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 .
Remarks
This action is in response to the amendments received on 1/20/26. Claims 1-32 are pending in the application. Claims 31 and 32 have been added. Applicants' arguments have been carefully and respectfully considered.
Claims 1-9, 16-20, and 24-27 are provisionally rejected on the ground of nonstatutory double patenting.
Claim(s) 1-3, 11-14, and 22-25 are rejected under 35 U.S.C. 103 as being unpatentable over Pathak et al. (US 2024/0394479), and further in view of Agarwal et al. (US 11,947,916).
Claims 4, 10, 15, 21, 26, and 30 are rejected under 35 U.S.C. 103 as being unpatentable over Pathak in view of Agarwal, and further in view of Goto et al. (US 2023/0259711).
Claim(s) 5-8, 16-19, and 27-29 are rejected under 35 U.S.C. 103 as being unpatentable over Pathak in view of Agarwal, and further in view of Reza et al. (US 2023/0237277).
Claims 31 and 32 are rejected under 35 U.S.C. 103 as being unpatentable over Pathak et al. (US 2024/0394479), and further in view of Bayliss et al. (US 2025/0112878).
Double Patenting
The nonstatutory double patenting rejection is based on a judicially created doctrine grounded in public policy (a policy reflected in the statute) so as to prevent the unjustified or improper timewise extension of the “right to exclude” granted by a patent and to prevent possible harassment by multiple assignees. A nonstatutory double patenting rejection is appropriate where the conflicting claims are not identical, but at least one examined application claim is not patentably distinct from the reference claim(s) because the examined application claim is either anticipated by, or would have been obvious over, the reference claim(s). See, e.g., In re Berg, 140 F.3d 1428, 46 USPQ2d 1226 (Fed. Cir. 1998); In re Goodman, 11 F.3d 1046, 29 USPQ2d 2010 (Fed. Cir. 1993); In re Longi, 759 F.2d 887, 225 USPQ 645 (Fed. Cir. 1985); In re Van Ornum, 686 F.2d 937, 214 USPQ 761 (CCPA 1982); In re Vogel, 422 F.2d 438, 164 USPQ 619 (CCPA 1970); In re Thorington, 418 F.2d 528, 163 USPQ 644 (CCPA 1969).
A timely filed terminal disclaimer in compliance with 37 CFR 1.321(c) or 1.321(d) may be used to overcome an actual or provisional rejection based on nonstatutory double patenting provided the reference application or patent either is shown to be commonly owned with the examined application, or claims an invention made as a result of activities undertaken within the scope of a joint research agreement. See MPEP § 717.02 for applications subject to examination under the first inventor to file provisions of the AIA as explained in MPEP § 2159. See MPEP § 2146 et seq. for applications not subject to examination under the first inventor to file provisions of the AIA . A terminal disclaimer must be signed in compliance with 37 CFR 1.321(b).
The filing of a terminal disclaimer by itself is not a complete reply to a nonstatutory double patenting (NSDP) rejection. A complete reply requires that the terminal disclaimer be accompanied by a reply requesting reconsideration of the prior Office action. Even where the NSDP rejection is provisional the reply must be complete. See MPEP § 804, subsection I.B.1. For a reply to a non-final Office action, see 37 CFR 1.111(a). For a reply to final Office action, see 37 CFR 1.113(c). A request for reconsideration while not provided for in 37 CFR 1.113(c) may be filed after final for consideration. See MPEP §§ 706.07(e) and 714.13.
The USPTO Internet website contains terminal disclaimer forms which may be used. Please visit www.uspto.gov/patent/patents-forms. The actual filing date of the application in which the form is filed determines what form (e.g., PTO/SB/25, PTO/SB/26, PTO/AIA /25, or PTO/AIA /26) should be used. A web-based eTerminal Disclaimer may be filled out completely online using web-screens. An eTerminal Disclaimer that meets all requirements is auto-processed and approved immediately upon submission. For more information about eTerminal Disclaimers, refer to www.uspto.gov/patents/apply/applying-online/eterminal-disclaimer.
This is a provisional nonstatutory double patenting rejection because the patentably indistinct claims have not in fact been patented.
19/027199
19/028561
A non-transitory computer-readable medium comprising computer-readable instructions stored thereon that when executed by a processor cause the processor to:
receive a set of documents for generating a prompt to be input into a language model having a context window with a token limit, from which to generate a topic label and a topic description for a topic,
wherein the topic label comprises a name for the topic and the topic description comprises a description of the topic in a human-understandable format;
A non-transitory computer-readable medium comprising computer-readable instructions stored thereon that when executed by a processor cause the processor to:
receive a set of documents for generating a prompt to be input into a language model having a context window with a token limit, from which to generate a topic label and a topic description for a topic;
input the set of documents into an unsupervised machine learning model;
input the set of documents into an unsupervised machine learning model;
execute the unsupervised machine learning model to output a plurality of topics for the set of the documents, each of the plurality of topics comprising a plurality of topic terms and each of the plurality of topic terms associated with a first weight value;
execute the unsupervised machine learning model to output the topic for the set of the documents, the topic comprising a plurality of topic terms;
select a first subset of topic terms for each topic of the plurality of topics, wherein the first subset of topic terms for each topic are selected from the plurality of topic terms of that topic based on the first weight value assigned to each of the plurality of topic terms of that topic;
select a first subset of topic documents from the set of documents that belong to the topic;
compute an inverse document frequency weight value for each topic term in the first subset of topic terms of each topic;
compute a second weight value for each topic term in the first subset of topic terms based on the first weight value and the inverse document frequency weight value for that topic term;
select a second subset of topic terms for each topic from the first subset of topic terms, wherein the second subset of topic terms are selected based on the second weight value of each topic term in the first subset of topic terms;
select a second subset of topic documents from the first subset of topic documents based on the plurality of topic terms;
identify a title from each of the second subset of topic documents to obtain a plurality of titles;
generate a compressed representation of the set of documents from the second subset of topic terms of each topic to include in a prompt for each topic, wherein the compressed representation having a first number of tokens to be stored in a computer memory that is less than a second number of tokens in the plurality of topic terms;
generate a compressed representation of the set of documents based on the plurality of titles to include in a prompt, wherein the compressed representation having a first number of tokens stored in a computer memory that is less than a second number of tokens in the plurality of topic terms,
the compressed representation being generated by concatenating the plurality of titles identified from the second subset of topic documents and excluding unselected titles from the set of documents that belong to the topic;
wherein:
the compressed representation reduces a token count of the prompt to fit within the context window of the language model
wherein:
the compressed representation reduces a token count of the prompt to fit within the context window of the language model
the compressed representation is generated based on (i) text segments extracted by an information extraction model and (ii) topic terms and associated weight values output by the unsupervised machine learning model;
the compressed representation is generated based on (i) text segments extracted by an information extraction model and (ii) topic terms and associated weight values output by the unsupervised machine learning model;
generating the compressed representation comprises selecting, ordering, and concatenating topic terms based on the weight values to form a machine-consumable prompt;
generating the compressed representation comprises selecting, ordering, and concatenating topic terms based on the weight values to form a machine-consumable prompt;
the machine-consumable prompt has a token count that fits within a context window token limit of the language model;
the machine-consumable prompt has a token count that fits within a context window token limit of the language model;
input the machine-consumable prompt of each topic into the language model that is distinct from the unsupervised machine learning model; and
input the machine-consumable prompt of each topic into the language model that is distinct from the unsupervised machine learning model; and
generate the topic label and topic description for each topic of the plurality of topics by executing the language model based on the input of the machine-consumable prompt,
the compressed representation being generated by concatenating the selected subset of topic terms and excluding unselected topic terms of the second number of tokens in the plurality of topic terms.
generate the topic label and topic description for the topic by executing execute the language model based on the input of the machine-consumable prompt.
Claims 1-9, 16-20, and 24-27 are provisionally rejected on the ground of nonstatutory double patenting as being unpatentable over claim 1-9, 12-20, and 23-29 of copending Application No. 19/028561 in view of Agarwal et al. (US 11,947,916). Copending application does not claim “compute an inverse document frequency weight value for each topic term in the first subset of topic terms of each topic; compute a second weight value for each topic term in the first subset of topic terms based on the first weight value and the inverse document frequency weight value for that topic term;”
Agarwal teaches compute an inverse document frequency weight value for each topic term in the first subset of topic terms of each topic (Agarwal, Col. 3 Li. 18-22, Document corpus 212 is analyzed by a topic definition and modeler component 214. The topic definition and modeler component 214 may use techniques such as LDA, NLP, Term Frequency-Inverse Document Frequency (TF-IDF), or other algorithms.);
compute a second weight value for each topic term in the first subset of topic terms based on the first weight value and the inverse document frequency weight value for that topic term (Agarwal, Col. 3 Li. 35-60, a topic may include "paid" and the stems may include "payment," "pay," and the like. The expanded list of topic terms 216 may be passed to the occurrence ordering component 220. The occurrence ordering component 220 may order the topic terms by frequency of occurrence in the document corpus 212. The occurrence ordering component 220 may remove topic terms that are below a threshold frequency of occurrence. The resulting list of topic terms may be represented in a table such as: see table of col. 3 li. 46)
It would have been obvious at the effective filing date of the invention to a person having ordinary skill in the art to which said subject matter pertains to have included the teachings of Agarwal because it allows users to determine both topics and contexts of those topics over a large corpus of text (Agarwal, Col. 2 Li. 7-17).
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-3, 11-14, and 22-25 are rejected under 35 U.S.C. 103 as being unpatentable over Pathak et al. (US 2024/0394479), and further in view of Agarwal et al. (US 11,947,916).
With respect to claim 1, Pathak teaches a non-transitory computer-readable medium comprising computer-readable instructions stored thereon that when executed by a processor cause the processor to:
receive a set of documents for generating a prompt to be input into a language model (Pathak, pa 0069, The application system 108 includes functionality that enables a user to interact with an online resource that links related information items in a graph. In a particular dialogue turn, the user submits an input query that incorporates information pulled from the graph. Or the knowledge-supplementing component 136 extracts information from the graph) having a context window with a token limit (Pathak, pa 0053, the dialogue system 104 can adapt the way it constructs the prompt information 124 … the execution platform that runs the language model 106. & pa 0078, the resource availability-assessing component 606 receives an input signal that indicates the current processing capacity of the application system 108 that uses the dialogue system 104. The resource availability assessing component 606 uses a rules-based system and/or a machine-trained model and/or other functionality to map these factors into a complexity level.);
input the set of documents into an unsupervised machine learning model (Pathak, pa 0090, the compression component 138 compresses the content in the candidate context information 902, including the dialogue history and/or the knowledge information retrieved by the knowledge-supplementing component 136 & pa 0092, once the compression component 138 is invoked, the compression-managing component 914 invokes all of the individual compression components (906, 908, 910, 912), which can then operate in parallel. & pa 0097, The topic-modeling component 910 can likewise uses various rules-based logic and/or machine-trained models to extract topics associated with the source information 904, including Latent Dirichlet Allocation (LDA));
execute the unsupervised machine learning model to output a plurality of topics …, each of the plurality of topics comprising a plurality of topic terms and each of the plurality of topic terms associated with a first weight value (Pathak, pa 0097, The topic-modeling component 910 can likewise uses various rules-based logic and/or machine-trained models to extract topics associated with the source information 904, including Latent Dirichlet Allocation (LDA));
select a first subset of topic terms for each topic of the plurality of topics, wherein the first subset of topic terms for each topic are selected from the plurality of topic terms of that topic based on the first weight value assigned to each of the plurality of topic terms of that topic (Pathak, pa 0098, the compression component 138 also weights the relevance of selected terms (keywords, named entities, topics, etc.) based on one or more weighting factors, and uses those weights factors in determining which terms are to be included in the prompt information 124…. By favorably weighting a selected term, the compression component 138 promotes this term over other terms that are not similarly weighted, and increases the likelihood that the selected term will be included in the top K information items);
generate a compressed representation … to include in a prompt for each topic (Pathak, Fig. 1 & pa 0045, The dynamic prompt-generating component 128 also assembles information provided by the separate analysis components 130 into the prompt information 124. & pa 0098, By favorably weighting a selected term, the compression component 138 promotes this term over other terms that are not similarly weighted, and increases the likelihood that the selected term will be included in the top K information items), wherein the compressed representation having a first number of tokens to be stored in a computer memory that is less than a second number of tokens in the plurality of topic terms (Pathak, pa 0054, the dialogue system 104 compresses source information from which the prompt information 124 is constructed, e.g., by picking salient terms from the source information and/or removing redundant information from the source information. & pa 0059, the compression component 138 uses topic analysis to identify one or more topics that are pertinent to the source information. The compression component 138 has the effect of compressing the source information by using selected terms to describe it. & pa 0091, The compression component 138 uses different components and associated techniques to perform different types of compression. Generally, each of the techniques provides a reduced-sized representation of the source information that preserves at least some semantic content of the source information in an original form. The reduced-sized representation of the source information is included in the prompt information 124 in lieu of the source information in its original form.) wherein:
the compressed representation reduces a token count of the prompt to fit within the context window of the language model (Pathak, pa 0073, the content unit amount-assessing component 608 determines, based on the assessed complexity level, a maximum number of content units to include in the prompt information 124 for the current dialogue turn & pa 0091, Generally, each of the techniques provides a reduced-sized representation of the source information that preserves at least some semantic content of the source information in an original form.);
the compressed representation is generated based on (i) text segments extracted by an information extraction model (Pathak, pa 0094, The keyword-extracting component 906 uses any rules-based logic (e.g., any algorithm) or machine-trained model to detect prominent keywords or named entities associated with the source information 904.) and (ii) topic terms and associated weight values output by the unsupervised machine learning model (Pathak, pa 0098, the compression component 138 also weights the relevance of selected terms (keywords, named entities, topics, etc.) based on one or more weighting factors, and uses those weights factors in determining which terms are to be included in the prompt information 124.);
generating the compressed representation comprises selecting, ordering, and concatenating topic terms based on the weight values to form a machine-consumable prompt (Pathak, pa 0098, the compression component 138 selects the K top-ranked terms. By favorably weighting a selected term, the compression component 138 promotes this term over other terms that are not similarly weighted, and increases the likelihood that the selected term will be included in the top K information items.);
the machine-consumable prompt has a token count that fits within a context window token limit of the language model (Pathak, pa 0002, an application typically limits the size of the prompt that can be input to the language model. & pa 0073, the content unit amount-assessing component 608 determines, based on the assessed complexity level, a maximum number of content units to include in the prompt information 124 for the current dialogue turn);
input the machine-consumable prompt of each topic into the language model that is distinct from the unsupervised machine learning model (Pathak, pa 0116, The language model 1402 commences with the receipt of the model-input information, e.g., corresponding to the prompt information 124. The model-input information is expressed as a series of linguistic tokens 1406 & Fig. 16, pa 0129, In block 1608, the dialogue system submits the prompt information to the machine trained language model, and receives a response (e.g., the response 126) from the machine-trained language model based on the prompt information.) & Fig. 1, compression component (having topic model as shown in Fig. 9) and separate language model 106 & Fig. 9, compression managing component 914 (compression component 138) including topic modeling component 910 creating compressed source information & ); and
generate the topic label … the topic by executing the language model based on the input of the machine-consumable prompt (Pathak, Fig. 16, pa 0129, In block 1610, the dialogue system 104 generates output information (e.g., the output information 120) based on the response), the compressed representation being generated by concatenating the selected subset of topic terms and excluding unselected topic terms of the second number of tokens in the plurality of topic terms (Pathak, pa 0116, The language model 1402 commences with the receipt of the model-input information, e.g., corresponding to the prompt information 124. The model-input information is expressed as a series of linguistic tokens 1406).
Pathak doesn't expressly discuss receive a set of documents for generating a prompt to be input into a language model having a context window with a token limit, from which to generate a topic label and a topic description for a topic; select a first subset of topic terms for each topic of the plurality of topics, wherein the first subset of topic terms for each topic are selected from the plurality of topic terms of that topic based on the first weight value assigned to each of the plurality of topic terms of that topic; compute an inverse document frequency weight value for each topic term in the first subset of topic terms of each topic; compute a second weight value for each topic term in the first subset of topic terms based on the first weight value and the inverse document frequency weight value for that topic term and generate a compressed representation of the set of documents, generate the topic label and topic description for each topic of the plurality of topics by executing the language model based on the input of the machine-consumable prompt.
Agarwal teaches receive a set of documents … from which to generate a topic label and a topic description for a topic (Agarwal, Col. 3 Li. 18-19, Document corpus 212 is analyzed by a topic definition and modeler component 214), wherein the topic label comprises a name for the topic and the topic description comprises a description of the topic in a human-understandable format (Agarwal, Col. 2 Li. 7-13, Disclosed in some examples are methods, systems, and machine readable mediums which provide topic sentences summarizing topics determined within a corpus of documents. These summaries may be used by customer service associates, analysts, or other users to quickly determine both topics discussed and contexts of those topics over a large corpus of text.);
input the set of documents into an unsupervised machine learning model (Agarwal, Col. 3 Li. 18-22, Document corpus 212 is analyzed by a topic definition and modeler component 214. The topic definition and modeler component 214 may use techniques such as LDA, NLP, Term Frequency-Inverse Document Frequency (TF-IDF), or other algorithms);
execute the unsupervised machine learning model to output a plurality of topics for the set of the documents (Agarwal, Col. 3 Li. 15-17, Topic definition and modeler component 214 may execute on one or more computing device of a topic definition service 130), each of the plurality of topics comprising a plurality of topic terms and each of the plurality of topic terms associated with a first weight value (Agarwal, Col. 3 Li. 24-27, Sentences in the corpus may then be separated into contexts by topic modeling algorithms such as LDA based upon similarities in topic terms for each sentence. In some examples, the contexts may be defined by adjacent sentences having similar topic terms. Similarity may be measured by TF-IDF scores);
compute an inverse document frequency weight value for each topic term in the first subset of topic terms of each topic (Agarwal, Col. 3 Li. 18-22, Document corpus 212 is analyzed by a topic definition and modeler component 214. The topic definition and modeler component 214 may use techniques such as LDA, NLP, Term Frequency-Inverse Document Frequency (TF-IDF), or other algorithms.);
compute a second weight value for each topic term in the first subset of topic terms based on the first weight value and the inverse document frequency weight value for that topic term (Agarwal, Col. 3 Li. 35-60, a topic may include "paid" and the stems may include "payment," "pay," and the like. The expanded list of topic terms 216 may be passed to the occurrence ordering component 220. The occurrence ordering component 220 may order the topic terms by frequency of occurrence in the document corpus 212. The occurrence ordering component 220 may remove topic terms that are below a threshold frequency of occurrence. The resulting list of topic terms may be represented in a table such as: see table of col. 3 li. 46)
generate a compressed representation of the set of documents from the second subset of topic terms of each topic (Agarwal, Col. 5 Li. 61-63, Based upon the ordered list of topic terms, one or more topic definition sentences may be generated by the content generator 230. & Col. 6 Li. 59-62, each topic term in a context is used to build a different sentence).
It would have been obvious at the effective filing date of the invention to a person having ordinary skill in the art to which said subject matter pertains to have modified Pathak with the teachings of Agarwal because it allows users to determine both topics and contexts of those topics over a large corpus of text (Agarwal, Col. 2 Li. 7-17).
With respect to claim 2, Pathak in view of Agarwal teaches the non-transitory computer-readable medium of claim 1, wherein the unsupervised machine learning model is a topic model (Pathak, pa 0097, The topic-modeling component 910 can likewise uses various rules-based logic and/or machine-trained models to extract topics associated with the source information 904, including Latent Dirichlet Allocation (LDA)) and the language model is a Large Language Model (LLM) (Pathak, pa 0005, language model refers to a machine-trained model that is capable of processing language-based input information and, optionally, any other kind of input information (including video information, image information, audio information, etc.). As such, a language model can correspond to a multi-modal machine-trained model.).
With respect to claim 3, Pathak in view of Agarwal teaches the non-transitory computer-readable medium of claim 2, wherein the topic model is a Latent Dirichlet Allocation (LDA) clustering model (Pathak, pa 0097, The topic-modeling component 910 can likewise uses various rules-based logic and/or machine-trained models to extract topics associated with the source information 904, including Latent Dirichlet Allocation (LDA)) or a Singular Value Decomposition (SVD) model.
With respect to claim 11, Pathak in view of Agarwal teaches the non-transitory computer-readable medium of claim 1, wherein a number of topic terms in the second subset of topic terms is less than the number of topic terms in the first subset of topic terms (Pathak, pa 0098, By favorably weighting a selected term, the compression component 138 promotes this term over other terms that are not similarly weighted, and increases the likelihood that the selected term will be included in the top K information items).
With respect to claims 12-14 and 22, the limitations are essentially the same as claims 1-3 and 11, and are rejected for the same reasons.
With respect to claims 23-25, the limitations are essentially the same as claims 1-3, and are rejected for the same reasons.
Claims 4, 10, 15, 21, 26, and 30 are rejected under 35 U.S.C. 103 as being unpatentable over Pathak in view of Agarwal, and further in view of Goto et al. (US 2023/0259711).
With respect to claim 4, Pathak in view of Agarwal teaches the non-transitory computer-readable medium of claim 1, as discussed above.
Goto teaches wherein the computer- readable instructions further cause the processor to compute the inverse document frequency weight value for each topic term using:
PNG
media_image1.png
36
490
media_image1.png
Greyscale
where IDFtopicterm is the inverse document frequency weight value for a topic term of the plurality of topic terms of a topic, total number of topics is a number of the plurality of topics, and number of topics containing the topicterm is the number of the plurality of topics that have the topic term included in the plurality of topic terms (Goto, pa 0035,
PNG
media_image2.png
86
446
media_image2.png
Greyscale
In Equation 3, the numerator |T| can refer to the total count of topics and the denominator can refer to the total count of topics containing token wi. The IDF is the log of the quotient.).
It would have been obvious at the effective filing date of the invention to a person having ordinary skill in the art to which said subject matter pertains to have modified Pathak in view of Agarwal with the teachings of Goto because it can indicate relatively more accurate, useful, and meaningful topic labels (Goto, pa 0032).
With respect to claim 10, Pathak in view of Agarwal teaches the non-transitory computer-readable medium of claim 1, as discussed above.
Goto teaches wherein the computer- readable instructions further cause the processor to compute the second weight value for each topic term in the first subset of topic terms of each topic by multiplying the first weight value of that topic term and the inverse document frequency weight value of that topic term (Goto, pa 0033, As shown in Equation 1, the TF-IDF can be the term frequency (TF) multiplied by the inverse document frequency (IDF).).
It would have been obvious at the effective filing date of the invention to a person having ordinary skill in the art to which said subject matter pertains to have modified Pathak in view of Agarwal with the teachings of Goto because it can indicate relatively more accurate, useful, and meaningful topic labels (Goto, pa 0032).
With respect to claims 12-15, 21, and 22, the limitations are essentially the same as claims 1-4, 10, and 11, and are rejected for the same reasons.
With respect to claims 23-26 and 30, the limitations are essentially the same as claims 1-4 and 10, and are rejected for the same reasons.
Claim(s) 5-8, 16-19, and 27-29 are rejected under 35 U.S.C. 103 as being unpatentable over Pathak in view of Agarwal, and further in view of Reza et al. (US 2023/0237277).
With respect to claim 5, Pathak in view of Agarwal teaches the non-transitory computer-readable medium of claim 1, as discussed above.
Reza teaches wherein to generate the compressed representation of the set of documents from the second subset of topic terms, the computer-readable instructions further cause the processor to concatenate the second subset of topic terms to generate a string for each topic (Reza, pa 0059, A domain adaptation algorithm may be used to train T5 to generate unique domain relevant features (DRFs; a set of keywords that characterize domain information) for each input. Then those DRFs can be concatenated with the input to form a template).
It would have been obvious at the effective filing date of the invention to a person having ordinary skill in the art to which said subject matter pertains to have modified Pathak in view of Agarwal with the teachings of Reza because it provides dynamic prompting which can be highly beneficial to develop a pre-trained model by appending the prompts to each set of input with an opinion and aspect. This will provide a better in-context learning and capture the opinion context information, which can lead to effective semantic information modelling (Reza, pa 0034).
With respect to claim 6, Pathak in view of Agarwal and Reza teaches the non-transitory computer-readable medium of claim 5, wherein the machine-consumable prompt for each topic comprises the string for that topic, an output definition defining a format for the topic label and the topic description for that topic, and one or more constraints (Reza, Fig. 1 & pa 0039, The original input 105 is then concatenated with the respective generated prompting template 115 to create a prompting function 120. For example, the original input 105 and the prompting template 115 may be used as input for a concatenate function configured to join the two text strings into a single text string: the original input 105; the prompting template 115, such that the two text strings are now linked or associated with one another.).
With respect to claim 7, Pathak in view of Agarwal, and Reza teaches the non-transitory computer-readable medium of claim 6, wherein the one or more constraints include a system role and a user role to provide a framework for how to generate the topic label and topic description for each topic (Reza, pa 0048, The training data may be acquired from the public domain or private domain. For example, a user such as a customer in the food and service industry may provide training data for fine-tuning a model to analyze sentiment in online food blog posts. & pa 0036, FIG. 1 is a block diagram illustrating the overall concept 100 of dynamic aspect based prompting and its influence on improving the confidence in a downstream task. As shown, original input 105 is obtained from a set of training data. The original input 105 is a text example such as “I like the food but not the service” from the set of training data. …The set of training data includes labels. The labels comprise: (i) text that relate to possible solutions for the given task to be learned by the model, and (ii) the specified solutions (e.g., a class identifier or ground truth for the text example). The labels may be provided by a user (e.g., a customer) and may be particular to a domain that the user intends to train the model within.).
With respect to claim 8, Pathak in view of Agarwal, and Reza teaches the non-transitory computer-readable medium of claim 6, wherein the one or more constraints further include a summary of what to include in the topic description (Reza, pa 0036, The labels may be provided by a user (e.g., a customer) and may be particular to a domain that the user intends to train the model within. For example, the text labels may be words such as terrible, bland, flavorful, delicious, disgusting, sour, sweet, poison, enjoyable, spicy, etc. that relate to various semantic classes (e.g., positive, negative, neutral, or the like) to be predicted for each text example within the domain of food. In other words, the original input 105 may include text that relates to possible solutions for the given task (e.g., the food was good but the service was bad—with good and bad being text that relate to possible sentiment solutions or classes);).
With respect to claims 16-19 and 27-29, the limitations are essentially the same as claims 5-8, and are rejected for the same reasons.
Claims 31 and 32 are rejected under 35 U.S.C. 103 as being unpatentable over Pathak et al. (US 2024/0394479), and further in view of Bayliss et al. (US 2025/0112878).
With respect to claim 31, Pathak teaches a computer-program product comprising a non-transitory computer-readable medium having computer-readable instructions stored thereon that, when executed by one or more processors, cause the one or more processors to:
execute an information extraction model over the set of documents to extract a plurality of text segments (Pathak, pa 0094, The keyword-extracting component 906 uses any rules-based logic (e.g., any algorithm) or machine-trained model to detect prominent keywords or named entities associated with the source information 904.);
execute an unsupervised topic model using at least a subset of the plurality of text segments to output a plurality of topics, each topic comprising (i) a plurality of topic terms (Pathak, pa 0097, The topic-modeling component 910 can likewise uses various rules-based logic and/or machine-trained models to extract topics associated with the source information 904, including Latent Dirichlet Allocation (LDA)) and (ii) a respective first weight value for each topic term (Pathak, pa 0094, the keyword-extracting component 906 can identify prominent words in the source information 904 using Term Frequency-Inverse Document Frequency (TF-IDF) or the TextRank algorithm);
compute an inverse document frequency value for each topic term of at least one topic of the plurality of topics based on the set of documents (Pathak, pa 0094, the keyword-extracting component 906 can identify prominent words in the source information 904 using Term Frequency-Inverse Document Frequency (TF-IDF));
compute a respective second weight value for each topic term of the at least one topic based on a combination of the respective first weight value and the inverse document frequency value (Pathak, pa 0094, the keyword-extracting component 906 can identify prominent words in the source information 904 using Term Frequency-Inverse Document Frequency (TF-IDF));
select a subset of topic terms of the at least one topic based on the respective second weight values (Pathak, pa 0092, once the compression component 138 is invoked, the compression-managing component 914 invokes all of the individual compression components (906, 908, 910, 912), which can then operate in parallel. & pa 0094, the keyword-extracting component 906 can identify prominent words in the source information 904 using Term Frequency-Inverse Document Frequency (TF-IDF));
construct a machine-consumable compressed prompt for the language model by:
ordering the subset of topic terms from a highest second weight value to a lowest second weight value (Pathak, pa 0098, By favorably weighting a selected term, the compression component 138 promotes this term over other terms that are not similarly weighted, and increases the likelihood that the selected term will be included in the top K information items), and
concatenating the ordered subset of topic terms into a token sequence (Pathak, pa 0059, the compression component 138 uses topic analysis to identify one or more topics that are pertinent to the source information. The compression component 138 has the effect of compressing the source information by using selected terms to describe it. …The dynamic prompt-generating component 128 uses the selected terms to compose the prompt information 124 for the current input query 118.).
Pathak doesn't expressly discuss receive a set of documents to be summarized by executing a language model having a bounded context window with a maximum token capacity; determine that a token count of the machine-consumable compressed prompt is less than or equal to the maximum token capacity of the bounded context window of the language model; in response to determining that the token count of the machine-consumable compressed prompt is less than or equal to the maximum token capacity, input the machine-consumable compressed prompt into the language model; and execute the language model based on the machine-consumable compressed prompt to generate a summary of the set of documents.
Bayliss teaches receive a set of documents to be summarized by executing a language model having a bounded context window with a maximum token capacity (Bayliss, Fig. 5, relevant portion of knowledge graph 502 & pa 0093, the prompt-engineering component 208 may determine which portion of the knowledge graph 122 is semantically most relevant, or has answers that are relevant for the query. The prompt-engineering component 208 may then provide the relevant portion 502 along with the query 128 to the LLM 118.);
determine that a token count of the machine-consumable compressed prompt is less than or equal to the maximum token capacity of the bounded context window of the language model (Bayless, Fig. 6 & pa 0096, In the initial prompt 604, the prompt-engineering component 208 may include the query 128 in the natural language, and also the assignments 302 that have been determined and stored in the knowledge graph 122. The response 606 provided by the LLM 118 indicates that the initial prompt 604 exceeds the context window size for the LLM 118. Accordingly, the prompt-engineering component 208 may generate a summary (potentially harnessing an LLM) of the assignments 302 and/or other data in the initial prompt to generate the prompt 608.);
in response to determining that the token count of the machine-consumable compressed prompt is less than or equal to the maximum token capacity, input the machine-consumable compressed prompt into the language model (Bayless, Fig. 6 & pa 0096, The LLM 118 is then able to provide a response 610 that includes the answers 134 that were desired by the prompt-engineering component 208.); and
execute the language model based on the machine-consumable compressed prompt to generate a summary of the set of documents (Bayless, Fig. 6 & pa 0096, The LLM 118 is then able to provide a response 610 that includes the answers 134 that were desired by the prompt-engineering component 208.).
It would have been obvious at the effective filing date of the invention to a person having ordinary skill in the art to which said subject matter pertains to have modified Pathak with the teachings of Bayliss because LLMs typically have context windows that represent how much information the LLMs can consider and the limits of context windows can be handled by providing the LLM with the most relevant information (Bayliss, pa 0029-0030).
With respect to claim 32, the limitations are essentially the same as claim 31, and are rejected for the same reasons.
Response to Arguments
35 U.S.C. 101
With regard to claims 1-30, the amendments to the claims have overcome the 35 U.S.C. 101 rejection. The Examiner withdraws the 35 U.S.C. 101 rejection to claims 1-30.
35 U.S.C. 103 rejections
Applicant argues that Pathak in view of Agarwal fails to teach “generating the compressed representation comprises selecting, ordering, and concatenating topic terms based on the weight values” because there is not a prompt artifact having this structural requirement (topic-term ordering/concatenation based on topic-term weights). The Examiner respectfully disagrees. Pathak teaches selecting the K top-ranked terms (pa 0098). This teaches selecting and ordering based on weight values by ordering terms weighted higher than others and selecting them. The terms are concatenated by creating a prompt to be submitted (pa 0045). Also see Fig. 1 for the analysis components 130 that create the prompt information 124.
Applicant argues that Pathak in view of Agarwal does not disclose multi-model chaining using outputs of an unsupervised model as inputs into a language model. The Examiner respectfully disagrees. The topic modeling component 910 uses rules-based logic such as Latent Dirichlet Allocation (LDA) which is an unsupervised model. Topic modeling component 910 is part of the compression component 914 and represented in Fig. 1 as 138. Fig. 1 shows that compression component 138 and language model 106 are separate, distinct models.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Kelly et al. (US 2024/0289560) teaches generating a contextual classification for the one or more request text fields and identifying a refined document subset based on the contextual classification and generating, using a large language model, one or more generative text fields using a generative model prompt based on the prompt document subset and the one or more request text fields.
Ailem et al. (US 2026/0004135) teaches using one or more large language models and the plurality of instructive, generate a plurality of annotated clusters, wherein an annotated cluster comprises a category annotation and a summary annotation.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BRITTANY N ALLEN whose telephone number is (571)270-3566. The examiner can normally be reached M-F 9 am - 5:00 pm 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, Sherief Badawi can be reached on 571-272-9782. 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.
/BRITTANY N ALLEN/ Primary Examiner, Art Unit 2169