DETAILED ACTION
This communication is in response to the Application filed on 11/07/2024. Claims 1-14 are pending and have been examined.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 13, 2013, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 11/07/2024 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Priority
Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d). The certified copy has been filed in parent Application No. EP 232085795, filed on 11/08/2023.
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph:
(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and
(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function.
Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function.
Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action.
This application includes one or more claim limitations that do not use the word “means,” but are nonetheless being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, because the claim limitation(s) uses a generic placeholder that is coupled with functional language without reciting sufficient structure to perform the recited function and the generic placeholder is not preceded by a structural modifier. Such claim limitation(s) is/are: “a user interaction recording module” in claim 12.
Because this/these claim limitation(s) is/are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, it/they is/are being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof.
If applicant does not intend to have this/these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitation(s) recite(s) sufficient structure to perform the claimed function so as to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-8 rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Claim 1 recites A [computer]-implemented method for generating domain specific training data for a [large language model], the computer-implemented method comprising: providing a domain specific ontology relating to a domain; providing domain specific information relating to the domain; and processing the domain specific information in a data processing-pipeline for structuring data for training of the large language model, wherein the domain specific ontology is provided as a recognition pattern in a step of the data processing-pipeline, such that the structured training data comprises domain specific ontology annotations.
The limitations in this claim, as drafted, are a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. A human is capable of providing information, such as a text document, and ontology information, such as a reference table for domain specific terms. A processing-pipeline can be represented by the steps the human mind takes to interpret data, in this case it could be someone using the reference table to better understand the document and identify which text could be useful. The selection of data for training an LLM is a design decision that the human mind is capable of doing. A recognition pattern that results in ontology annotations is the human equivalent of using recognized terms from a reference table to label domain specific terminology in a document which can be done using pen and paper. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claim recites the additional components of a computer and a large language model. The computer is merely used to apply the mental process. The computer is described in paragraph 39 of the specification with a generic description of the device. The large language model is merely used for the post-solution activity of having the created data by input into it. Paragraph 6 of the specification describes the LLM and gives no specific implantation or description of the component. Accordingly, the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claim is not patent eligible.
Claim 2 recites The computer-implemented method of claim 1, wherein the domain specific information comprises domain specific text documents relating to the domain, wherein the processing-pipeline is a natural language processing pipeline NPP for structuring data for training of the large language model, and wherein the domain specific ontology is provided as a recognition pattern in a step of named entity recognizing of the natural language processing pipeline NPP, such that the structured training data comprises domain specific ontology annotations.
The limitations in this claim, as drafted, are a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. A human can interpret a text document and structure it in a different way. The decision to use that training data for an LLM is a design decision the human mind is capable of making. A human can perform named entity recognition by using a reference of domain specific terminology to annotate the text document using pen and paper. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claim does not recite any additional components that were not present in the independent claim. Accordingly, the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claim is not patent eligible.
Claim 3 recites The computer-implemented method of claim 1, wherein the domain specific training data is software application specific training data for the large language model, wherein the software application is related to the domain of a specific factual context, a technical domain, a physical domain, a technical and physical domain, or any combination thereof, wherein the domain specific information is provided as a software application computer program, source code, an API definition of the software application, or any combination thereof, wherein the processing-pipeline is a code parsing pipeline for structuring data for training of the large language model, and wherein the domain specific ontology is provided as a recognition pattern in a step of semantic analyzing of the code parsing pipeline such that the structured training data comprises domain specific ontology annotations.
The limitations in this claim, as drafted, are a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. A human is capable of taking software information and relating it to a specific domain, for instance, looking at handwritten code and identifying what language it was written in. This code could be directly copied from something a software app, source code, or API definition. A human is capable of restructuring something written in code-based syntax to be training data, for example, splitting it up into smaller, tagged portions. A human could also use a reference guide to analyze the semantics of the syntax and label terms according to the reference. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claim does not recite any additional components that were not present in the independent claim. Accordingly, the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claim is not patent eligible.
Claim 4 recites The computer-implemented method of claim 2, wherein the natural language processing pipeline NPP comprises: preprocessing, tokenizing, part-of-speech-tagging, named entity recognizing, and post-processing.
The limitations in this claim, as drafted, are a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. These steps can all be performed by the human mind. Pre-processing can be any form preparing a document such as changing the structure it’s written in. Tokenizing can be segmenting the text into individual pieces that each have separate meanings. POS tagging can just be labelling elements of the text to provide more explanation. Named entity recognition can be identifying any terms that belong to a specific domain. Post-processing can be converting the format of the writing once again after the other steps have been performed. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claim does not recite any additional components that were not present in the independent claim. Accordingly, the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claim is not patent eligible.
Claim 5 recites The computer-implemented method of claim 3, wherein the code parsing pipeline comprises: tokenizing, parsing, semantic-analyzing, and post-processing.
The limitations in this claim, as drafted, are a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. A human can perform these steps on a piece of handwritten code. Tokenizing can be segmenting the code into individual pieces that each have separate meanings. Parsing can be manually organizing the syntax into a syntax tree structure on another piece of paper. Semantic-analyzing can be using a reference to annotate domain specific terms within the syntax tree. Post-processing can be converting the format of the writing after the other steps have been performed. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claim does not recite any additional components that were not present in the independent claim. Accordingly, the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claim is not patent eligible.
Claim 6 recites The computer-implemented method of claim 1, further comprising: training the large language model using the structured training data comprising domain specific ontology annotations.
The limitations in this claim, as drafted, are a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. The type of data used to train a large language model is a design decision that the human mind is capable of making. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claim does not recite any additional components that were not present in the independent claim. Accordingly, the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claim is not patent eligible.
Claim 7 recites A [computer]-implemented method for generating a text report from a data file, the computer-implemented method comprising: generating domain specific training data for a [large language model], the generating of the domain specific training data comprising: providing a domain specific ontology relating to a domain; providing domain specific information relating to the domain; and processing the domain specific information in a data processing-pipeline for structuring data for training of the large language model, wherein the domain specific ontology is provided as a recognition pattern in a step of the data processing-pipeline, such that the structured training data comprises domain specific ontology annotations, and wherein the data file contains passages that are formulated in computer semantics; processing the data file by the trained large language model; and generating a human-readable text report.
The limitations in this claim, as drafted, are a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. A human is capable of creating text reports and/or training data. A human is capable of providing information, such as a text document, and ontology information, such as a reference table for domain specific terms. A processing-pipeline can be represented by the steps the human mind takes to interpret data, in this case it could be someone using the reference table to better understand the document and identify which text could be useful. The selection of data for training an LLM is a design decision that the human mind is capable of doing. A recognition pattern that results in ontology annotations is the human equivalent of using recognized terms from a reference table to label domain specific terminology in a document which can be done using pen and paper. This could be done for a piece of text written in computer semantics such as the syntax of a coding language. A human is then capable of using this gathered information to create a text report based on the computer semantics. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claim recites the additional components of a computer and a large language model. The computer is merely used to apply the mental process. The computer is described in paragraph 39 of the specification with a generic description of the device. The large language model is merely used to apply the task of text creation via a computer, which is something a human can do already. Paragraph 6 of the specification describes the LLM and gives no specific implantation or description of the component. Accordingly, the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claim is not patent eligible.
Claim 8 recites The computer-implemented method of claim 7, wherein the text report is of a user interaction with a software application running on a computer, wherein the computer-implemented method further comprises: recording of user interaction, wherein the data file contains recordings of the user interaction; processing the data file that contains the recordings of the user interaction by the trained large language model; and generating a user action report.
The limitations in this claim, as drafted, are a process that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. A human is capable of recording an interaction with a software application, for example, watching someone use the software and writing down what they’re doing. A human is than capable of using what they reordered to create a report of the interaction. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. The claim does not recite any additional components that were not present in the independent claim. Accordingly, the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claim is not patent eligible.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1, 2, 4, 6, and 11 are rejected under 35 U.S.C. 102(a)(2) as being anticipated by US Patent Publication 11687827 B2 (Pondicherry Murugappan et al.) (Hence fourth, Pondicherry).
Regarding Claim 1, Pondicherry teaches A computer-implemented method for generating domain specific training data for a large language model, the computer-implemented method comprising:
(Accordingly, a domain-specific regulatory text corpus is selected to provide training data for training the ML models in order to extract information from the regulatory text accurately.) (Col. 2, Lines 52-55).
(FIG. 3 shows a networked computer system 300 that can be employed to implement the data processing system 100 in accordance with the examples disclosed herein.) (Col. 10, Lines 8-10).
providing a domain specific ontology relating to a domain;
(Accordingly a domain-specific regulatory text corpus is selected to provide training data for training the ML models in order to extract information from the regulatory text accurately. The regulatory text corpus includes prior regulatory documents pertaining to the specific domain. Whenever new regulations are published by a governing body corresponding to a specific domain, the data processing system can be configured to receive the domain-specific regulatory text document either via a user upload or automatically via monitoring various communication channels.) (Col. 2, Lines 51-62).
A text corpus is built to form the domain specific ontology for processing texts.
providing domain specific information relating to the domain;
(The topic extractor 110 includes a topic extraction model 112 for topic extraction modelling prediction. The topic extractor 110 can predict with a certain confidence level, the relationships for a given input regulatory document e.g., the domain-specific document 150 with existing regulations.) (Col. 6, Lines 49-53).
The system receives a domain specific document as an input which is considered information.
and processing the domain specific information in a data processing-pipeline for structuring data for training of the large language model,
(The topic extraction model is based on Latent Semantic Indexing (LSI) and is trained via unsupervised learning on the prior domain-specific documents in the regulatory text corpus. Identification of the entities in the domain-specific document includes initially obtaining linguistic features from the textual content of the domain-specific document. The linguistic features are obtained using an entity feature selection model. The entity feature selection model is based on a sequence labelling technique and learns to select features via supervised learning on annotated training data. The linguistic features thus obtained are employed by the entity identification model to identify the entities in the domain-specific document as belonging to entities of the entities identification model.) (Col. 3, Lines 32-45).
(The entity feature selection model 136 is trained on the feature training data 294 to extract various linguistic features of a body of text at token, sentence and document levels for identification of entities by the entity extractor 130.) (Col. 8, Lines 66-67).
A pipeline of models is used to process documents in which previously processed documents are used in training data for the models.
wherein the domain specific ontology is provided as a recognition pattern in a step of the data processing-pipeline,
(Again, training the entity identification model 132 on domain-specific training data as provided by the regulatory text corpus 190 produces a trained entity identification model and enables the entity extractor 130 to produce more accurate output. In an example, the entity identification model 132 can be trained in named entity recognition (NER) for producing the output 160 that includes the entities/keywords (nouns). In an example, supervised learning techniques as detailed herein can be implemented in training the entity identification model 132.) (Col. 7, Lines 30-39).
The domain specific document corpus is used in a named entity recognition step of the processing pipeline.
such that the structured training data comprises domain specific ontology annotations.
(The entity identification model 132 is trained by the model trainer 194 which provides annotated entity training data 296 from the regulatory text corpus 190.) (Col. 9, Lines 15-17).
Training data is annotated according to the named entities.
Regarding Claim 2, Pondicherry teaches the system of claim 1.
Pondicherry further teaches wherein the domain specific information comprises domain specific text documents relating to the domain,
(The data processing system 100 extracts data from regulatory textual documents and automatically converts the textual data into an output 160 that can include one or more of topics, entities, process rules, requirements and definitions that can be consumed by the downstream processes for further implementations such as RPA.) (Col. 5, Lines 40-45).
Information is in the form of text documents.
wherein the processing-pipeline is a natural language processing pipeline NPP for structuring data for training of the large language model,
(The topics, entities and sections thus extracted can be employed for various purposes such as for interpreting regulatory texts by applying NLP and ML techniques.) (Col. 7, Lines 52-54).
(The entity identification model 132 is trained by the model trainer 194 which provides annotated entity training data 296 from the regulatory text corpus 190.) (Col. 9, Lines 15-17).
The pipeline employs NLP techniques to create training data.
and wherein the domain specific ontology is provided as a recognition pattern in a step of named entity recognizing of the natural language processing pipeline NPP,
(Again, training the entity identification model 132 on domain-specific training data as provided by the regulatory text corpus 190 produces a trained entity identification model and enables the entity extractor 130 to produce more accurate output. In an example, the entity identification model 132 can be trained in named entity recognition (NER) for producing the output 160 that includes the entities/keywords (nouns). In an example, supervised learning techniques as detailed herein can be implemented in training the entity identification model 132.) (Col. 7, Lines 30-39).
Named entity recognition is performed.
such that the structured training data comprises domain specific ontology annotations.
(The entity identification model 132 is trained by the model trainer 194 which provides annotated entity training data 296 from the regulatory text corpus 190.) (Col. 9, Lines 15-17).
The training data is annotated based on the NER.
Regarding Claim 4, Pondicherry teaches the system of claim 2.
Pondicherry further teaches wherein the natural language processing pipeline NPP comprises: preprocessing, tokenizing, part-of-speech-tagging, named entity recognizing, and post-processing.
(In an example, the entity identification model 132 can be trained in named entity recognition (NER) for producing the output 160 that includes the entities/keywords (nouns).) (Col. 7, Lines 30-39).
(The entity feature selection model 136 is trained on the feature training data 294 to extract various linguistic features of a body of text at token, sentence and document levels for identification of entities by the entity extractor 130. The feature training data 294 includes training data wherein the different linguistic features in a large volume of regulatory documents are annotated or labelled. Various features are extracted using syntactic component POS tags, corpus feature components like bag-of-words, Term Frequency-Inverse Document Frequency (TF-IDF) and language modelling component (n-gram), etc.) (Col. 8, Line 66 to Col. 9, Line 9).
The input text is broken down into features at the token level (tokenized) and input into an entity extractor model (pre-processing required to put in machine-readable format). Features are extracted using POS tags and entities are identified using NER. Finally, the information is used as training data for multiple models meaning post-processing is performed to format the data for this purpose.
Regarding Claim 6, Pondicherry teaches the system of claim 1.
Pondicherry further teaches further comprising: training the large language model using the structured training data comprising domain specific ontology annotations.
(The entity feature selection model is based on a sequence labelling technique and learns to select features via supervised learning on annotated training data. The linguistic features thus obtained are employed by the entity identification model to identify the entities in the domain-specific document as belonging to entities of the entities identification model. The entity identification model itself is based on conditional random fields (CRF) methodology and can be trained on entity training data which includes the prior domain-specific documents with the various entities annotated and classified. The section identification also uses linguistic feature extraction from the regulatory text of the received domain-specific document. However, a section feature selection model which is employed for section identification is based on a classification technique and is also trained on labelled training data with the linguistic features annotated on a subset of the prior domain-specific documents. The section classification model which classifies portions of the domain-specific document as belonging to one of a plurality of predetermined domain-specific sections is based on Multinomial Naïve Bayes (MNB) classification type algorithm. The MNB classification type algorithm is trained via supervised learning on section training data which includes a subset of the domain-specific documents in the regulatory text corpus with annotated sections.) (Col. 3, Lines 39-64).
The models used for processing future documents are trained on the created training data.
Regarding Claim 11, Pondicherry teaches A machine comprising: a user input interface;
(FIG. 9 shows a graphical user interface (GUI) that displays output from a document processor in accordance with the examples disclosed herein.) (Col. 2, Lines1-3).
a user interaction recording module for recording user interaction into a data file;
(Whenever new regulations are published by a governing body corresponding to a specific domain, the data processing system can be configured to receive the domain-specific regulatory text document either via a user upload or automatically via monitoring various communication channels.) (Col. 2, Lines 57-62).
a processor configured to:
(The computer system 1400 includes processor(s) 1402, such as a central processing unit, ASIC or other type of processing circuit) (Col. 16, Lines 5-15).
generate a text report from a data file,
(FIG. 9 shows a GUI 900 that displays the output 160 from the data processing system 100 in accordance with the examples disclosed herein. The output 160 includes identities of related documents 902 as determined by the topic extractor 110, the related topics 904 obtained by the entity extractor 130 for a received input document such as the domain-specific document 150.) (Col. 14, Lines 3-9).
Fig. 9 shows an example of a text report output.
the generation of the text report comprising: generation of domain specific training data for a large language model, provision of a domain specific ontology relating to a domain;
(Accordingly a domain-specific regulatory text corpus is selected to provide training data for training the ML models in order to extract information from the regulatory text accurately. The regulatory text corpus includes prior regulatory documents pertaining to the specific domain. Whenever new regulations are published by a governing body corresponding to a specific domain, the data processing system can be configured to receive the domain-specific regulatory text document either via a user upload or automatically via monitoring various communication channels.) (Col. 2, Lines 51-62).
A text corpus is built to form the domain specific ontology for processing texts.
provision of a domain specific information relating to the domain;
(The topic extractor 110 includes a topic extraction model 112 for topic extraction modelling prediction. The topic extractor 110 can predict with a certain confidence level, the relationships for a given input regulatory document e.g., the domain-specific document 150 with existing regulations.) (Col. 6, Lines 49-53).
The system receives a domain specific document as an input which is considered information.
and processing the domain specific information in a data processing-pipeline for structuring data for training of the large language model,
(The topic extraction model is based on Latent Semantic Indexing (LSI) and is trained via unsupervised learning on the prior domain-specific documents in the regulatory text corpus. Identification of the entities in the domain-specific document includes initially obtaining linguistic features from the textual content of the domain-specific document. The linguistic features are obtained using an entity feature selection model. The entity feature selection model is based on a sequence labelling technique and learns to select features via supervised learning on annotated training data. The linguistic features thus obtained are employed by the entity identification model to identify the entities in the domain-specific document as belonging to entities of the entities identification model.) (Col. 3, Lines 32-45).
(The entity feature selection model 136 is trained on the feature training data 294 to extract various linguistic features of a body of text at token, sentence and document levels for identification of entities by the entity extractor 130.) (Col. 8, Lines 66-67).
A pipeline of models is used to process documents in which previously processed documents are used in training data for the models.
wherein the domain specific ontology is provided as a recognition pattern in a step of the data processing-pipeline,
(Again, training the entity identification model 132 on domain-specific training data as provided by the regulatory text corpus 190 produces a trained entity identification model and enables the entity extractor 130 to produce more accurate output. In an example, the entity identification model 132 can be trained in named entity recognition (NER) for producing the output 160 that includes the entities/keywords (nouns). In an example, supervised learning techniques as detailed herein can be implemented in training the entity identification model 132.) (Col. 7, Lines 30-39).
The domain specific document corpus is used in a named entity recognition step of the processing pipeline.
such that the structured training data comprises domain specific ontology annotations.
(The entity identification model 132 is trained by the model trainer 194 which provides annotated entity training data 296 from the regulatory text corpus 190.) (Col. 9, Lines 15-17).
Training data is annotated according to the named entities.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 3, 5, and 7 are rejected under 35 U.S.C. 103 as being unpatentable over US Patent Publication 11687827 B2 (Pondicherry Murugappan et al.) (Hence fourth, Pondicherry) in view of US Patent Publication US 20220261241 A1 (Balasubramanian et al.).
Regarding Claim 3, Pondicherry teaches the system of claim 1.
Pondicherry does not explicitly teach: wherein the domain specific training data is software application specific training data for the large language model, wherein the software application is related to the domain of a specific factual context, a technical domain, a physical domain, a technical and physical domain, or any combination thereof, wherein the domain specific information is provided as a software application computer program, source code, an API definition of the software application, or any combination thereof, wherein the processing-pipeline is a code parsing pipeline for structuring data for training of the large language model, and wherein the domain specific ontology is provided as a recognition pattern in a step of semantic analyzing of the code parsing pipeline such that the structured training data comprises domain specific ontology annotations.
However, Balasubramanian et al. teaches wherein the domain specific training data is software application specific training data for the large language model,
(It uses the data services to create the training data for training these models. These models are used to generate the documentation for a program statement line of code and to predict the documentation for a given function code snippet.) (Paragraph 23).
wherein the software application is related to the domain of a specific factual context, a technical domain, a physical domain, a technical and physical domain, or any combination thereof,
(Therefore, the NL Summarizer Service 112 extracts meaningful sentences from a given paragraph by evaluating their structure, subject and context using NLP techniques and encodes the extracted sentences and summarizing the extracted sentences using the machine learnt summarizer model for software domain.) (Paragraph 69).
The input has an associated technical domain related to the software.
wherein the domain specific information is provided as a software application computer program, source code, an API definition of the software application, or any combination thereof,
(The operations include parsing a source code file to extract a function and generate an abstract syntax tree, generating first natural language documentation for each of a plurality of program statements within the function using a programming language neural network model, generating second natural language documentation for the function as a whole by processing a code snippet of the function using a function documentation neural network model, consolidating the first natural language documentation and the second natural language documentation at a function level, a source file level, and a project level to create consolidated natural language documentation, and summarizing multiple sentences of the consolidated natural language documentation into an abstract summary of the source code file by applying a set of rules.) (Paragraph 29).
The input is a source code file.
wherein the processing-pipeline is a code parsing pipeline for structuring data for training of the large language model,
(In some embodiments, the operations include building the programming language neural network model by extracting data by reading language specifications and reference documentation, preparing training data comprising key constructs of a programming language comprising at least one of syntax expression, functions, function signatures, programming language key words, and associated documentation, encoding the training data for training the programming language neural network model using a machine learning algorithm, and building the programming language neural network model using the training data and saving the programming language neural network model to file storage.) (Paragraph 33).
A processing-pipeline is performed on the input to create training data.
and wherein the domain specific ontology is provided as a recognition pattern in a step of semantic analyzing of the code parsing pipeline such that the structured training data comprises domain specific ontology annotations.
(In some embodiments, parsing the source code file to extract the function and generate the abstract syntax tree includes detecting a programming language and syntax of the source code file, parsing the source code file to extract the code snippet of the function from the source code file, generating an abstract syntax tree and a parse tree along with the code snippet of the function, and clean the code snippet of the function to remove non-executable content comprising one or more comments.) (Paragraph 32).
The input is used to build an abstract syntax tree as a form of ontology used for a recognition pattern.
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify the training data processing pipeline as taught by as taught by Pondicherry to also be capable of performing this process on software-based inputs such as source code as taught by Balasubramanian et al. This would have been an obvious improvement to expand available inputs to include source code which is something that commonly struggles with poor documentation (Balasubramanian et al. Paragraph 4).
Regarding Claim 5, Pondicherry in view of Balasubramanian et al. teaches the system of claim 3.
Furthermore, Balasubramanian et al. teaches wherein the code parsing pipeline comprises: tokenizing, parsing, semantic-analyzing, and post-processing.
(In some embodiments, the operations include building the function documentation neural network model by extracting functions and associated documentation from a set of training source code files, evaluating the functions and the associated documentation relative to predetermined quality thresholds comprising at least one of a number of sentences, semantic quality of the documentation, and a date at which a corresponding training source code file was most recently updated, creating training data comprising multiple representations of source code of the training source code files by substituting variable names with auto-generated synthetic names, encoding the training data for training the function documentation neural network model using a the machine learning algorithm, and building the function documentation neural network model and saving the function documentation neural network model to file storage.) (Paragraph 34).
(In some embodiments, the operations include refining the consolidated natural language documentation by assessing parameters of the consolidated natural language documentation using rules and natural language processing techniques, the parameters comprising at least one of semantic coherence, clarity, and conciseness.) (Paragraph 39).
(The Code Processing Service 109 reads the source file and detects the programming language and the syntax. After detecting the language, the Code Processing Service 109 uses the appropriate parser to parse the source file to extract the functions code snippets from the source file. The Code Processing Service 109 also generates the abstract syntax tree and parse tree along with the function code snippet using a parser loaded with the programming language syntax rules. The function signature is extracted and stored along with the function code snippet and the abstract syntax tree. The Code Processing Service 109 also cleans up the code snippet to remove comments and other non-executable content.) (Paragraph 62).
The process of creating training data uses many steps including encoding (tokenizing), identifying building an abstract syntax tree (parsing), identifying variable names and replacing them (semantic analyzing), and cleaning up code snippets by removing non-executable content before creating training data (post-processing).
Regarding Claim 7, Pondicherry teaches A computer-implemented method for generating a text report from a data file, the computer-implemented method comprising:
(In an example, the output from the document processor can be used to generate reports to meet the reporting requirements associated with the new regulations.) (Col. 4, Lines 21-23).
(FIG. 3 shows a networked computer system 300 that can be employed to implement the data processing system 100 in accordance with the examples disclosed herein.) (Col. 10, Lines 8-10).
generating domain specific training data for a large language model,
(A model trainer 194 can supply the appropriate training data as detailed herein to train the plurality of ML models on a regulatory text corpus 190 which is designed for improving the models' prediction confidence level. In an example, the regulatory text corpus 190 can pertain to a specific domain wherein a domain-specific data dictionary can also be built within the data processing system 100.) (Col. 6, Lines 29-37).
the generating of the domain specific training data comprising: providing a domain specific ontology relating to a domain;
(Accordingly a domain-specific regulatory text corpus is selected to provide training data for training the ML models in order to extract information from the regulatory text accurately. The regulatory text corpus includes prior regulatory documents pertaining to the specific domain. Whenever new regulations are published by a governing body corresponding to a specific domain, the data processing system can be configured to receive the domain-specific regulatory text document either via a user upload or automatically via monitoring various communication channels.) (Col. 2, Lines 51-62).
A text corpus is built to form the domain specific ontology for processing texts.
providing domain specific information relating to the domain;
(The topic extractor 110 includes a topic extraction model 112 for topic extraction modelling prediction. The topic extractor 110 can predict with a certain confidence level, the relationships for a given input regulatory document e.g., the domain-specific document 150 with existing regulations.) (Col. 6, Lines 49-53).
The system receives a domain specific document as an input which is considered information.
and processing the domain specific information in a data processing-pipeline for structuring data for training of the large language model,
(The topic extraction model is based on Latent Semantic Indexing (LSI) and is trained via unsupervised learning on the prior domain-specific documents in the regulatory text corpus. Identification of the entities in the domain-specific document includes initially obtaining linguistic features from the textual content of the domain-specific document. The linguistic features are obtained using an entity feature selection model. The entity feature selection model is based on a sequence labelling technique and learns to select features via supervised learning on annotated training data. The linguistic features thus obtained are employed by the entity identification model to identify the entities in the domain-specific document as belonging to entities of the entities identification model.) (Col. 3, Lines 32-45).
(The entity feature selection model 136 is trained on the feature training data 294 to extract various linguistic features of a body of text at token, sentence and document levels for identification of entities by the entity extractor 130.) (Col. 8, Lines 66-67).
A pipeline of models is used to process documents in which previously processed documents are used in training data for the models.
wherein the domain specific ontology is provided as a recognition pattern in a step of the data processing-pipeline,
(Again, training the entity identification model 132 on domain-specific training data as provided by the regulatory text corpus 190 produces a trained entity identification model and enables the entity extractor 130 to produce more accurate output. In an example, the entity identification model 132 can be trained in named entity recognition (NER) for producing the output 160 that includes the entities/keywords (nouns). In an example, supervised learning techniques as detailed herein can be implemented in training the entity identification model 132.) (Col. 7, Lines 30-39).
The domain specific document corpus is used in a named entity recognition step of the processing pipeline.
such that the structured training data comprises domain specific ontology annotations.
(The entity identification model 132 is trained by the model trainer 194 which provides annotated entity training data 296 from the regulatory text corpus 190.) (Col. 9, Lines 15-17).
Training data is annotated according to the named entities.
Pondicherry does not explicitly teach: and wherein the data file contains passages that are formulated in computer semantics; processing the data file by the trained large language model; and generating a human-readable text report.
However, Balasubramanian et al. teaches and wherein the data file contains passages that are formulated in computer semantics;
(The downloaded source code goes as an input to the Code Processing Service 109. The Code Processing Service 109 reads the source file and detects the programming language and the syntax.) (Paragraph 62).
Source code (computer semantics) are taken in as input.
processing the data file by the trained large language model;
(The Documentation Processing service 113 feeds this individual program statement line of code to the NL Program Statement Service 110 to generate the documentation for the each of the program statement line of code via a natural language neural network trained model.) (Paragraph 64).
(In some embodiments, the operations include building the function documentation neural network model by extracting functions and associated documentation from a set of training source code files, evaluating the functions and the associated documentation relative to predetermined quality thresholds comprising at least one of a number of sentences, semantic quality of the documentation, and a date at which a corresponding training source code file was most recently updated, creating training data comprising multiple representations of source code of the training source code files by substituting variable names with auto-generated synthetic names, encoding the training data for training the function documentation neural network model using a the machine learning algorithm, and building the function documentation neural network model and saving the function documentation neural network model to file storage.) (Paragraph 34).
Training data is created for natural language neural network
and generating a human-readable text report.
(The Documentation Processing service 113 feeds this individual program statement line of code to the NL Program Statement Service 110 to generate the documentation for the each of the program statement line of code via a natural language neural network trained model.) (Paragraph 64).
Documentation is generated for source code received as input.
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify the training data processing pipeline as taught by as taught by Pondicherry to also be capable of performing this process on software-based inputs such as source code as taught by Balasubramanian et al. This would have been an obvious improvement to expand available inputs to include source code which is something that commonly struggles with poor documentation (Balasubramanian et al. Paragraph 4).
Claim 8 are rejected under 35 U.S.C. 103 as being unpatentable over US Patent Publication 11687827 B2 (Pondicherry Murugappan et al.) (Hence fourth, Pondicherry) in view of US Patent Application Publication US 20220261241 A1 (Balasubramanian et al.) and further in view of US Patent Publication US 20100229112 A1 (Ergan et al.).
Regarding Claim 8, Pondicherry in view of Balasubramanian et al. teaches the system of claim 7.
Pondicherry in view of Balasubramanian et al. does not explicitly teach: wherein the text report is of a user interaction with a software application running on a computer, wherein the computer-implemented method further comprises: recording of user interaction, wherein the data file contains recordings of the user interaction; processing the data file that contains the recordings of the user interaction by the trained large language model; and generating a user action report.
However, Ergan et al. teaches wherein the text report is of a user interaction with a software application running on a computer,
(The tool records user interactions, including input events such as mouse and keyboard commands and metadata about the objects within a graphical user interface with which these interactions occur. Upon occurrence of the stop event, the tool then creates a report containing an event record from recorded user interactions.) (Paragraph 29).
A user interaction with software is recorded.
wherein the computer-implemented method further comprises: recording of user interaction, wherein the data file contains recordings of the user interaction;
(In some embodiments, recording tool 230 may capture and store indications of user interactions during "sessions" whenever the tool is enabled. The recording tool 230 may record user interaction in a recording session. If an event to be reported occurs during a session, the tool 230 may generate an appropriate report based on recorded user interactions.) (Paragraph 52).
The recording is stored in case a report is needed.
processing the data file that contains the recordings of the user interaction by (the trained large language model) (taught by Pondicherry).
(The compressor 236 may optionally be configured to translate the relevant event records from the event log 235 into a human readable format such as a natural language of the user. For example, an event record in the event log 235 may be translated into simple language such as "user left click on notepad from the start menu.") (Paragraph 73).
The record events are converted to a natural language format.
and generating a user action report.
(The report received by the back end server at step 722 contains a number of event records corresponding to a sequence of user interactions in the context of a graphical user interface. The event records include information about the user input and metadata describing the user interaction. In some embodiments, the report may also include screen shot information.) (Paragraph 102).
A report is generated with the details from the user interaction recording.
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify the training data processing pipeline as taught by as taught by Pondicherry in view of Balasubramanian et al. to also be capable of performing this process on user-interactions with software as taught by Ergan et al. This would have been an obvious improvement to assist in resolving software errors where as a detailed report can be made based on the last user interaction. (Ergan et al. Paragraph 7).
Allowable Subject Matter
Claims 9 and 10 objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Claims 12-14 allowed.
The following is an examiner’s statement of reasons for allowance:
The closest prior art of reference is US Patent Publication 11687827 B2 (Pondicherry Murugappan et al.) in view of US Patent Application Publication US 20220261241 A1 (Balasubramanian et al.) and further in view of US Patent Publication US 20100229112 A1 (Ergan et al.).
Pondicherry Murugappan et al. teaches A machine comprising: a user input interface; (Col. 2, Lines 1-3).
a processor configured to: (Col. 16, Lines 5-15).
generate a text report from a data file, the generation of the text report comprising: generation of domain specific training data for a large language model, (Col. 2, Lines 51-62).
the generation of the domain specific training data comprising: provision of a domain specific ontology relating to a domain; (Col. 2, Lines 51-62).
provision of a domain specific information relating to the domain; (Col. 6, Lines 49-53).
and process of the domain specific information in a data processing-pipeline for structuring data for training of the large language model, (Col. 3, Lines 32-45) (Col. 8, Lines 66-67).
wherein the domain specific ontology is provided as a recognition pattern in a step of the data processing-pipeline, such that the structured training data comprises domain specific ontology annotations, (Col. 7, Lines 30-39).
Pondicherry Murugappan et al. does not teach a user interaction recording module for recording user interaction into a data file; and wherein the data file contains passages that are formulated in computer semantics; process of the data file by the trained large language model; generation of a human-readable text report, wherein the human-readable text report is of an automation process of a machine being controlled by the automation process implemented as a programmable logic controller language file; generation of the data file in a programmable logic controller semantics from the programmable logic controller language file; and process of the data file by the trained large language model and generating the text report of the automation process; and an output interface configured to output the user action report.
Balasubramanian et al. teaches and wherein the data file contains passages that are formulated in computer semantics; (Paragraph 62).
process of the data file by the trained large language model; (Paragraph 64).
Ergan et al. teaches a user interaction recording module for recording user interaction into a data file; (Paragraph 29).
However, none of the prior art references, either alone or in combination teach the limitations of independent claim 12.
Any comments considered necessary by applicant must be submitted no later than the payment of the issue fee and, to avoid processing delays, should preferably accompany the issue fee. Such submissions should be clearly labeled “Comments on Statement of Reasons for Allowance.”
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NICHOLAS DANIEL LOWEN whose telephone number is (571)272-5828. The examiner can normally be reached Mon-Fri 8:00am - 4:00pm.
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, Paras D Shah can be reached at (571) 270-1650. 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.
/NICHOLAS D LOWEN/Examiner, Art Unit 2653
/Paras D Shah/Supervisory Patent Examiner, Art Unit 2653
06/15/2026
Prosecution Insights