within the field.
Accordingly, claims 1, 10, and 17 are directed to an abstract idea and do not include significantly
more than the abstract idea itself.
With respect to claims 2 and 3, the claims relate to using a psycholinguistic filter on the prompts. This is a mental process that could be performed by a human using pen and paper or by purely mental reasoning. No additional elements are present.
With respect to claims 4-7 and 13, the claims relate to adjusting the perplexity of the prompts by changing parameters of the LLMs such as temperature, top-k, top-p, or through Mirostat sampling. This is a mental process that could be performed by a human using pen and paper or by purely mental reasoning. The only additional element is “LLMs”, which is a generic instruction to perform the abstract idea on/using a computer and does not impose a meaningful limit on the judicial exception. No additional elements are present.
With respect to claim 8, the claim relates to the target population being a patient population with medical conditions that affect communication. This is pre-solution activity and therefore does not add a meaningful limitation. No additional elements are present.
With respect to claim 9, the claim relates to adjusting the level of perplexity based on patient information, communication history, and a target perplexity level. This is a mental process that could be performed by a human using pen and paper or by purely mental reasoning. No additional elements are present.
With respect to claim 11, the claim relates to the level of perplexity corresponding to a complexity of language of the target patient population. This is pre-solution activity and therefore does not add a meaningful limitation. No additional elements are present.
With respect to claim 12, the claim relates to analyzing a dataset to determine perplexity or communication style characteristics. This is a mental process that could be performed by a human using pen and paper or by purely mental reasoning. No additional elements are present. With respect to claim 14, the claim relates to generating the prompts based on medical queries by the target population. This is a mental process that could be performed by a human using pen and paper or by purely mental reasoning. No additional elements are present.
With respect to claim 15, the claim relates to refining the prompts based on feedback from adjusting and filtering. This is a mental process that could be performed by a human using pen and paper or by purely mental reasoning. No additional elements are present.
With respect to claim 16, the claim relates to implementing a feedback-loop workflow, wherein the adjusting and filtering function as value functions in a reinforcement learning framework, tuning control variables, and iteratively improving prompt generation. This is a mental process that could be performed by a human using pen and paper or by purely mental reasoning. No additional elements are present.
With respect to claim 18, the claim relates to generating the prompts based on a predicted temperature and a predicted set of top candidates. This is a mental process that could be performed by a human using pen and paper or by purely mental reasoning. No additional elements are present.
With respect to claim 19, the claim relates to adjusting the perplexity level based on a complexity of a medical condition associated with the patient population or language characteristics of the patient population. This is a mental process that could be performed by a human using pen and paper or by purely mental reasoning. No additional elements are present.
With respect to claim 20, the claim relates to comparing characteristics of the prompts to a baseline communication pattern and maintaining a perplexity range and linguistic diversity for the prompts. This is a mental process that could be performed by a human using pen and paper or by purely mental reasoning. No additional elements are present.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1, 4-7, and 9 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang et al. ("PATIENT-Ψ: Using Large Language Models to Simulate Patients for Training Mental Health Professionals", 10/03/2024), hereinafter referred to as Wang, in view of Basu et al. ("MIROSTAT: A NEURAL TEXT DECODING ALGORITHM THAT DIRECTLY CONTROLS PERPLEXITY", 01/14/2021), hereinafter referred to as Basu.
Regarding claim 1, Wang discloses a method comprising: generating, based on different large-language models (LLMs) ("To leverage LLMs’ capabilities as a judge for evaluating open-ended tasks, we use state-of-art LLMs, including GPT-4 and Llama 3 70B, as evaluators," Wang pg. 5), prompts that are diverse (Wang Figures 11-16 show the different prompts and conversation styles generated, i.e. producing "diverse" prompts);
to correspond to one or more communication styles of a target population (Wang Table 1 pg. 4 shows the different conversational styles that PATIENT-Ψ can simulate, i.e. a target population).
However, Wang fails to disclose and adjusting a level of perplexity in the generated prompts. Basu teaches a neural text decoding algorithm that directly controls perplexity.
Basu teaches and adjusting a level of perplexity in the generated prompts ("We introduce mirostat, a neural text decoding algorithm that actively controls the generative process to maintain the perplexity of generated text at a certain desired value. Mirostat uses an adaptive top-k sampling algorithm to actively tune the value of k which helps maintain the overall perplexity of the text," Basu Section 1 pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of adjusting the perplexity level of the prompts themselves. Perplexity is a measure used to evaluate how well an LLM predicts text. Both humans and models prefer when the perplexity of a text or a prompt is not too low or not too high, ensuring that both the human and models are able to understand the text or prompt clearly. Therefore, it would be obvious to adjust the perplexity to stay in this range and ensure a smooth exchange of information.
Regarding claim 4, Wang, in view of Basu, discloses all of the limitations of claim 1. However, Wang fails to disclose wherein the adjusting of the level of perplexity comprises adjusting the level of perplexity based on modification of one or more parameters of the different LLMs.
Basu teaches wherein the adjusting of the level of perplexity comprises adjusting the level of perplexity based on modification of one or more parameters of the different LLMs ("We introduce mirostat, a neural text decoding algorithm that actively controls the generative process to maintain the perplexity of generated text at a certain desired value. Mirostat uses an adaptive top-k sampling algorithm to actively tune the value of k which helps maintain the overall perplexity of the text," Basu Section 1 pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of adjusting the perplexity level of the prompts themselves by adjusting parameters of the LLMs. Perplexity is a measure used to evaluate how well an LLM predicts text. Both humans and models prefer when the perplexity of a text or a prompt is not too low or not too high, ensuring that both the human and models are able to understand the text or prompt clearly. Changing the parameters of the utilized LLMs would be the simplest way to affect their output perplexity. Therefore, it would be obvious to adjust the perplexity by changing LLM parameters to stay in this range and ensure a smooth exchange of information.
Regarding claim 5, Wang, in view of Basu, discloses all of the limitations of claim 4. However, Wang fails to disclose wherein the one or more parameters comprise a temperature.
Basu teaches wherein the one or more parameters comprise a temperature ("Distorting probability distributions, as in top-k, top-p, or temperature sampling help improve quality of generated texts, if parameters are properly tuned," Basu Section 1.1 pg. 2 and "In Fig. 3b, we use top-k sampling with varying temperature values," Basu Section 5.2 pg. 6).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of adjusting the perplexity level of the prompts themselves by adjusting parameters of the LLMs. Perplexity is a measure used to evaluate how well an LLM predicts text. Both humans and models prefer when the perplexity of a text or a prompt is not too low or not too high, ensuring that both the human and models are able to understand the text or prompt clearly. Changing the parameters of the utilized LLMs would be the simplest way to affect their output perplexity. Therefore, it would be obvious to adjust the perplexity by changing LLM parameters to stay in this range and ensure a smooth exchange of information.
Regarding claim 6, Wang, in view of Basu, discloses all of the limitations of claim 4. However, Wang fails to disclose wherein the one or more parameters comprise a top-k parameter or a top-p parameter.
Basu teaches wherein the one or more parameters comprise a top-k parameter or a top-p parameter ("We introduce mirostat, a neural text decoding algorithm that actively controls the generative process to maintain the perplexity of generated text at a certain desired value. Mirostat uses an adaptive top-k sampling algorithm to actively tune the value of k which helps maintain the overall perplexity of the text," Basu Section 1 pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of adjusting the perplexity level of the prompts themselves by adjusting parameters of the LLMs. Perplexity is a measure used to evaluate how well an LLM predicts text. Both humans and models prefer when the perplexity of a text or a prompt is not too low or not too high, ensuring that both the human and models are able to understand the text or prompt clearly. Changing the parameters of the utilized LLMs would be the simplest way to affect their output perplexity. Therefore, it would be obvious to adjust the perplexity by changing LLM parameters to stay in this range and ensure a smooth exchange of information.
Regarding claim 7, Wang, in view of Basu, discloses all of the limitations of claim 4. However, Wang fails to disclose wherein the adjusting of the level of perplexity comprises adjusting the level of perplexity based on Mirostat sampling of the prompts.
Basu teaches wherein the adjusting of the level of perplexity comprises adjusting the level of perplexity based on Mirostat sampling of the prompts ("We introduce mirostat, a neural text decoding algorithm that actively controls the generative process to maintain the perplexity of generated text at a certain desired value. Mirostat uses an adaptive top-k sampling algorithm to actively tune the value of k which helps maintain the overall perplexity of the text," Basu Section 1 pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of adjusting the perplexity level of the prompts themselves by adjusting parameters of the LLMs. Perplexity is a measure used to evaluate how well an LLM predicts text. Both humans and models prefer when the perplexity of a text or a prompt is not too low or not too high, ensuring that both the human and models are able to understand the text or prompt clearly. Changing the parameters of the utilized LLMs would be the simplest way to affect their output perplexity. Therefore, it would be obvious to adjust the perplexity by changing LLM parameters to stay in this range and ensure a smooth exchange of information.
Regarding claim 9, Wang, in view of Basu, discloses all of the limitations of claim 1. Wang further discloses patient information, a communication history (Wang Figure 2 pg. 3 shows Patient relevant history and various communications).
However, Wang fails to disclose wherein the adjusting of the level of perplexity comprises adjusting the level of perplexity based on a target perplexity level.
Basu teaches wherein the adjusting of the level of perplexity comprises adjusting the level of perplexity based on a target perplexity level ("We introduce mirostat, a neural text decoding algorithm that actively controls the generative process to maintain the perplexity of generated text at a certain desired value. Mirostat uses an adaptive top-k sampling algorithm to actively tune the value of k which helps maintain the overall perplexity of the text," Basu Section 1 pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of adjusting the perplexity level of the prompts themselves. Perplexity is a measure used to evaluate how well an LLM predicts text. Both humans and models prefer when the perplexity of a text or a prompt is not too low or not too high, ensuring that both the human and models are able to understand the text or prompt clearly. Therefore, it would be obvious to adjust the perplexity to stay in this range and ensure a smooth exchange of information.
Claim(s) 2-3, 8, and 10-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang, in view of Basu, and further in view of Elvevag et al. ("Quantifying incoherence in speech: An automated methodology and novel application to schizophrenia", 04/16/2007), hereinafter referred to as Elvevag.
Regarding claim 2, Wang, in view of Basu, discloses all of the limitations of claim 1. However, Wang fails to disclose applying a psycholinguistic filter that filters communication styles that are different than the one or more communication styles of the target population.
Elvevag teaches an automated methodology and novel application to schizophrenia by quantifying incoherence in speech.
Elvevag teaches applying a psycholinguistic filter that filters communication styles that are different than the one or more communication styles of the target population ("In word association and generation tasks LSA derived coherence scores were sensitive to differences between patients and controls, and correlated with clinical measures of thought disorder," Elvevag Abstract pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Elvevag’s teaching of using a psycholinguistic filter to filter out different communication and conversational styles. Filtering out text that departs from the control group or the target population is a predictable usage, and it is also well-known within the art of natural language processing (NLP).
Regarding claim 3, Wang, in view of Basu and further in view of Elvevag, discloses all of the limitations of claim 2. However, Wang fails to disclose wherein the psycholinguistic filter uses Latent Semantic Analysis (LSA) or machine learning classifiers to measure semantic coherence and linguistic features of the prompts.
Elvevag teaches wherein the psycholinguistic filter uses Latent Semantic Analysis (LSA) or machine learning classifiers to measure semantic coherence and linguistic features of the prompts ("In word association and generation tasks LSA derived coherence scores were sensitive to differences between patients and controls, and correlated with clinical measures of thought disorder," Elvevag Abstract pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Elvevag’s teaching of using a psycholinguistic filter to filter out different communication and conversational styles. Filtering out text that departs from the control group or the target population is a predictable usage, and LSA is a well-known technique within the art of NLP to measure semantic coherence and linguist features.
Regarding claim 8, Wang, in view of Basu, discloses all of the limitations of claim 1. However, Wang fails to disclose wherein the target population comprises a patient population with a specific medical condition that affects a communication style.
Elvevag teaches wherein the target population comprises a patient population with a specific medical condition that affects a communication style ("Patients with schizophrenia and healthy control volunteers were administered a variety of language tasks. The speech generated was transcribed and the coherence computed using Latent Semantic Analysis (LSA)," Elvevag Abstract pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Elvevag’s teaching of analyzing speech patterns and habits of schizophrenic patients. Using common NLP techniques to enhance communication with conversationally compromised patients is an obvious and predictable use.
Regarding claim 10, Wang discloses a computer system comprising: a processor set ("To build PATIENT-Ψ, we construct diverse patient cognitive models based on CBT principles and use large language models (LLMs) programmed with these cognitive models to act as a simulated therapy patient," Wang Abstract pg. 1, LLMs inherently require processors);
one or more computer-readable storage media ("To build PATIENT-Ψ, we construct diverse patient cognitive models based on CBT principles and use large language models (LLMs) programmed with these cognitive models to act as a simulated therapy patient," Wang Abstract pg. 1, LLMs inherently require storage);
and program instructions stored on the one or more computer-readable storage media to cause the processor set to perform operations comprising ("To build PATIENT-Ψ, we construct diverse patient cognitive models based on CBT principles and use large language models (LLMs) programmed with these cognitive models to act as a simulated therapy patient," Wang Abstract pg. 1, LLMs inherently require program instructions): generating diverse prompts (Wang Figures 11-16 show the different prompts and conversation styles generated, i.e. producing "diverse" prompts);
to match a communication style of a target patient population (Wang Table 1 pg. 4 shows the different conversational styles that PATIENT-Ψ can simulate, i.e. a target population).
However, Wang fails to disclose adjusting a level of perplexity in the diverse prompts; and filtering the diverse prompts based on a comparison of the diverse prompts to linguistic characteristics of observed patient communications of the target patient population.
Basu teaches adjusting a level of perplexity in the diverse prompts ("We introduce mirostat, a neural text decoding algorithm that actively controls the generative process to maintain the perplexity of generated text at a certain desired value. Mirostat uses an adaptive top-k sampling algorithm to actively tune the value of k which helps maintain the overall perplexity of the text," Basu Section 1 pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of adjusting the perplexity level of the prompts themselves. Perplexity is a measure used to evaluate how well an LLM predicts text. Both humans and models prefer when the perplexity of a text or a prompt is not too low or not too high, ensuring that both the human and models are able to understand the text or prompt clearly. Therefore, it would be obvious to adjust the perplexity to stay in this range and ensure a smooth exchange of information.
Elvevag teaches and filtering the diverse prompts based on a comparison of the diverse prompts to linguistic characteristics of observed patient communications of the target patient population ("In word association and generation tasks LSA derived coherence scores were sensitive to differences between patients and controls, and correlated with clinical measures of thought disorder," Elvevag Abstract pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Elvevag’s teaching of filtering out different communication and conversational styles from a target patient population. Filtering out text that departs from the control group or the target population is a predictable usage, and it is also well-known within the art of natural language processing (NLP).
Regarding claim 11, Wang, in view of Basu and further in view of Elvevag, discloses all of the limitations of claim 10. However, Wang fails to disclose wherein the level of perplexity corresponds to a complexity of a language of the target patient population.
Basu teaches wherein the level of perplexity corresponds to a complexity of a language of the target patient population ("Perplexity is a statistical metric used to evaluate quality of neural text generation, and is closely related to average surprise as shown in Fig. 7 in Appendix A and formalized in Sec. 2. A large-scale human subject experiment by Zhang et al. (2020) showed human-evaluated quality is closely related to the likelihood of the generated text for fixed number of tokens. In particular, reducing perplexity increases quality up to some point before the quality starts dropping. This implies that good control over perplexity of the generated text would give direct control over the quality of generated text (as evaluated by humans). Generating texts with an appropriately chosen target perplexity value may maximize quality of generated text," Basu Section 1 pg. 2).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of adjusting the perplexity level of the prompts themselves. Perplexity is a measure used to evaluate how well an LLM predicts text. Both humans and models prefer when the perplexity of a text or a prompt is not too low or not too high, ensuring that both the human and models are able to understand the text or prompt clearly. Therefore, it would be obvious to adjust the perplexity to stay in this range and ensure a smooth exchange of information.
Regarding claim 12, Wang, in view of Basu and further in view of Elvevag, discloses all of the limitations of claim 10. Wang further discloses analyzing a dataset of patient communications or communication style characteristics of the target patient population (Wang Figure 2 pg. 3 shows Patient relevant history and various communications).
However, Wang fails to disclose to determine the level of perplexity.
Basu teaches to determine the level of perplexity ("We introduce mirostat, a neural text decoding algorithm that actively controls the generative process to maintain the perplexity of generated text at a certain desired value. Mirostat uses an adaptive top-k sampling algorithm to actively tune the value of k which helps maintain the overall perplexity of the text," Basu Section 1 pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of adjusting the perplexity level of the prompts themselves. Perplexity is a measure used to evaluate how well an LLM predicts text. Both humans and models prefer when the perplexity of a text or a prompt is not too low or not too high, ensuring that both the human and models are able to understand the text or prompt clearly. Therefore, it would be obvious to adjust the perplexity to stay in this range and ensure a smooth exchange of information.
Regarding claim 13, Wang, in view of Basu and further in view of Elvevag, discloses all of the limitations of claim 10. However, Wang fails to disclose wherein the adjusting of the level of perplexity comprises adjusting the level of perplexity based on temperature scaling, top-k sampling, and top-p sampling to control randomness and diversity of the diverse prompts.
Basu teaches wherein the adjusting of the level of perplexity comprises adjusting the level of perplexity based on temperature scaling ("Distorting probability distributions, as in top-k, top-p, or temperature sampling help improve quality of generated texts, if parameters are properly tuned," Basu Section 1.1 pg. 2 and "In Fig. 3b, we use top-k sampling with varying temperature values," Basu Section 5.2 pg. 6), top-k sampling ("We introduce mirostat, a neural text decoding algorithm that actively controls the generative process to maintain the perplexity of generated text at a certain desired value. Mirostat uses an adaptive top-k sampling algorithm to actively tune the value of k which helps maintain the overall perplexity of the text," Basu Section 1 pg. 1), and top-p sampling to control randomness and diversity of the diverse prompts ("Another sampling method is top-p, also known as nucleus sampling, where the next word is chosen from the top x probable choices, where x is the smallest integer such that their cumulative probability mass is at least p," Basu Section 1 pg. 1-2).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of adjusting the perplexity level of the prompts themselves by adjusting parameters of the LLMs. Perplexity is a measure used to evaluate how well an LLM predicts text. Both humans and models prefer when the perplexity of a text or a prompt is not too low or not too high, ensuring that both the human and models are able to understand the text or prompt clearly. Changing the parameters of the utilized LLMs would be the simplest way to affect their output perplexity. Therefore, it would be obvious to adjust the perplexity by changing LLM parameters to stay in this range and ensure a smooth exchange of information.
Regarding claim 14, Wang, in view of Basu and further in view of Elvevag, discloses all of the limitations of claim 10. Wang further discloses wherein the generating of the diverse prompts comprises generating the diverse prompts based on multiple medical queries by the target patient population in association with a medical condition (Wang Figure 2 pg. 3 shows medical queries and comments in the Cognitive Model section).
Regarding claim 15, Wang, in view of Basu and further in view of Elvevag, discloses all of the limitations of claim 10. However, Wang fails to disclose wherein the operations further comprise refining the diverse prompts based on feedback from the adjusting and filtering.
Basu teaches wherein the operations further comprise refining the diverse prompts based on feedback from the adjusting and filtering (Basu Algorithm 2 shows error between observed and target and updating based on the error).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of refining the prompts based on feedback. Feedback is an important part of creating an optimized model, and its inclusion here acts to better the LLMs’ outputs. Using feedback (i.e. error) is a well-known and understood technique within the art of NLP.
Regarding claim 16, Wang, in view of Basu and further in view of Elvevag, discloses all of the limitations of claim 10. However, Wang fails to disclose implementing a feedback-loop workflow, wherein the adjusting and the filtering function as value functions in a reinforcement learning framework; automatically tuning control variables for the adjusting and the filtering based on a quality assessment of the diverse prompts; and iteratively improving prompt generation based on the automatic tuning.
Basu teaches implementing a feedback-loop workflow, wherein the adjusting and the filtering function as value functions in a reinforcement learning framework (Basu Algorithm 2 shows error between observed and target and updating based on the error);
automatically tuning control variables for the adjusting and the filtering based on a quality assessment of the diverse prompts (Basu Algorithm 2 shows error between observed and target and updating automatically based on the error);
and iteratively improving prompt generation based on the automatic tuning (Basu Algorithm 2 shows error between observed and target and updating automatically based on the error, thereby converging and improving the algorithm).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of refining the prompts based on feedback. Feedback is an important part of creating an optimized model, and its inclusion here acts to better the LLMs’ outputs. Using feedback (i.e. error) is a well-known and understood technique within the art of NLP.
Regarding claim 17, Wang discloses a computer program product comprising: one or more computer-readable storage media ("To build PATIENT-Ψ, we construct diverse patient cognitive models based on CBT principles and use large language models (LLMs) programmed with these cognitive models to act as a simulated therapy patient," Wang Abstract pg. 1, LLMs inherently require storage);
and program instructions stored on the one or more computer-readable storage media to perform operations comprising ("To build PATIENT-Ψ, we construct diverse patient cognitive models based on CBT principles and use large language models (LLMs) programmed with these cognitive models to act as a simulated therapy patient," Wang Abstract pg. 1, LLMs inherently require program instructions): associated with a communication style of a patient population (Wang Table 1 pg. 4 shows the different conversational styles that PATIENT-Ψ can simulate, i.e. a target population);
generating diverse and unique prompts (Wang Figures 11-16 show the different prompts and conversation styles generated, i.e. producing "diverse" prompts) using multiple large language models (LLMs) ("To leverage LLMs’ capabilities as a judge for evaluating open-ended tasks, we use state-of-art LLMs, including GPT-4 and Llama 3 70B, as evaluators," Wang pg. 5).
However, Wang fails to disclose receiving a target perplexity value; having a level of perplexity that corresponds to the target perplexity value; and analyzing the prompts using patient communication datasets and psycholinguistic information of the patient population.
Basu teaches receiving a target perplexity value ("Then, we use this analysis to design a feedback-based adaptive top-k text decoding algorithm called mirostat that generates text (of any length) with a predetermined target value of perplexity without any tuning," Basu Abstract pg. 1);
having a level of perplexity that corresponds to the target perplexity value ("Then, we use this analysis to design a feedback-based adaptive top-k text decoding algorithm called mirostat that generates text (of any length) with a predetermined target value of perplexity without any tuning," Basu Abstract pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of adjusting the perplexity level of the prompts themselves. Perplexity is a measure used to evaluate how well an LLM predicts text. Both humans and models prefer when the perplexity of a text or a prompt is not too low or not too high, ensuring that both the human and models are able to understand the text or prompt clearly. Therefore, it would be obvious to adjust the perplexity to stay in this range and ensure a smooth exchange of information.
Elvevag teaches and analyzing the prompts using patient communication datasets and psycholinguistic information of the patient population ("In word association and generation tasks LSA derived coherence scores were sensitive to differences between patients and controls, and correlated with clinical measures of thought disorder," Elvevag Abstract pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Elvevag’s teaching of filtering out different communication and conversational styles from a target patient population. Filtering out text that departs from the control group or the target population is a predictable usage, and it is also well-known within the art of natural language processing (NLP).
Regarding claim 18, Wang, in view of Basu and further in view of Elvevag, discloses all of the limitations of claim 17. However, Wang fails to disclose wherein the generating of the prompts includes generating the prompts based on a predicted temperature and a predicted set of top candidates and probabilities.
Basu teaches wherein the generating of the prompts includes generating the prompts based on a predicted temperature ("Distorting probability distributions, as in top-k, top-p, or temperature sampling help improve quality of generated texts, if parameters are properly tuned," Basu Section 1.1 pg. 2 and "In Fig. 3b, we use top-k sampling with varying temperature values," Basu Section 5.2 pg. 6) and a predicted set of top candidates and probabilities ("We introduce mirostat, a neural text decoding algorithm that actively controls the generative process to maintain the perplexity of generated text at a certain desired value. Mirostat uses an adaptive top-k sampling algorithm to actively tune the value of k which helps maintain the overall perplexity of the text," Basu Section 1 pg. 1 and "Another sampling method is top-p, also known as nucleus sampling, where the next word is chosen from the top x probable choices, where x is the smallest integer such that their cumulative probability mass is at least p," Basu Section 1 pg. 1-2).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of adjusting the perplexity level of the prompts themselves by adjusting parameters of the LLMs. Perplexity is a measure used to evaluate how well an LLM predicts text. Both humans and models prefer when the perplexity of a text or a prompt is not too low or not too high, ensuring that both the human and models are able to understand the text or prompt clearly. Changing the parameters of the utilized LLMs would be the simplest way to affect their output perplexity. Therefore, it would be obvious to adjust the perplexity by changing LLM parameters to stay in this range and ensure a smooth exchange of information.
Regarding claim 19, Wang, in view of Basu and further in view of Elvevag, discloses all of the limitations of claim 17. However, Wang fails to disclose wherein the operations further comprise adjusting the level of perplexity based on a complexity of a medical condition associated with the patient population or one or more language characteristics of the patient population.
Basu teaches wherein the operations further comprise adjusting the level of perplexity ("We introduce mirostat, a neural text decoding algorithm that actively controls the generative process to maintain the perplexity of generated text at a certain desired value. Mirostat uses an adaptive top-k sampling algorithm to actively tune the value of k which helps maintain the overall perplexity of the text," Basu Section 1 pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of adjusting the perplexity level of the prompts themselves. Perplexity is a measure used to evaluate how well an LLM predicts text. Both humans and models prefer when the perplexity of a text or a prompt is not too low or not too high, ensuring that both the human and models are able to understand the text or prompt clearly. Therefore, it would be obvious to adjust the perplexity to stay in this range and ensure a smooth exchange of information.
Elvevag teaches based on a complexity of a medical condition associated with the patient population or one or more language characteristics of the patient population ("In word association and generation tasks LSA derived coherence scores were sensitive to differences between patients and controls, and correlated with clinical measures of thought disorder," Elvevag Abstract pg. 1 and "Patients with schizophrenia and healthy control volunteers were administered a variety of language tasks. The speech generated was transcribed and the coherence computed using Latent Semantic Analysis (LSA)," Elvevag Abstract pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Elvevag’s teaching of analyzing speech patterns and habits of schizophrenic patients. Using common NLP techniques to enhance communication with conversationally compromised patients is an obvious and predictable use.
Regarding claim 20, Wang, in view of Basu and further in view of Elvevag, discloses all of the limitations of claim 17. However, Wang fails to disclose comparing one or more characteristics of the prompts to a baseline communication pattern; and maintaining a perplexity range and linguistic diversity for the prompts based on the comparing.
Basu teaches and maintaining a perplexity range and linguistic diversity for the prompts based on the comparing ("Experiments show that for low values of k and p, perplexity drops significantly with generated text length and leads to excessive repetitions (the boredom trap). Contrarily, for large values of k and p, perplexity increases with generated text length and leads to incoherence (confusion trap). Mirostat avoids both traps. Specifically, we show that setting target perplexity value beyond a threshold yields negligible sentence-level repetitions," Basu Abstract pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Basu’s teaching of adjusting the perplexity level of the prompts themselves. Perplexity is a measure used to evaluate how well an LLM predicts text. Both humans and models prefer when the perplexity of a text or a prompt is not too low or not too high, ensuring that both the human and models are able to understand the text or prompt clearly. Therefore, it would be obvious to adjust the perplexity to stay in this range and ensure a smooth exchange of information.
Elvevag teaches comparing one or more characteristics of the prompts to a baseline communication pattern ("In word association and generation tasks LSA derived coherence scores were sensitive to differences between patients and controls, and correlated with clinical measures of thought disorder," Elvevag Abstract pg. 1).
It would have been obvious to one of ordinary skill within the art before the effective filing date of the claimed invention to have modified Wang’s disclosure of generating diverse prompts that are similar to communication styles of a target population by including Elvevag’s teaching of analyzing speech patterns and habits of schizophrenic patients. Using common NLP techniques to enhance communication with conversationally compromised patients is an obvious and predictable use.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
US Patent Application Publication No. 2025/0103621
US Patent Application Publication No. 2023/0320642
US Patent No. 12,367,308
Cohen et al., “A Tale of Two Perplexities: Sensitivity of Neural Language Models to Lexical Retrieval Deficits in Dementia of the Alzheimer’s Type”, 07/10/2020
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ADAM MICHAEL WEAVER whose telephone number is (571)272-7062. The examiner can normally be reached Monday-Friday, 8AM-5PM 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, Richemond Dorvil can be reached at (571) 272-7602. 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.
/ADAM MICHAEL WEAVER/ Examiner, Art Unit 2658
/RICHEMOND DORVIL/ Supervisory Patent Examiner, Art Unit 2658
Prosecution Insights