Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
DETAILED ACTION
The Action is responsive to the Request for Continued Examination, Amendments, and Remarks filed on 12/8/2025. Claims 1, 3-11, 13-17, 19-20, and 22-23 are pending claims. Claims 1, 11, and 17 are written in independent form. Claims 2, 12, 18, and 21 have been cancelled by Applicant.
Priority
Applicant’s claim for benefit of prior-filed provisional application 63/568,851 dated 3/22/2024 under 35. U.S.C. 119(e) or under 35 U.S.C. 120, 121, or 365(c) is acknowledged.
Claim Interpretation
Claims 4-7, 10, 14, and 20 recite the phrase “the LLM is fine-tuned for…” which is being understood as the intent to use the LLM, but is not actively performing any step/limitation of using the LLM. Examiner suggests to amend the claim limitations to recite all of the steps in a positive manner.
Claims 4, 14, and 20 recite the phrase “the LLM is fine-tuned for a cloud security application” which is being interpreted to have a scope of “the LLM is fine-tuned”. However, for the purpose of compact prosecution, the limitation is being addressed herein as if all of the steps are recited in a positive manner.
Claim 5 recites the phrase “the LLM is fine-tuned for performing automated entity extraction for multi-domain security applications” which is being interpretated to have a scope of “the LLM is fine-tuned”. However, for the purpose of compact prosecution, the limitation is being addressed herein as if all of the steps are recited in a positive manner.
Claims 6 and 10 recite the phrase “the LLM is fine-tuned for performing a cross-domain search…” which is being interpreted to have a scope of “the LLM is fine-tuned”. However, for the purpose of compact prosecution, the limitation is being addressed herein as if all of the steps are recited in a positive manner.
Claim 7 recites the phrase “the LLM is fine-tuned for performing a natural language (NL) query…” which is being interpreted to have a scope of “the LLM is fine-tuned”. However, for the purpose of compact prosecution, the limitation is being addressed herein as if all of the steps are recited in a positive manner.
Claims 6, 10 recites the phrase “to generate” which is being understand as the intent to generate, but is not actively performing any generating step/limitation. Examiner suggests to amend the claim limitation to recite all of the steps in a positive manner.
Claims 6 and 10 recite the phrase “…a cross-domain search to generate a search result…” which is being interpreted as not further narrowing the scope because the cross-domain search is not being actively performed. However, for the purpose of compact prosecution, the limitation is being addressed herein as if all of the steps are recited in a positive manner.
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, 3-11, 13-17, 19-20, and 22-23 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-patentable subject matter. The claimed invention is directed to one or more abstract ideas without significantly more. The judicial exception is not integrated into a practical application. The claims do not include additional elements that are sufficient to amount to significantly more than judicial exception. The eligibility analysis in support of these findings is provided below.
As per Claims 1, 11, and 17,
STEP 1:In accordance with Step 1 of the eligibility inquiry (as explained in MPEP 2106), the claimed system (claims 1, 3-10, and 22-23), method (claims 11 and 13-16), and computer program product (claims 17 and 19-20) are directed to one of the eligible categories of subject matter and therefore satisfies Step 1.
STEP 2A Prong One:The independent claims 1, 11, and 17 recite the following limitations directed to an abstract idea:
Automatically generate a seed dataset for a domain specific language (DSL);
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by making a judgement and/or opinion to generate (or think of) a seed dataset based on the observation and evaluation.
Wherein the automatically generating of the seed dataset comprises to:
Generate a natural language representation of the RQL to obtain the seed dataset;
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating a resource query language, and based on the observation and evaluation, making a judgement and/or opinion of a natural language representation of the RQL.
Expand, using a Large language Model (LLM), the seed dataset for the DSL; and
The limitation recites a mathematical concept of executing a mathematical function in the form of an LLM that takes as input the seed dataset and outputs an expanded seed data set.
Comprising to perform one or more of the following:
A) for a first set of samples using single attribute values having same asset attributes, generate, using the LLM, queries having a plurality of attribute values for the same asset attribute to obtain the expanded dataset for the DSL; or
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating a first set of samples using single attribute value having the same asset attribute and an LLM, and based on the observation and evaluation, making a judgement and/or opinion of queries having a plurality of attribute values for the same asset attribute to obtain the expanded dataset for the DSL.
B) for a second set of samples using single attribute values having the same asset attribute and have different filtering criteria, generate, using the LLM, queries that mix the natural language of the RQL to obtain the expanded dataset for the DSL,
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating an LLM, the natural language representation of the RQL, a second set of samples using single asset values having same asset attributes and have different filtering criteria, and based on the observation and evaluation, making a judgement and/or opinion of queries that mix the natural langue of the RQL to obtain the expanded dataset for the DSL.
Validate the expanded dataset for the DSL,
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by observing and evaluating the expanded dataset for the DSL and making a judgement and/or opinion, based on the observation and evaluation, of the validity of the expanded dataset.
wherein the expanded dataset for the DSL is input to the LLM for fine tune training of the LLM;
The limitation recites a mathematical concept of executing a mathematical function in the form of fine tune training an LLM that takes as input the expanded dataset.
STEP 2A Prong Two:The claims recite using “a processor”, “a memory”, and “a non-transitory computer readable medium”, which is a high-level recitation of generic computer components and represents mere instructions to apply on a computer as in MPEP 2106.05(f), which does not provide integration into a practical application.
The claims recite the following additional elements:
Wherein the DSL includes a resource query language (RQL);
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent the domain specific language as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
Wherein the same asset attribute includes one or more of the following: asset name, asset type, asset class, cloud type, cloud account, finding name, finding type, and/or finding severity;
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent the asset attribute as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
Wherein the filtering criteria includes one or more of the following: a vulnerability having a Common Vulnerability Scoring System (CVSS) score within an allowed range, a vulnerability having an allowed severity, and/or a vulnerability having an allowed vulnerability identifier;
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent the filtering criteria as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
output search results for a query based on the validated expanded dataset;
The limitation recites an insignificant extra solution activity as retrieval of data (ie. Mere data gathering) such as ‘obtaining information’ as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
STEP 2B:
The conclusions for the mere implementation using a computer are carried over and does not provide significantly more.
With respect to “Wherein the DSL includes a resource query language (RQL).” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to “Wherein the same asset attribute includes one or more of the following: asset name, asset type, asset class, cloud type, cloud account, finding name, finding type, and/or finding severity;” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to “Wherein the filtering criteria includes one or more of the following: a vulnerability having a Common Vulnerability Scoring System (CVSS) score within an allowed range, a vulnerability having an allowed severity, and/or a vulnerability having an allowed vulnerability identifier;” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to “output search results for a query based on the validated expanded dataset;” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(i).
Looking at the claim as a whole does not change this conclusion and the claim is ineligible.
As per Dependent Claims 3-10, 13-16, 19-20, and 22-23,
STEP 1:In accordance with Step 1 of the eligibility inquiry (as explained in MPEP 2106), the claimed system (claims 1, 3-10, and 22-23), method (claims 11 and 13-16), and computer program product (claims 17 and 19-20) are directed to one of the eligible categories of subject matter and therefore satisfies Step 1.
STEP 2A Prong One:The dependent claims 3-10, 13-16, 19-20, and 22-23 recite the following limitations directed to an abstract idea:
The limitation of Dependent Claims 4, 14, and 20 include the step(s) of:
Wherein the LLM is fine-tuned.
The limitation recites a mathematical concept of executing a mathematical function in the form of fine tune training an LLM that takes as input the seed dataset.
The limitation of Dependent Claim 5 includes the step(s) of:
Wherein the LLM is fine-tuned for performing automated entity extraction for multi-domain security applications.
The limitation recites a mathematical concept of executing a mathematical function in the form of fine tune training an LLM that takes as input the seed dataset.
Analyzing the limitation as if the intended use language was recited in a positive step, the limitation recites a mathematical concept of executing a mathematical function in the form of the LLM that performs automated entity extraction.
The limitation of Dependent Claim 6 includes the step(s) of:
Wherein the LLM is fine-tuned for performing a cross-domain search to generate a search result using a plurality of data source domains that includes using a planner, executor, and aggregator to collect distinct results from each of the plurality of data source domains.
The limitation recites a mathematical concept of executing a mathematical function in the form of fine tune training an LLM that takes as input the seed dataset.
Analyzing the limitation as if the intended use language was recited in a positive step, the limitation recites a mathematical concept of executing a mathematical function in the form of the LLM that performs a cross-domain search and outputs search results.
The limitation of Dependent Claim 7 includes the step(s) of:
Wherein the LLM is fine-tuned for performing a natural language (NL) query for a plurality of data source domains for a cloud security application,
The limitation recites a mathematical concept of executing a mathematical function in the form of fine tune training an LLM that takes as input the seed dataset.
Analyzing the limitation as if the intended use language was recited in a positive step, the limitation recites a mathematical concept of executing a mathematical function in the form of the LLM that performs a natural language query for a plurality of data source domains.
The limitation of Dependent Claims 8 and 15 include the step(s) of:
Generate an RQL query in response to a natural language query using the fine-tuned LLM.
The limitation recites a mathematical concept of executing a mathematical function in the form of an LLM that takes as input a natural language query and outputs an RQL query.
The limitation of Dependent Claims 9 and 16 include the step(s) of:
automatically generate a configuration policy in RQL from a natural language (NL) input.
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating a natural language input, and making a judgement and/or opinion of a configuration policy in RQL based on the observation and evaluation.
The limitation of Dependent Claim 10 includes the step(s) of:
Wherein the LLM is fine-tuned for performing a cross-domain search to generate a search result using a plurality of data source domains to collect distinct results from each of the plurality of data source domains, and
Analyzing the limitation as if the intended use language was recited in a positive step, the limitation recites a mathematical concept of executing a mathematical function in the form of the LLM that performs a cross-domain search and outputs search results.
automatically generate an output in response to a natural language query, wherein one or more of the plurality of data source domains are searched using queries in RQL.
The limitation recites a mathematical concept of executing a mathematical function in the form of the LLM that takes as input a natural language query and outputs a response based on performing a search of one or more data source domains using queries in RQL (based on the input natural language query).
The limitation of Dependent Claim 21 includes the step(s) of:
wherein the expanding of the seed dataset comprises to generate a plurality of variations of the natural language representation to obtain the expanded dataset for the DSL;
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating the natural language representation and based on the observation and evaluation, making a judgement and/or opinion of variations of the natural language representation to obtain the expanded dataset for the DSL.
The limitation of Dependent Claim 22 includes the step(s) of:
wherein the expanding of the seed dataset comprises to, for a first set of samples using single attribute values having same asset attribute, generate, using the LLM, queries having a plurality of attribute values for the same asset attribute to obtain the expanded dataset for the DSL,
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating a first set of samples using single asset values having same asset attributes, and based on the observation and evaluation, making a judgement and/or opinion of queries having a plurality of asset values for the same asset attributes to obtain the expanded dataset for the DSL.
The limitation of Dependent Claim 23 includes the step(s) of:
wherein the expanding of the seed dataset comprises to, for a second set of samples using single attribute values having the same asset attribute and have different filtering criteria, generate, using the LLM, queries that mix the natural language of the RQL to obtain the expanded dataset for the DSL,
The limitation recites a mental process of observation, evaluation, judgement, and/or opinion capable of being performed by the human mind by observing and evaluating an LLM, the natural language representation of the RQL, a second set of samples using single asset values having same asset attributes and have different filtering criteria, and based on the observation and evaluation, making a judgement and/or opinion of queries that mix the natural langue of the RQL to obtain the expanded dataset for the DSL.
STEP 2A Prong Two:The claim(s) recite the following additional elements:
The limitation of Dependent Claims 3, 13, and 19 include the step(s) of:
Wherein the RQL is generated for RQL for multi-domain security applications.
The limitation recites an insignificant extra-solution activity as selecting a particular type of resource data being used to represent the resource query language as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
The limitation of Dependent Claims 4, 14, and 20 include the step(s) of:
Wherein the LLM is fine-tuned for a cloud security application.
The limitation recites an insignificant extra-solution activity as selecting a particular type of data/application being performed by the LLM as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
The limitation of Dependent Claim 7 includes the step(s) of:
Wherein the DSL is a resource query language (RQL),
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent the domain specific language as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
Wherein the LLM is fine-tuned for performing a natural language (NL) query for a plurality of data source domains for a cloud security application, and
The limitation recites an insignificant extra solution activity as retrieval of data (ie. Mere data gathering) such as ‘obtaining information’ through executing a query/search as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
Wherein the plurality of data source domains includes a configuration data set, an Identity and Asset Management (IAM) data set, and a vulnerability data set.
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being represented in the data source domains as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
The limitation of Dependent Claims 9 and 16 include the step(s) of:
Wherein the DSL is a resource query language (RQL),
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent the domain specific language as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
The limitation of Dependent Claim 22 includes the step(s) of:
wherein the same asset attribute includes one or more of the following: asset name, asset type, asset class, cloud type, cloud account, finding name, finding type, and/or finding severity.
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent the asset attribute as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
The limitation of Dependent Claim 23 includes the step(s) of:
Wherein the filtering criteria includes one or more of the following: a vulnerability having a Common Vulnerability Scoring System (CVSS) score within an allowed range, a vulnerability having an allowed severity, and/or a vulnerability having an allowed vulnerability identifier;
The limitation recites an insignificant extra-solution activity as selecting a particular type of data being used to represent the filtering criteria as identified in MPEP 2106.05(g) and does not provide integration into a practical application.
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
STEP 2B:
The conclusions for the mere implementation using a computer are carried over and does not provide significantly more.
With respect to Claims 3, 13, and 19 reciting “Wherein the RQL is generated for RQL for multi-domain security applications.” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claims 4, 14, and 20 reciting “Wherein the LLM is fine-tuned for a cloud security application.” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claim 7 reciting “Wherein the DSL is a resource query language (RQL).” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claim 7 reciting “Wherein the plurality of data source domains includes a configuration data set, an Identity and Asset Management (IAM) data set, and a vulnerability data set.” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claims 9 and 16 reciting “Wherein the DSL is a resource query language (RQL).” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claim 22 reciting “Wherein the same asset attribute includes one or more of the following: asset name, asset type, asset class, cloud type, cloud account, finding name, finding type, and/or finding severity;” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
With respect to Claim 23 reciting “Wherein the filtering criteria includes one or more of the following: a vulnerability having a Common Vulnerability Scoring System (CVSS) score within an allowed range, a vulnerability having an allowed severity, and/or a vulnerability having an allowed vulnerability identifier;” identified as insignificant extra-solution activity above this is also WURC when claimed in a merely generic manner as court-identified see MPEP 2106.05(d)(II)(iv).
Looking at the claim as a whole does not change this conclusion and the claim is ineligible.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1, 3-11, 13-17, 19-20, and 22-23 are rejected under 35 U.S.C. 103 as being unpatentable over Malkiel et al. (U.S. Pre-Grant Publication No. 2021/0182935, hereinafter referred to as Malkiel), and further in view of Das et al. (U.S. Patent No. 11,017,764, hereinafter referred to as Das) and Kim et al. (U.S. Pre-Grant Publication NO. 2021/0303555, hereinafter referred to as Kim).
Regarding Claim 1:
Malkiel teaches a system, comprising:
A processor configured to:
Automatically generate a seed dataset for a domain specific language (DSL),
Malkiel teaches “the seed item 108 is any item within the selected domain which has been selected or otherwise indicated by a user from a plurality of items within the selected domain” (Para. [0033]). thereby teaching a seed dataset for a domain specific language being generated from the selected content.
Expand, using a Large Language Model (LLM), the seed dataset for the DSL to obtain an expanded dataset for the DSL,
Malkiel teaches “the trained language model infers item similarities between a seed item and a plurality of candidate items via the trained language model at 1406. The trained language model outputs a set of recommended items from a plurality of candidate items in a catalog maximizing item similarities between the seed item and the candidate items at 1408.” (Para. [0109]).
Validate the expanded dataset for the DSL, wherein the expanded dataset for the DSL is input to the LLM for fine tune training of the LLM;
Malkiel teaches “the trained language model 106 generates a set of one or more recommended items 110 from a plurality of candidate items” and “The plurality of candidate items are described in the domain-specific corpus 104, which is used to train or fine-tune a specialist language model, such as, but not limited to, the trained language model 106” (Para. [0034]) thereby teaching that the expanded dataset of recommended items was input to the LLM for fine tune training.
Malkiel further teaches validating the expanded dataset by teaching “the domain-related corpus in other examples is split into two separate sets, a trained data set and a validation data set” (Para. [0044]) and “after every epoch, the model performance is evaluated on the hold-out validation data set” where “at the end of the language model training, the best model is selected according to the accuracy of the reconstructing masked words, reported on the validation set” (Para. [0051]).
Output search results for a query based on the validated expanded dataset; and
Malkiel teaches output search results (top K recommended items) based on an expanded dataset used to previously fine tune the LLM (the trained language model 202) by teaching “In order to produce similarity for a given seed item S, a calculation component 210 calculates a cosine similarity 212 between the embedding vector of the seed item “S” and the embedding vectors of all the other items in the catalog. A scoring component 216 optionally converts the cosine similarity 212 to a set of similarity scores 218 for each candidate item. A recommendation component 214 sorts the candidate items in a descending order according to this cosine score for each item in the set of similarity scores 218. The recommendation component 214 retrieves the top K items 220 from the plurality of candidate items 208 as recommendations to form a set of recommended items 222, where “K” represents a user-configurable or user-selected number of items. The set of recommended items 222 is a set of one or more recommended items which is the same or similar to the seed item with regard to one or more attributes or properties, such as, but not limited to, the recommended item 224.” (Para. [0040])
A memory coupled to the processor and configured to provide the processor with instructions.
Malkiel teaches “Some aspects and examples disclosed herein are directed to a system, method and/or computer executable instructions for generating cold-start recommendations based on title and description relationships comprising: a processor; and a computer-readable medium storing instructions that are operative upon execution by the processor” (Para. [0114]).
Malkiel explicitly teaches all of the elements of the claimed invention as recited above except:
Wherein the DSL includes a resource query language (RQL).
Wherein the automatically generating of the seed dataset comprises to: Generate a natural language representation of the RQL to obtain the seed dataset;
Wherein the expanding comprising to perform one or more of the following:
A) for a first set of samples using single attribute values having same asset attributes, generate, using the LLM, queries having a plurality of attribute values for the same asset attribute to obtain the expanded dataset for the DSL, wherein the same asset attribute includes one or more of the following: asset name, asset type, asset class, cloud type, cloud account, finding name, finding type, and/or finding severity; and/or
B) for a second set of samples using single attribute values having the same asset attribute and have different filtering criteria, generate, using the LLM, queries that mix the natural language of the RQL to obtain the expanded dataset for the DSL, wherein the filtering criteria includes one or more of the following: a vulnerability having a Common Vulnerability Scoring System (CVSS) score within an allowed range, a vulnerability having an allowed severity and/or a vulnerability having an allowed vulnerability identifier;
However, in the related field of endeavor of requesting domain specific information, Das teaches:
Wherein the DSL includes a resource query language (RQL).
Das teaches “receiving the NL request” at a request processing engine 2020 for translating “the NL request 2015 to the domain-specific language (DSL) request 2045” (Col. 74 Lines 14-33 & Fig. 21) thereby teaching a query/request in a language for a specific resource/domain.
Thus, it would have been obvious to one of ordinary skill in the art, having the teachings of Das and Malkiel at the time that the claimed invention was effectively filed, to have combined the conversion of natural langue request to a domain specific language request, as taught by Das, with the systems and methods for recommending domain specific items, as taught by Malkiel.
One would have been motivated to make such combination because Das teaches allowing for a user to enter a natural language request and converting the natural language request to a domain specific language request (Col. 74 Lines 14-33) and it would have been obvious to a person having ordinary skill in the art that enabling users to request domain specific content through natural language input would expand the search options for users when combined with Malkiel.
Das and Malkiel explicitly teach all of the elements of the claimed invention as recited above except:
Wherein the automatically generating of the seed dataset comprises to: Generate a natural language representation of the RQL to obtain the seed dataset;
Wherein the expanding comprising to perform one or more of the following:
A) for a first set of samples using single attribute values having same asset attributes, generate, using the LLM, queries having a plurality of attribute values for the same asset attribute to obtain the expanded dataset for the DSL, wherein the same asset attribute includes one or more of the following: asset name, asset type, asset class, cloud type, cloud account, finding name, finding type, and/or finding severity; and/or
B) for a second set of samples using single attribute values having the same asset attribute and have different filtering criteria, generate, using the LLM, queries that mix the natural language of the RQL to obtain the expanded dataset for the DSL, wherein the filtering criteria includes one or more of the following: a vulnerability having a Common Vulnerability Scoring System (CVSS) score within an allowed range, a vulnerability having an allowed severity and/or a vulnerability having an allowed vulnerability identifier;
However, in the related field of endeavor of enhancing training data, Kim teaches:
Wherein the automatically generating of the seed dataset comprises to: Generate a natural language representation of the RQL to obtain the seed dataset;
Kim teaches “facilitating increased levels of scaling training data for, in turn, more sophisticated and robust machine-learning models” (Para. [0024]) where “improved NL2QL machine-learning models from training on enhanced training data comprising NL2QL pairs generated via logical-form dialogue construction” (Para.[0093]) and “for an applicable predicate of the subsequent dialogue sequence, the NL2QL pair generation system can generate both a natural language query and a query-language representation by using the NL2QL template” (Para. [0022])).
Wherein the expanding comprising to perform one or more of the following:
A) for a first set of samples using single attribute values having same asset attributes, generate, using the LLM, queries having a plurality of attribute values for the same asset attribute to obtain the expanded dataset for the DSL,
It is noted that Para. [0260] of the present Specification states that “It should be noted that one sample includes two parts: (1) the RQL query generated, and (2) the corresponding natural language (NL) representation for that query.”.
Malkiel teaches “The language model 600 iterates over each candidate item comparing each candidate item to the seed item 640 to generate a set of lists 642.” (Para. [0076]) and “the language model 700 iterates over the data samples of the training set, including textual data 702.” (Para. [0079]) and “The set of recommended items 222 is a set of one or more recommended items which is the same or similar to the seed item with regard to one or more attributes or properties,” (Para. [0040]).Kim teaches “the NL2QL pair generation system exploits inputs comprising a database, lexicon, and an ontology to rapidly generate a large-scale dataset of paired natural language queries and query-language representations” (Para. [0018]) and “the NL2QL pair generation system can generate both a natural language query and a query-language representation by using the NL2QL template” (Para. [0022]).
Therefore, Malkiel in combination with Kim teach for samples using attribute values and asset attributes (data samples of the training set including textual data - Malkiel), generating (using the LLM taught by Malkiel) queries having a plurality of attribute values for the same asset attribute to obtain the expanded dataset for the DSL (generating NL2QL pairs – Kim)
Wherein the same asset attribute includes one or more of the following: asset name, asset type, asset class, cloud type, cloud account, finding name, finding type, and/or finding severity; and/or
Kim teaches asset attributes as at least asset name, asset class, finding name, and finding type, given their broadest reasonable interpretations, by teaching: “For example, and as illustrated in FIG. 2, a first entity (‘Employee’) may correspond to two properties, ‘name’ and ‘hire year.’ Likewise, for instance, a second entity (‘Office’) may correspond to a third property, ‘floor.’” (Para. [0047]).
Malkiel also teaches asset attributes as at least asset name, asset type, asset class, finding name, and finding type, given their broadest reasonable interpretations, by teaching: “Likewise, a domain for wines in another example includes a list of wines. Each wine item in the catalog include a name of a wine and a description of the wine. The description of each item in this non-limiting example may include the color (red or white), the producer or vineyard from which the wine originated, year of bottling, or other descriptive information.” (Para.[0025])
B) for a second set of samples using single attribute values having the same asset attribute and have different filtering criteria, generate, using the LLM, queries that mix the natural language of the RQL to obtain the expanded dataset for the DSL,
It is noted that Para. [0260] of the present Specification states that “It should be noted that one sample includes two parts: (1) the RQL query generated, and (2) the corresponding natural language (NL) representation for that query.”
Malkiel teaches “The language model 600 iterates over each candidate item comparing each candidate item to the seed item 640 to generate a set of lists 642.” (Para. [0076]) and “the language model 700 iterates over the data samples of the training set, including textual data 702.” (Para. [0079]) and “The set of recommended items 222 is a set of one or more recommended items which is the same or similar to the seed item with regard to one or more attributes or properties,” (Para. [0040]).Kim teaches “the NL2QL pair generation system exploits inputs comprising a database, lexicon, and an ontology to rapidly generate a large-scale dataset of paired natural language queries and query-language representations” (Para. [0018]) and “the NL2QL pair generation system can generate both a natural language query and a query-language representation by using the NL2QL template” (Para. [0022]).
Therefore, Malkiel in combination with Kim teach for samples using asset values and asset attributes (data samples of the training set including textual data - Malkiel), generating (using the LLM taught by Malkiel) queries having a plurality of assets or values for the same asset attributes to obtain the expanded dataset for the DSL (generating NL2QL pairs – Kim)
Wherein the filtering criteria includes one or more of the following: a vulnerability having a Common Vulnerability Scoring System (CVSS) score within an allowed range, a vulnerability having an allowed severity and/or a vulnerability having an allowed vulnerability identifier;
Das teaches “Examples of data models can include electronic mail, authentication, databases, intrusion detection, malware, application state, alerts, compute inventory, network sessions, network traffic, performance, audits, updates, vulnerabilities, etc. Data models and their objects can be designed by knowledge managers in an organization, and they can enable downstream users to quickly focus on a specific set of data.” (Col. 42 Lines 39-50).
Thus, it would have been obvious to one of ordinary skill in the art, having the teachings of Kim, Das, and Malkiel at the time that the claimed invention was effectively filed, to have combined the improved NL2QL machine-learning models due to being trained on enhanced training data comprising NL2QL pairs generated via logical-form dialogue construction, as taught by Kim, with the conversion of natural langue request to a domain specific language request, as taught by Das, and the systems and methods for recommending domain specific items, as taught by Malkiel.
One would have been motivated to make such combination because Kim teaches “the third-party crowdsourcing system 374 can complement and add variety to NL2QL pairs, thereby further improving the NL2QL pair database 378” (Para. [0087]) and it would have been obvious to a person having ordinary skill in the art that allowing a third-party to complement and add variety to the NL2QL pairs would improve the diversity of the data.
Regarding Claim 3:
Kim, Das, and Malkiel further teach:
Wherein the RQL is generated for RQL for multi-domain security applications.
Das teaches “the search head combines the partial results and/or events received from the indexers to produce a final result for the query. In some examples, the results of the query are indicative of performance or security of the IT environment” (215)
Regarding Claim 4:
Kim, Das, and Malkiel further teach:
Wherein the LLM is fine-tuned for a cloud security application.
Malkiel teaches “The language model 102 in some examples is an untrained general language model or a pre-trained language model that is trained or fine-tuned using a domain-specific corpus 104 to create a more domain-specific, specialized trained language model 106.” (Para. [0033]).
Regarding Claim 5:
Kim, Das, and Malkiel further teach:
Wherein the LLM is fine-tuned for performing automated entity extraction for multi-domain security applications.
Malkiel teaches “The language model 102 in some examples is an untrained general language model or a pre-trained language model that is trained or fine-tuned using a domain-specific corpus 104 to create a more domain-specific, specialized trained language model 106.” (Para. [0033]).
Regarding Claim 6:
Kim, Das, and Malkiel further teach:
Wherein the LLM is fine-tuned for performing a cross-domain search to generate a search result using a plurality of data source domains that includes using a planner, executor, and aggregator to collect distinct results from each of the plurality of data source domains.
Malkiel teaches “The language model 102 in some examples is an untrained general language model or a pre-trained language model that is trained or fine-tuned using a domain-specific corpus 104 to create a more domain-specific, specialized trained language model 106.” (Para. [0033]).
Regarding Claim 7:
Kim, Das, and Malkiel further teach:
Wherein the DSL is a resource query language (RQL),
Das teaches “receiving the NL request” at a request processing engine 2020 for translating “the NL request 2015 to the domain-specific language (DSL) request 2045” (Col. 74 Lines 14-33 & Fig. 21).
Wherein the LLM is fine-tuned for performing a natural language (NL) query for a plurality of data source domains for a cloud security application, and
Malkiel teaches “The language model 102 in some examples is an untrained general language model or a pre-trained language model that is trained or fine-tuned using a domain-specific corpus 104 to create a more domain-specific, specialized trained language model 106.” (Para. [0033]).
Wherein the plurality of data source domains includes a configuration data set, an Identity and Asset Management (IAM) data set, and a vulnerability data set.
Das teaches “The enterprise security application can process many types of security-related information. In general, this security-related information can include any information that can be used to identify security threats. For example, the security-related information can include network-related information, such as IP addresses, domain names, asset identifiers, network traffic volume, uniform resource locator strings, and source addresses” and “Security-related information can also include malware infection data and system configuration information, as well as access control information, such as login/logout information and access failure notifications. The security-related information can originate from various sources within a data center, such as hosts, virtual machines, storage devices and sensors. The security-related information can also originate from various sources in a network, such as routers, switches, email servers, proxy servers, gateways, firewalls and intrusion-detection systems” (Col. 55 Lines 12-49).
Regarding Claim 8:
Kim, Das, and Malkiel further teach the processor is further configured to:
Generate an RQL query in response to a natural language query using the fine-tuned LLM.
Das teaches “receiving the NL request” at a request processing engine 2020 for translating “the NL request 2015 to the domain-specific language (DSL) request 2045” (Col. 74 Lines 14-33 & Fig. 21).
Regarding Claim 9:
Kim, Das, and Malkiel further teach:
Wherein the DSL is a resource query language (RQL), and
Das teaches “receiving the NL request” at a request processing engine 2020 for translating “the NL request 2015 to the domain-specific language (DSL) request 2045” (Col. 74 Lines 14-33 & Fig. 21).
Wherein the processor is further configured to automatically generate a configuration policy in RQL from a natural language (NL) input.
Das teaches “receiving the NL request” at a request processing engine 2020 for translating “the NL request 2015 to the domain-specific language (DSL) request 2045” (Col. 74 Lines 14-33 & Fig. 21).
Das further teaches “events are field-searchable using one or more configuration files associated with the events. Each configuration file includes one or more field names, where each field name is associated with a corresponding extraction rule and a set of events to which that extraction rule applies” (Col. 7 Line 52 – Col. 8 Line 15).
Regarding Claim 10:
Kim, Das, and Malkiel further teach:
Wherein the LLM is fine-tuned for performing a cross-domain search to generate a search result using a plurality of data source domains to collect distinct results from each of the plurality of data source domains, and
It is noted that the limitation recites intended use language of “…for performing…”, “…to generate…”, and “…to collect…” and is therefore not being given patentable weight. The scope of the limitation is understood as reciting “wherein the LLM is fine-tuned”.
Malkiel teaches “The language model 102 in some examples is an untrained general language model or a pre-trained language model that is trained or fine-tuned using a domain-specific corpus 104 to create a more domain-specific, specialized trained language model 106.” (Para. [0033]).
Wherein the processor is further configured to automatically generate an output in response to a natural language query, wherein one or more of the plurality of data source domains are searched using queries in RQL.
Das teaches “receiving the NL request” at a request processing engine 2020 for translating “the NL request 2015 to the domain-specific language (DSL) request 2045” and “execute the DSL request 2045” (Col. 74 Lines 14-33 & Fig. 21). Das further teaches “the NL application 1940 applies the DSL request 2045 to the associated domain-specific data source 1920” (Col. 75 Lines 18-25).
Regarding Claim 11:
All of the limitations herein are similar to some or all of the limitations of Claim 1.
Regarding Claim 13:
All of the limitations herein are similar to some or all of the limitations of Claim 3.
Regarding Claim 14:
All of the limitations herein are similar to some or all of the limitations of Claim 4.
Regarding Claim 15:
All of the limitations herein are similar to some or all of the limitations of Claim 8.
Regarding Claim 16:
All of the limitations herein are similar to some or all of the limitations of Claim 9.
Regarding Claim 17:
Some of the limitations herein are similar to some or all of the limitations of Claim 1.
Kim, Das, and Malkiel further teach:
A non-transitory computer readable medium comprising computer instructions.
Malkiel teaches “a system, method and/or computer executable instructions for generating cold-start recommendations based on title and description relationships comprising: a processor; and a computer-readable medium storing instructions that are operative upon execution by the processor” (Para. [0114]).
Regarding Claim 19:
All of the limitations herein are similar to some or all of the limitations of Claim 3.
Regarding Claim 20:
All of the limitations herein are similar to some or all of the limitations of Claim 4.
Regarding Claim 22:
All of the limitations herein are similar to some or all of the limitations of Claim 1.
Regarding Claim 23:
All of the limitations herein are similar to some or all of the limitations of Claim 1.
Response to Amendment
Applicant’s Amendments, filed on 12/8/2025, are acknowledged and addressed herein.
Response to Arguments
On pages 9-10 of the Remarks filed on 12/8/2025, Applicant states that “Kim does not teach the recited asset attribute or the recited filtering criteria. The other applied references fail to cure the deficiencies of Kim. Accordingly, the applied references fail to teach or render obvious [the amended expanding step of the independent claims]”.
Applicant’s argument is agreed that Kim alone does not teach all of the amended claim language, however, upon further review, Kim in combination with Malkiel and Das does appear to teach the amended limitations as is addressed in full in the rejection above.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Peraud et al. (U.S. Pre-Grant Publication No. 2022/0230089) teaches a classifier may be trained with less than all datasets manually annotated with labels. A small subset of verbatims may be manually labeled with topic labels as seeds. Data augmentations can be used to acquire seed verbatim sets for known topics and to assign temporary pseudo labels to the rest of the verbatims based on their vector space proximity to the labeled seed verbatims. The training may involve classification epochs during which embeddings are updated with the assumption that the pseudo labels are ground-truth labels. The training may also involve labeling epochs during which the updated embeddings are used to update the vectors corresponding to the verbatims, and pseudo labels are updated based on updated vector coordinates in the vector space. As the training process progresses through the epochs, the embeddings will converge.The reference further teaches “This base natural language model 116 may then be fine-tuned using relevant domain language to develop a domain natural language model 118. For example, if the relevant domain is gaming, then the base natural language model 116 may be fine-tuned on a corpus of millions of examples of gaming domain language spanning a variety of gaming products and a variety of entry points for gaming user feedback” (Para. -0075])The reference further teaches “This fine-tuned domain natural language model 118 may then be further trained and scored with the seed verbatims 112 that have been manually labeled by human SMEs 110.” (Para. [0076]).
Akbacak et al. (U.S. Pre-Grant Publication No. 2015/0332670) teaches training language models using in-domain-like data collected automatically from one or more data sources. The data sources (such as text data or user-interactional data) are mined for specific types of data, including data related to style, content, and probability of relevance, which are then used for language model training. In one embodiment, a language model is trained from features extracted from a knowledge graph modified into a probabilistic graph, where entity popularities are represented and the popularity information is obtained from data sources related to the knowledge. Embodiments of language models trained from this data are particularly suitable for domain-specific conversational understanding tasks where natural language is used, such as user interaction with a game console or a personal assistant application on personal device.
Poirier et al. (U.S. Pre-Grant Publication No. 2024/0202464) teaches managing a plurality of agents to generate a response to a query using a multimodal model. An example method uses the plurality of agents to iteratively determine subsequent outputs of the multimodal model satisfies the query. It can generate a respective context associated with a respective output of the multimodal model. And determine, by the multimodal model based on the respective context, whether the respective subsequent output satisfies the query.
Wang et al. (U.S. Pre-Grant Publication No. 2025/0036616) teaches managing report requests are described. A device associated with an identity management platform may receive a message from a user of an organization via a client device. The message may include a natural language and may indicate a request for information associated with the organization. In response to the request, the device may generate a first query based on translating the message into an intermediary language using a machine learning model. The intermediary language may be associated with the identity management platform. The device may transmit a report to the user via the client device. The report may be based on the first query and include the information associated with the organization.The reference further teaches “the query component 530 may be configured to support training the machine learning model to translate the natural language into the intermediary language, where translating the message into the first query is based on the training.” (Para. [0083]).
Gottlob et al. (U.S. Pre-Grant Publication No. 2025/0045256) teaches (i) accessing and/or modifying a database to be automatically curated, (ii) optionally accessing additional data or information sources for further useful data or information, (iii) using one or more pre-trained large language models (LLMs) accessed via API or other connections, issuing prompts and retrieving prompt-answers and executing database curation requests that specify database curation tasks to be performed on at least one sub-structure of the database, the tasks comprising (a) a database enrichment task to compute new data records to be inserted into the database sub-structure, (b) a database verification task to verify, using the one or more LLMs, data contained in the sub-structure, and identify incorrect data, (c) a database update, and (d) a null-value or a missing value replacement task. The requested tasks are automatically performed via a computation comprising an adaptively generated prompt sequence.
Foreign Reference CN117726897A teaches a large language model can be used, for example, a large language model can be used, for example, a large language model can be used, for example, a large language model can be used, for example, a large language model can be used, for example, a large language model can ChatGPT, Wen Hsin and so on, expanding the attribute information. For example, the target attribute value corresponding to each type of attribute information can be determined specifically; expanding the text template based on the target attribute value to obtain at least one image description information.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ROBERT F MAY whose telephone number is (571)272-3195. The examiner can normally be reached Monday-Friday 9:30am to 6pm.
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, Boris Gorney can be reached on 571-270-5626. 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.
/ROBERT F MAY/Examiner, Art Unit 2154 1/9/2025
/SYED H HASAN/Primary Examiner, Art Unit 2154