FINAL OFFICE ACTION
Status of the Claims
Claims 1-10, 12-14, 16-18, and 20-23 are rejected under 35 U.S.C. 103
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 1-10, 12-14, 16-18, and 20-23 are rejected under 35 U.S.C. 103 as being unpatentable over Agrawal et al. (U.S. Publication No. 2025/0147832 A1), hereinafter referred to as Agrawal, in view of Joshi et al. (U.S. Publication No. 2025/0036800 A1), hereinafter referred to as Joshi, in further view of Crabtree et al. (U.S. Publication No. 2025/0259144 A1), hereinafter referred to as Crabtree.
Regarding Claim 1, Agrawal teaches:
A device, comprising:
at least one processor; and at least one memory that stores executable instructions that, when executed by the at least one processor, facilitate performance of operations, comprising: ([0142]);
receiving a group of alert messages within a predefined time window, wherein each alert message of the group of alert messages comprises: ([0036]; regarding, “the system may receive or access a log (e.g., an error log) that includes one or more error messages.”; [0038]; regarding, “The log may be generated by a software when implementing an application or a service (e.g., running a data processing pipeline, compiling a software package, or the like).”)
a respective description of a respective alert associated with the alert message in a text modality, and respective timestamp data indicative of a respective time at which the alert message was generated; ([0038]; regarding, “The log may include information related to a code error (also referred to as “an error”), such as an error message that informs the error, a type of the error (e.g., compile time error, run time error, a syntax error, an overflow error, or the like), the name and or file-path of the code associated with the error, timestamps associated with the error, or other information related to the error.”);
for each alert message of the group of alert messages, based on the respective timestamp data, retrieving, from a telemetry store… respective temporal context data that are stored according to a temporal modality, ([0036]; regarding, “the system may receive or access a log (e.g., an error log) that includes one or more error messages… Based on the log and/or the error message, the system may further search and determine a context associated with the error. The context may include the code, portions of the log that are close or more related to the error message, portions of one or more documents associated with the code, and/or additional information (e.g., ontology associated with a service that utilizes the code, search results from search engines) relevant to the error or useful for one or more LLMs to explain the error message. For example, the additional information may be generated by retrieving data (e.g., certain text relevant to the log, the code, and/or the code error) stored in the ontology associated with the service that utilizes the code. The retrieved data may be included in the context. Additionally, the system may generate detailed and/or specific instructions”);
wherein temporal context data comprises the respective temporal context data for each alert message of the group of alert messages, wherein the temporal context data of the group of alert messages are indicative of respective states of an associated system at times at which alert messages of the group of alert messages were generated; ([0038]; regarding, “The log may include information related to a code error (also referred to as “an error”), such as an error message that informs the error, a type of the error (e.g., compile time error, run time error, a syntax error, an overflow error, or the like), the name and or file-path of the code associated with the error, timestamps associated with the error, or other information related to the error.”);
generating a prompt chain comprising a sequence of prompts, wherein each prompt of the sequence of prompts is associated with a respective alert message from the group of alert messages, and wherein each prompt is generated based on: text data based on the alert message associated with the prompt, ([0070]; regarding, “The prompt generation module 110 is configured to generate one or more prompts to one or more language models, such as LLM 130a and/or LLM 130b. The prompt generation module 110 may generate a prompt that includes an error message indicating a code error and context associated with the error for the LLM 130a and/or LLM 130b to explain the error message.”);
and submitting the sequence of prompts to the second artificial intelligence model. ([0071]; regarding, “the error analysis system 102 may be capable of interfacing with multiple LLMs. This allows for experimentation, hot-swapping and/or adaptation to different models based on specific use cases or requirements, providing versatility and scalability to the system. In various implementations, the error analysis system 102 may interface with a second LLM 130b in order to, for example, generate some of context associated with an error for the first LLM 130a to explain the error.”).
Agrawal fails to explicitly disclose but Joshi teaches:
and performing, using a first artificial intelligence… a temporal embedding that encodes features of the temporal context data in the temporal modality into respective first embedding vectors according to an embedding-space of a second artificial intelligence model that is configured to receive input according to the text modality, wherein the second artificial intelligence model is different from the first artificial intelligence model; ([0036]; regarding, “The embeddings unit 302 is configured to access the prompt information stored in the privacy preserving datastore 204 and to provide each prompt as an input to the embeddings language model 304 to obtain corresponding embeddings vectors. The embeddings language model 304 is a natural language model (NLP) that is configured to analyze a textual input and to output embeddings vectors”; [0037]; regarding, “The data clustering unit 308 analyzes the embedding vectors using one or more privacy preserving clustering algorithms 310. The privacy preserving clustering algorithms 310 group embedding vectors representing the user prompts into groups having similar characteristics. In some implementations a k-means clustering algorithm that partitions the embedding vectors into k clusters in which each embedding vector belongs to a cluster having a nearest mean to the embedding vector.”);
…each prompt is generated based on:… a respective first embedding vector of the respective first embedding vectors generated from the temporal context data associated with the alert message associated with the prompt; ([0032]; regarding, “The prompt anonymization pipeline 124 analyzes and anonymizes the prompt data stored in the privacy preserving datastore 204 using the prompt data analysis unit 206. The prompt data analysis unit 206 accesses the prompt data stored in the privacy preserving datastore 204, determines embedding vectors for the prompt data, analyzes the embeddings to identify clusters, and summarizes the clusters to generate anonymized prompt data to be stored in the anonymized prompt data datastore 208.”).
Therefore, it would have been obvious before the effective filing date of the claimed invention to one of ordinary skill in the art to which said subject matter pertains to combine Agrawal with the teachings of Joshi. Doing so can improve predictions provided by the model. (Joshi, [0054]).
Agrawal in view of Joshi fails to explicitly disclose but Crabtree teaches:
…retrieving, from a telemetry store that stores telemetry data and metric data as time series data… ([0203]; regarding, “Data persistence stores such as the multiple dimension time series data store module 1020 may receive streaming data from a large plurality of sensors that may be of several different types. The multiple dimension time series data store module may also store any time series data”);
…using a first artificial intelligence model implemented as a time series soft prompt encoder and configured to operate in the temporal modality… ([0281]; regarding, “A first such variation comprises Auto-Encoding Models”; [0282]; regarding, “The primary goal of an autoencoder is to learn efficient representations of input data by encoding the data into a lower-dimensional space and then reconstructing the original data from the encoded representation.”; [0203]; regarding, “distributed computational graph computing system utilizing an AI enhanced decision platform for external network reconnaissance and contextual data collection… Much of the enterprise knowledge/context data analyzed by the system both from sources within the confines of the enterprise business, and from cloud based sources, also enter the system through the cloud interface 1010, data being passed to the connector module 1035 which may possess the API routines 1035a needed to accept and convert the external data and then pass the normalized information to other analysis and transformation components of the system, the… multidimensional time series database (MDTSDB) 1020”);
wherein generating the prompt chain comprises temporal prompt chaining in which prompts later in the sequence are constructed based on information derived from earlier prompts in the sequence and correlations among the timestamp data of the group of alert messages; ([0186]; regarding, “During a prompt execution process, experience curation 340 can send user query to DCG 330 which can orchestrate the retrieval of context and a response. Using its declarative roots, DCG 330 can abstract away many of the details of prompt chaining;… retrieving contextual data from vector databases 330; and maintaining memory across multiple LLM calls. The DCG output may be a prompt, or series of prompts, to submit to a language model via LLM services 360 (which may be potentially prompt tuned). In turn, the LLM processes the prompts, contextual data, and user query to generate a contextually aware response”).
Therefore, it would have been obvious before the effective filing date of the claimed invention to one of ordinary skill in the art to which said subject matter pertains to combine Agrawal and Joshi with the teachings of Crabtree. Doing so allows for better resource utilization and improved performance. (Crabtree, [0218]).
Regarding Claim 2, Agrawal in view of Joshi in further view of Crabtree teaches the device of claim 1 as cited above. Agrawal in view of Joshi in further view of Crabtree further teaches:
wherein the group of alert messages are received from a health monitoring device of the associated system and received in response to a number of alert message in the group being greater than or equal to a defined threshold. (Agrawal, [0036]; regarding, “the system may receive or access a log (e.g., an error log) that includes one or more error messages.”; [0038]; regarding, “The log may be generated by a software when implementing an application or a service”).
Regarding Claim 3, Agrawal in view of Joshi in further view of Crabtree teaches the device of claim 1 as cited above. Agrawal in view of Joshi in further view of Crabtree further teaches:
wherein the temporal context data comprises time series data. (Agrawal, [0043]).
Regarding Claim 4, Agrawal in view of Joshi in further view of Crabtree teaches the device of claim 1 as cited above. Agrawal in view of Joshi in further view of Crabtree further teaches:
wherein the second artificial intelligence model comprises a large language model. (Joshi, [0039]; regarding, “In one implementation, the summarization model 314 is an LLM that is fine-tuned using DP on each cluster of embedding vectors.”).
Regarding Claim 5, Agrawal in view of Joshi in further view of Crabtree teaches the device of claim 1 as cited above. Agrawal in view of Joshi in further view of Crabtree further teaches:
wherein the first artificial intelligence model comprises a deep learning model. (Joshi, [0036]; regarding, “The embeddings language model 304 is a natural language model (NLP)”).
Regarding Claim 6, Agrawal in view of Joshi in further view of Crabtree teaches the device of claim 1 as cited above. Agrawal in view of Joshi in further view of Crabtree further teaches:
wherein the second artificial intelligence model encodes the text data of the prompts into respective second embedding vectors associated with the alert messages. (Joshi, [0038]; regarding, “The cluster summarization unit 312 utilizes the summarization model 314 to summarize the embeddings”; [0039]; regarding, “The LLM is prompted by the cluster summarization unit 312 to generate a plurality of synthetic representative data points that represent embedding vectors associated with each cluster.”).
Regarding Claim 7, Agrawal in view of Joshi in further view of Crabtree teaches the device of claim 6 as cited above. Agrawal in view of Joshi in further view of Crabtree further teaches:
wherein the second artificial intelligence model, for each alert message concatenates the first embedding vector and the second embedding vector associated with the alert message as part of an embedding layer of the second artificial intelligence model. (Joshi, [0038]; regarding, “The cluster summarization unit 312 utilizes the summarization model 314 to summarize the embeddings associated with at least a subset of the clusters to determine a “theme” representing the subject matter of the cluster.”; [0036]; regarding, “The embeddings unit 302 is configured to access the prompt information stored in the privacy preserving datastore 204 and to provide each prompt as an input to the embeddings language model 304 to obtain corresponding embeddings vectors.”).
Regarding Claim 8, Agrawal in view of Joshi in further view of Crabtree teaches the device of claim 1 as cited above. Agrawal in view of Joshi in further view of Crabtree further teaches:
wherein the operations further comprise, for each alert message of the group of alert messages, in response to examination of a knowledge store, determining, using a third artificial intelligence model, respective alert context data that indicates additional context information regarding the alert message in the text modality, and wherein the third artificial intelligence model is different from the first artificial intelligence model and the second artificial intelligence model. (Agrawal, [0080]; regarding, “the context generation module 106 may access portions of a log that are close or more related to the error message to determine the context associated with the error.”; [0083]; regarding, “the context generation module 106 may execute the similarity search using at least one of a language model, an artificial intelligence (“AI”) model, a generative model, a machine learning (“ML”) model, a neural network (“NN”), or an LLM.”).
Regarding Claim 9, Agrawal in view of Joshi in further view of Crabtree teaches the device of claim 8 as cited above. Agrawal in view of Joshi in further view of Crabtree further teaches:
wherein the text data associated with the prompt further comprises the respective alert context data associated with the alert message associated with the prompt. (Agrawal, [0037]; regarding, “The prompt may include the error message, the context associated with the error…”).
Regarding Claim 10, Agrawal in view of Joshi in further view of Crabtree teaches the device of claim 9 as cited above Agrawal in view of Joshi in further view of Crabtree further teaches:
wherein the third artificial intelligence model comprises a retrieval augmented generation model. (Agrawal, [0035]; regarding, “various implementations of the systems and methods of the present disclosure can advantageously employ one or more LLMs”; [0036]; regarding, “The context may include… portions of the log that are close or more related to the error message… and/or additional information… relevant to the error or useful for one or more LLMs to explain the error message…. the additional information may be generated by retrieving data”; [0037]; regarding, “Based on the error message, the context associated with the error, and/or the instructions, the system may generate a prompt for an LLM.”; [0082]; regarding, “he context generation module 106 may further vectorize the plurality of portions of the set of documents to generate a plurality of vectors.”).
Regarding Claim 12, Agrawal teaches:
A method, comprising:
receiving, by a device comprising at least one processor, a group of alert messages within a predefined time window, wherein each alert message of the group of alert messages comprises: ([0036]; regarding, “the system may receive or access a log (e.g., an error log) that includes one or more error messages.”; [0038]; regarding, “The log may be generated by a software when implementing an application or a service (e.g., running a data processing pipeline, compiling a software package, or the like).”)
a respective description of a respective alert associated with the alert message in a text modality, and respective timestamp data indicative of a respective time at which the alert message was generated; ([0038]; regarding, “The log may include information related to a code error (also referred to as “an error”), such as an error message that informs the error, a type of the error (e.g., compile time error, run time error, a syntax error, an overflow error, or the like), the name and or file-path of the code associated with the error, timestamps associated with the error, or other information related to the error.”);
for each alert message of the group of alert messages, based on the respective timestamp data, retrieving, by the device, respective temporal context data from a telemetry store… ([0036]; regarding, “the system may receive or access a log (e.g., an error log) that includes one or more error messages… Based on the log and/or the error message, the system may further search and determine a context associated with the error. The context may include the code, portions of the log that are close or more related to the error message, portions of one or more documents associated with the code, and/or additional information (e.g., ontology associated with a service that utilizes the code, search results from search engines) relevant to the error or useful for one or more LLMs to explain the error message. For example, the additional information may be generated by retrieving data (e.g., certain text relevant to the log, the code, and/or the code error) stored in the ontology associated with the service that utilizes the code. The retrieved data may be included in the context. Additionally, the system may generate detailed and/or specific instructions”);
wherein temporal context data comprises the respective temporal context data for each alert message of the group of alert messages, wherein the temporal context data of the group of alert messages are indicative of respective states of an associated system at times at which alert messages of the group of alert messages were generated; ([0038]; regarding, “The log may include information related to a code error (also referred to as “an error”), such as an error message that informs the error, a type of the error (e.g., compile time error, run time error, a syntax error, an overflow error, or the like), the name and or file-path of the code associated with the error, timestamps associated with the error, or other information related to the error.”);
generating a prompt chain comprising a sequence of prompts, wherein each prompt of the sequence of prompts is associated with a respective alert message from the group of alert messages, and wherein each prompt is generated based on: text data based on the alert message associated with the prompt, ([0070]; regarding, “The prompt generation module 110 is configured to generate one or more prompts to one or more language models, such as LLM 130a and/or LLM 130b. The prompt generation module 110 may generate a prompt that includes an error message indicating a code error and context associated with the error for the LLM 130a and/or LLM 130b to explain the error message.”);
inputting, by the device, the sequence of prompts to the second artificial intelligence model. ([0071]; regarding, “the error analysis system 102 may be capable of interfacing with multiple LLMs. This allows for experimentation, hot-swapping and/or adaptation to different models based on specific use cases or requirements, providing versatility and scalability to the system. In various implementations, the error analysis system 102 may interface with a second LLM 130b in order to, for example, generate some of context associated with an error for the first LLM 130a to explain the error.”).
Agrawal fails to explicitly disclose but Joshi teaches:
encoding, by the device, using a first artificial intelligence model… features of the temporal context data in the temporal-based modality into an embedding space of a second artificial intelligence model that is configured to receive input according to the text-based modality, wherein the second artificial intelligence model is different from the first artificial intelligence model; ([0036]; regarding, “The embeddings unit 302 is configured to access the prompt information stored in the privacy preserving datastore 204 and to provide each prompt as an input to the embeddings language model 304 to obtain corresponding embeddings vectors. The embeddings language model 304 is a natural language model (NLP) that is configured to analyze a textual input and to output embeddings vectors”; [0037]; regarding, “The data clustering unit 308 analyzes the embedding vectors using one or more privacy preserving clustering algorithms 310. The privacy preserving clustering algorithms 310 group embedding vectors representing the user prompts into groups having similar characteristics. In some implementations a k-means clustering algorithm that partitions the embedding vectors into k clusters in which each embedding vector belongs to a cluster having a nearest mean to the embedding vector.”);
…each prompt is generated based on:… a respective first embedding vector of the respective first embedding vectors generated from the respective temporal context data associated with the alert message associated with the prompt; ([0032]; regarding, “The prompt anonymization pipeline 124 analyzes and anonymizes the prompt data stored in the privacy preserving datastore 204 using the prompt data analysis unit 206. The prompt data analysis unit 206 accesses the prompt data stored in the privacy preserving datastore 204, determines embedding vectors for the prompt data, analyzes the embeddings to identify clusters, and summarizes the clusters to generate anonymized prompt data to be stored in the anonymized prompt data datastore 208.”).
Therefore, it would have been obvious before the effective filing date of the claimed invention to one of ordinary skill in the art to which said subject matter pertains to combine Agrawal with the teachings of Joshi. Doing so can improve predictions provided by the model. (Joshi, [0054]).
Agrawal in view of Joshi fails to explicitly disclose but Crabtree teaches:
…retrieving, from a telemetry store that stores telemetry data and metric data as time series data… ([0203]; regarding, “Data persistence stores such as the multiple dimension time series data store module 1020 may receive streaming data from a large plurality of sensors that may be of several different types. The multiple dimension time series data store module may also store any time series data”);
…using a first artificial intelligence model implemented as a time series soft prompt encoder and configured to operate in the temporal modality… ([0281]; regarding, “A first such variation comprises Auto-Encoding Models”; [0282]; regarding, “The primary goal of an autoencoder is to learn efficient representations of input data by encoding the data into a lower-dimensional space and then reconstructing the original data from the encoded representation.”; [0203]; regarding, “distributed computational graph computing system utilizing an AI enhanced decision platform for external network reconnaissance and contextual data collection… Much of the enterprise knowledge/context data analyzed by the system both from sources within the confines of the enterprise business, and from cloud based sources, also enter the system through the cloud interface 1010, data being passed to the connector module 1035 which may possess the API routines 1035a needed to accept and convert the external data and then pass the normalized information to other analysis and transformation components of the system, the… multidimensional time series database (MDTSDB) 1020”);
wherein generating the prompt chain comprises temporal prompt chaining in which prompts later in the sequence are constructed based on information derived from earlier prompts in the sequence and correlations among the timestamp data of the group of alert messages; ([0186]; regarding, “During a prompt execution process, experience curation 340 can send user query to DCG 330 which can orchestrate the retrieval of context and a response. Using its declarative roots, DCG 330 can abstract away many of the details of prompt chaining;… retrieving contextual data from vector databases 330; and maintaining memory across multiple LLM calls. The DCG output may be a prompt, or series of prompts, to submit to a language model via LLM services 360 (which may be potentially prompt tuned). In turn, the LLM processes the prompts, contextual data, and user query to generate a contextually aware response”).
Therefore, it would have been obvious before the effective filing date of the claimed invention to one of ordinary skill in the art to which said subject matter pertains to combine Agrawal and Joshi with the teachings of Crabtree. Doing so allows for better resource utilization and improved performance. (Crabtree, [0218]).
Claims 13-14 are rejected under 35 U.S.C. 103 under the same grounds of rejection as claims 8-9 respectively.
Regarding Claim 16, Agrawal teaches:
A non-transitory computer-readable medium comprising instructions that, in response to execution, cause a system comprising a processor to perform operations, comprising:
receiving a group of alert messages within a predefined time window, wherein each alert message of the group of alert messages comprises: ([0036]; regarding, “the system may receive or access a log (e.g., an error log) that includes one or more error messages.”; [0038]; regarding, “The log may be generated by a software when implementing an application or a service (e.g., running a data processing pipeline, compiling a software package, or the like).”)
a respective description of a respective alert associated with the alert message in a text modality, and respective timestamp data indicative of a respective time at which the alert message was generated; ([0038]; regarding, “The log may include information related to a code error (also referred to as “an error”), such as an error message that informs the error, a type of the error (e.g., compile time error, run time error, a syntax error, an overflow error, or the like), the name and or file-path of the code associated with the error, timestamps associated with the error, or other information related to the error.”);
for each alert message of the group of alert messages, based on the respective timestamp data, retrieving, from a temporal context store, respective temporal context data that are stored according to a temporal modality, wherein the respective temporal context data are indicative of a respective state of an associated system at the respective time; ([0036]; regarding, “the system may receive or access a log (e.g., an error log) that includes one or more error messages… Based on the log and/or the error message, the system may further search and determine a context associated with the error. The context may include the code, portions of the log that are close or more related to the error message, portions of one or more documents associated with the code, and/or additional information (e.g., ontology associated with a service that utilizes the code, search results from search engines) relevant to the error or useful for one or more LLMs to explain the error message. For example, the additional information may be generated by retrieving data (e.g., certain text relevant to the log, the code, and/or the code error) stored in the ontology associated with the service that utilizes the code. The retrieved data may be included in the context. Additionally, the system may generate detailed and/or specific instructions”; [0038]; regarding, “The log may include information related to a code error (also referred to as “an error”), such as an error message that informs the error, a type of the error (e.g., compile time error, run time error, a syntax error, an overflow error, or the like), the name and or file-path of the code associated with the error, timestamps associated with the error, or other information related to the error.”);
generating a prompt chain comprising a sequence of prompts, wherein each prompt of the sequence of prompts is associated with a respective alert message from the group of alert messages, and wherein each prompt is generated based on: text data based on the alert message associated with the prompt, ([0070]; regarding, “The prompt generation module 110 is configured to generate one or more prompts to one or more language models, such as LLM 130a and/or LLM 130b. The prompt generation module 110 may generate a prompt that includes an error message indicating a code error and context associated with the error for the LLM 130a and/or LLM 130b to explain the error message.”);
submitting the sequence of prompts to the second artificial intelligence model. ([0071]; regarding, “the error analysis system 102 may be capable of interfacing with multiple LLMs. This allows for experimentation, hot-swapping and/or adaptation to different models based on specific use cases or requirements, providing versatility and scalability to the system. In various implementations, the error analysis system 102 may interface with a second LLM 130b in order to, for example, generate some of context associated with an error for the first LLM 130a to explain the error.”).
Agrawal fails to explicitly disclose but Joshi teaches:
encoding, by the device, using a first artificial intelligence model, features of the temporal context data in the temporal-based modality into an embedding space of a second artificial intelligence model that is configured to receive input according to the text-based modality, wherein the second artificial intelligence model is different from the first artificial intelligence model; ([0036]; regarding, “The embeddings unit 302 is configured to access the prompt information stored in the privacy preserving datastore 204 and to provide each prompt as an input to the embeddings language model 304 to obtain corresponding embeddings vectors. The embeddings language model 304 is a natural language model (NLP) that is configured to analyze a textual input and to output embeddings vectors”; [0037]; regarding, “The data clustering unit 308 analyzes the embedding vectors using one or more privacy preserving clustering algorithms 310. The privacy preserving clustering algorithms 310 group embedding vectors representing the user prompts into groups having similar characteristics. In some implementations a k-means clustering algorithm that partitions the embedding vectors into k clusters in which each embedding vector belongs to a cluster having a nearest mean to the embedding vector.”);
…each prompt is generated based on:… a respective first embedding vector of the respective first embedding vectors generated from the respective temporal context data associated with the alert message associated with the prompt; ([0032]; regarding, “The prompt anonymization pipeline 124 analyzes and anonymizes the prompt data stored in the privacy preserving datastore 204 using the prompt data analysis unit 206. The prompt data analysis unit 206 accesses the prompt data stored in the privacy preserving datastore 204, determines embedding vectors for the prompt data, analyzes the embeddings to identify clusters, and summarizes the clusters to generate anonymized prompt data to be stored in the anonymized prompt data datastore 208.”).
Therefore, it would have been obvious before the effective filing date of the claimed invention to one of ordinary skill in the art to which said subject matter pertains to combine Agrawal with the teachings of Joshi. Doing so can improve predictions provided by the model. (Joshi, [0054]).
Claims 17-18 are rejected under 35 U.S.C. 103 under the same grounds of rejection as claims 8-9 respectively.
Regarding Claim 20, Agrawal in view of Joshi in further view of Crabtree teaches the medium of claim 17 as cited above. Agrawal in view of Joshi in further view of Crabtree further teaches:
wherein the first artificial intelligence model comprises a deep learning model, (Joshi, [0036]; regarding, “The embeddings language model 304 is a natural language model (NLP)”).
the second artificial intelligence model comprises a large language model, (Joshi, [0039]; regarding, “In one implementation, the summarization model 314 is an LLM that is fine-tuned using DP on each cluster of embedding vectors.”).
and the third artificial intelligence model comprises a retrieval augmented generation model. (Agrawal, [0035]; regarding, “various implementations of the systems and methods of the present disclosure can advantageously employ one or more LLMs”; [0036]; regarding, “The context may include… portions of the log that are close or more related to the error message… and/or additional information… relevant to the error or useful for one or more LLMs to explain the error message…. the additional information may be generated by retrieving data”; [0037]; regarding, “Based on the error message, the context associated with the error, and/or the instructions, the system may generate a prompt for an LLM.”; [0082]; regarding, “he context generation module 106 may further vectorize the plurality of portions of the set of documents to generate a plurality of vectors.”).
Regarding Claim 21, Agrawal in view of Joshi in further view of Crabtree teaches the medium of claim 16 as cited above. Agrawal in view of Joshi in further view of Crabtree further teaches:
wherein the second artificial intelligence model encodes the text data of the prompts into respective second embedding vectors associated with the alert messages. (Fig. 3, Joshi, [0038]; regarding, “The cluster summarization unit 312 utilizes the summarization model 314 to summarize the embeddings”; [0039]; regarding, “The LLM is prompted by the cluster summarization unit 312 to generate a plurality of synthetic representative data points that represent embedding vectors associated with each cluster.”).
Regarding Claim 22, Agrawal in view of Joshi in further view of Crabtree teaches the medium of claim 21 as cited above. Agrawal in view of Joshi in further view of Crabtree further teaches:
wherein the second artificial intelligence model, for each alert message concatenates the first embedding vector and the second embedding vector associated with the alert message as part of an embedding layer of the second artificial intelligence model. (Joshi, [0038]; regarding, “The cluster summarization unit 312 utilizes the summarization model 314 to summarize the embeddings associated with at least a subset of the clusters to determine a “theme” representing the subject matter of the cluster.”; [0036]; regarding, “The embeddings unit 302 is configured to access the prompt information stored in the privacy preserving datastore 204 and to provide each prompt as an input to the embeddings language model 304 to obtain corresponding embeddings vectors.”).
Regarding Claim 23, Agrawal in view of Joshi in further view of Crabtree teaches the method of claim 12 as cited above. Agrawal in view of Joshi in further view of Crabtree further teaches:
wherein the second artificial intelligence model: encodes the text data of the prompts into respective second embedding vectors associated with the alert messages, and for each alert message concatenates the first embedding vector and the second embedding vector associated with the alert message as part of an embedding layer of the second artificial intelligence model. (Fig. 3, Joshi, [0038]; regarding, “The cluster summarization unit 312 utilizes the summarization model 314 to summarize the embeddings”; [0039]; regarding, “The LLM is prompted by the cluster summarization unit 312 to generate a plurality of synthetic representative data points that represent embedding vectors associated with each cluster.”; [0036]; regarding, “The embeddings unit 302 is configured to access the prompt information stored in the privacy preserving datastore 204 and to provide each prompt as an input to the embeddings language model 304 to obtain corresponding embeddings vectors.”).
Response to Arguments
Applicant’s arguments filed 06/23/2026 have been fully considered.
With respect to the 103 rejections of independent Claim 1 and similarly Claims 12 and 16, Applicant’s arguments have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument. Please see the above detailed rejection.
Newly cited reference Crabtree in combination with Agrawal and Joshi, teaches the newly claimed matter.
Conclusion
THIS ACTION IS MADE FINAL. 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 MATHEW GUSTAFSON whose telephone number is (571)272-5273. The examiner can normally be reached Monday-Friday 8:00-4:00.
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, Bryce Bonzo can be reached at (571) 272-3655. 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.
/MICHAEL MASKULINSKI/Primary Examiner, Art Unit 2113
/M.D.G./Examiner, Art Unit 2113
Prosecution Insights