DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
IDS submitted on 09/22/2025 has been considered by the examiner.
Claim Objections
Claim 4, 9 and 14 are objected to because of the following informalities: recited word "Max." in aforementioned claims are not elaborated. "Max." should be written as “Maximum”.
Appropriate corrections are required.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claim 1-4, 6-9 and 11-14 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Claim 1 is directed to a process. The claim recites “pre-processing, ……. the at least one query to obtain a pre-processed at least one query; segmenting, ……, the pre-processed at least one query into one or more sets, each comprising one or more sub-queries; computing, ……, an embedding for each of the one or more sub-queries in the one or more sets; computing, ….., a similarity metric between the embedding of the one or more sub-queries in the one or more sets, and an embedding of a natural language part of a plurality of templatized domain statements; computing, ……a weighted set score for each of the one or more sets using the one or more sub-queries in the one or more sets and the similarity metric, wherein the weighted set score represents extent of similarity of the at least one query with each of the plurality of templatized domain statements; and generating, …….., a SQL query for the at least one query in natural language, using the database schema, the database meta data”.
The processes of pre-processing queries, segmenting the query into sub-queries, embedding the sub-queries, computing similarity metric between embedded sub-queries and templatized statements, computing weighted score for sub-queries and generation of SQL query for natural language query involve observation, judgement and evaluation. Aforementioned processes can practically be performed in the human mind. Thus, the claim is directed to an abstract idea falling within the grouping of mental steps, see MPEP 2106.04(a)(2)(III).
At step 2A, prong 2, this judicial exception is not integrated into a practical application. In particular, the claim recites additional elements – “the retrieved one or more templatized domain statements”, “retrieving, ……., one or more of the plurality of templatized domain statements based on the weighted set score”; “receiving, via one or more hardware processors, a) at least one query in natural language, b) a database schema, and c) a database meta data from domain specific data source”. Above mentioned steps of retrieving and receiving contents recite insignificant extra-solution activity of mere data gathering is “obtaining information” as identified in MPEP 2106.05 (g). The claim also mentions generic computer and generic computer components such as “hardware processors”, use of a computer or generic computer components to execute abstract idea in the form of software constitutes use of the computer or its components as a tool. Accordingly, this additional elements do not integrate the abstract idea into a practical application. Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application. Therefore, claim is directed to an abstract idea.
At step 2B, the claims don’t include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above the additional elements recites insignificant extra-solution activity of data gathering and outputting/transmitting data such are also well- understood, routine, and conventional. Further, sending/transmitting data is insignificant extra-solution activity of data transmission, such is also well- understood, routine, and conventional (OIP Techs., Inc., v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014) (computer receives and sends information over a network)). Looking at the limitations in combination and the claim as a whole does not change this conclusion and the claim is ineligible.
Claim 6 differs from claim 1 in that the steps of the claimed method is implemented by instructions when executed by one or more processors. The invention of claim 6 is a system including one or more processors and a memory storing the instructions to perform recited steps. For reasons discussed above, the claimed steps are directed to mental steps. Use of a processor to execute instructions stored in memory constitutes use of a generic computer as a tool and does not constitute an application of significantly more than the abstract idea. Accordingly, claim 6 is not patent eligible.
Claim 11 differs from claim 1 in that it recites a non-transitory computer readable medium including a sequence of instructions which when executed perform the method of claim 1. For reasons discussed above, the claimed process is directed to mental steps. Use of a non-transitory medium to store instructions which when executed perform the method of claim 1 constitutes use of a component of a generic computer as a tool and does not constitute an application of significantly more than the abstract idea. Accordingly, claim 11 is not patent eligible.
Dependent claim 2-4 are directed to the same abstract idea as the independent claim from which they depend and further recite limitations – “pre-processing comprises replacing a numerical value and a date to a predefined fixed integer”, “creating….., a list of a plurality of individual words in the pre-processed at least one query; generating, ….. the one or more sets comprising a plurality of combinations of the one or more sub-queries in the one or more sets by iterating through the list of a plurality of individual words, and joining two or more of the plurality of individual words”, “iteratively generating, ……. the one or more sub-queries until all of the plurality of individual words are used in at least one of the one or more sets, wherein each of the one or more sets comprises one or more sub-queries of same or different length of plurality of individual words, and wherein each of the one or more sub-queries among the one or more sets and each of the one or more set among the one or more sets are unique; and
generating, …… the one or more sub-queries in the one or more sets until the one or more sub-queries are matching with the at least one query in natural language”. The process of replacing values in pre-processed query, creating a list words from preprocessed query, generation of combination of sub-queries by iterating through the word list and joining words, iteratively generating sub-queries until end of the list of words having different length, generation of unique sub-query sets, generating sub-query such that sub-queries match with query in natural language involve observation, judgement and evaluation and can practically be performed in human mind. Accordingly, recited limitations fall into abstract idea groupings of mental process (see MPEP 2106.04(a)(2)(III)) under Step 2A, prong 1 of the 2019 PEG. Therefore, aforementioned processes can practically be performed in the human mind and directed to an abstract idea.
Claims further recite “computing the weighted set score by using the length of sub-query representing number of words in the sub- query and the one or more similarity metric is represented as: Set_Score = ∑ (length of sub-string (number of words)) * (Max. similarity score with a templatized domain statement), where, the length of sub-string is the number of words in the one or more sub-queries, and the Max. similarity score with a templatized domain statement is the similarity metric of the one or more sub-queries in the one or more sets”. Computing weighted set score using the length of sub-query representing number of words in the sub- query and the one or more similarity metric is represented as: Set_Score = ∑ (length of sub-string (number of words)) * (Max. similarity score with a templatized domain statement),
where, the length of sub-string is the number of words in the one or more sub-queries, and the Max. similarity score with a templatized domain statement is the similarity metric of the one or more sub-queries in the one or more sets is mathematical formulas or mathematical calculations. Accordingly, recited limitations fall into abstract idea groupings of Mathematical concepts (see MPEP 2106.04(a)(2)(I)) under Step 2A, prong 1 of the 2019 PEG..
At step 2A, prong 2, this judicial exception is not integrated into a practical application. In particular, the claims recite additional elements –generic computer and generic computer components such as “hardware processors”, the use of a computer or generic computer components to execute abstract idea in the form of software constitutes use of the computer or its components as a tool. Accordingly, this additional elements do not integrate the abstract idea into a practical application. Viewing the additional limitations together and the claims individually as a whole, nothing provides integration into a practical application. Therefore, claim 2-4 directed to an abstract idea.
At step 2B, the claims don’t include additional elements that are sufficient to amount to significantly more than the judicial exception. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept, see MPEP 2106.05 (f). Looking at the limitations in combination and the claims individually as a whole does not change this conclusion and the claim is ineligible. Accordingly, claim 2-4 are not patent eligible.
Claim 7, 8 and 9 differ from claim 2, 3 and 4 respectively in that the steps of the claimed method are implemented by instructions when executed by one or more processors. The invention of claim 7, 8 and 9 is a system including one or more processors and a memory storing the instructions to perform recited steps. For reasons discussed above, the claimed steps are directed to mental steps. Use of a processor to execute instructions stored in memory constitutes use of a generic computer as a tool and does not constitute an application of significantly more than the abstract idea. Accordingly, claim 7, 8 and 9 are not patent eligible.
Claim 12, 13 and 14 differ from claim 2, 3 and 4 respectively in that they recite a non-transitory computer readable medium including a sequence of instructions which when executed perform the method of claim 2, 3 and 4. For reasons discussed above, the claimed process is directed to mental steps. Use of a non-transitory medium to store instructions which when executed perform the method of claim 2, 3 and 4 constitutes use of a component of a generic computer as a tool and does not constitute an application of significantly more than the abstract idea. Accordingly, claim 12, 13 and 14 are not patent eligible.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim 1-2, 6-7, and 11-12 are rejected under 35 U.S.C. 103 as being unpatentable over Brende, Hans et al (PGPUB Document No. 20240394251), hereafter referred as to “Brende”, in view of Ghosh, Dipanjan et al (PGPUB Document No. 20260056945), hereafter, referred to as “Ghosh”, in further view of Rizk, Yara et al (PGPUB Document No. 20250265489), hereafter, referred to as “Rizk”.
Claim 1, Brende teaches A processor implemented method, comprising: receiving, via one or more hardware processors, a) at least one query in natural language, b) a database schema, and c) a database meta data from domain specific data source(Brende, para 0116 discloses receiving a natural language query having databases schema (tables) and metadata (date field for date range) “abstraction language interface 213 obtains a natural language query from a user, which could be a simple question or a complex instruction pertaining to data retrieval or manipulation….abstraction language interface 213 determines the main data subjects by linking nouns and noun phrases to corresponding entities in the database schema, such as mapping “sales” to a “Sales” database table….”);
pre-processing, via the one or more hardware processors, the at least one query to obtain a pre-processed at least one query(Brende, para 0110 discloses pre-processing the queries by outlining the basic structure of the query while leaving the place holder for dynamic values of the queries “the prompt may comprise a predefined, partially complete portion that outlines the basic structure of the query, while the dynamic portion allows for the inclusion of specific details that are only available at the time of the query, such as the exact data fields to be retrieved…..”), segmenting, via the one or more hardware processors he pre-processed at least one query into one or more sets, each comprising one or more sub-queries(Brende, para 0180 discloses generation of sub-queries “the generator might split the query into smaller sub-queries that can be processed in parallel”);
But Brende does not explicitly teach computing, via the one or more hardware processors, an embedding for each of the one or more sub-queries in the one or more sets;
computing, via the one or more hardware processors, a similarity metric between the embedding of the one or more sub-queries in the one or more sets, and an embedding of a natural language part of a plurality of templatized domain statements;
computing, via the one or more hardware processors, a weighted set score for each of the one or more sets using the one or more sub-queries in the one or more sets and the similarity metric, wherein the weighted set score represents extent of similarity of the at least one query with each of the plurality of templatized domain statements;
retrieving, via the one or more hardware processors, one or more of the plurality of templatized domain statements based on the weighted set score;
and generating, via the one or more hardware processors, a SQL query for the at least one query in natural language, using the database schema, the database meta data, and the retrieved one or more templatized domain statements.
However, in the same field of endeavor of embedding feature matching Ghosh teaches computing, via the one or more hardware processors, an embedding for each of the one or more sub-queries in the one or more sets(Ghosh, para 0009 discloses embedding for natural language queries/sub-queries “in response to receiving a user question, e.g., in natural language format, using an embedding model to obtain an embedded user question” );
computing, via the one or more hardware processors, a similarity metric between the embedding of the one or more sub-queries in the one or more sets, and an embedding of a natural language part of a plurality of templatized domain statements(Ghosh, para 0055 discloses finding similarity matching of query embedding with template statement or template questions “performing a first similarity matching process to match the embedded user question with a standard question template in a question template vector database”);
and generating, via the one or more hardware processors, a SQL query for the at least one query in natural language, using the database schema, the database meta data, and the retrieved one or more templatized domain statements(Ghosh, para 0009 discloses matching a query template by similarity matching for SQL query generation “performing a first similarity matching process to match the embedded user question with a standard question template in a question template vector database; extracting parameters from the user question to populate the standard question template; populating the standard question template with the extracted parameters to generate a final question; and providing the final question and the standardized context information to a generative AI model, e.g., a transformer-based model that has been trained for SQL query generation, that converts the final question into a query code”; where Brende in para 0116 discloses receiving a natural language query having databases schema (tables) and metadata (date field for date range)).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of matching query template based on embedding similarity of Ghosh into natural language query conversion to SQL statement of Brende to produce an expected result of consistent SQL generation from natural language queries. The modification would be obvious because one of ordinary skill in the art would be motivated to present consistent output results by using similarity matching between embedded question and query template (Ghosh, para 0030).
But Brende and Ghosh do not explicitly teach computing, via the one or more hardware processors, a weighted set score for each of the one or more sets using the one or more sub-queries in the one or more sets and the similarity metric, wherein the weighted set score represents extent of similarity of the at least one query with each of the plurality of templatized domain statements;
retrieving, via the one or more hardware processors, one or more of the plurality of templatized domain statements based on the weighted set score;
However, in the same field of endeavor of feature matching Rizk teaches computing, via the one or more hardware processors, a weighted set score for each of the one or more sets using the one or more sub-queries in the one or more sets and the similarity metric, wherein the weighted set score represents extent of similarity of the at least one query with each of the plurality of templatized domain statements(Rizk, para 0075 discloses query and template matching bases on weighed similarity, which can similarly be applied to sub-queries and statement template matching “If the input query requested a response from a recommended model, one implementation of module 350 uses the model-prompt template pair with the highest predicted or highest average predicted confidence value to generate a candidate response to the input query, and incorporates the candidate response into a response to the input query. …….. Another implementation of module 350 computes a weighted average of each candidate's similarity scores”);
retrieving, via the one or more hardware processors, one or more of the plurality of templatized domain statements based on the weighted set score (Rizk, para 0075 discloses response selection by highest weighted similarity score and which can similarly be applied for retrieving/selecting statement template “Module 350 incorporates the candidate response with the highest combined (e.g., the highest average) similarity score into a response to the input query”);
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of matching query template based on weighted similarity scores of Rizk into natural language query conversion to SQL statement of Brende and Ghosh to produce an expected result of consistent SQL generation from natural language queries. The modification would be obvious because one of ordinary skill in the art would be motivated to present relevant response by selecting templates based predicted confidence values(Rizk, abstract).
Regarding claim 2, Brende, Ghosh and Rizk teach all the limitation of claim 1 and Brende further teaches wherein the pre-processing comprises replacing a numerical value and a date to a predefined fixed integer(Brende, para 0110 discloses pre-processing the queries by outlining the basic structure of the query while leaving the place holder for dynamic values of the queries and further replacing the place holders “the prompt may comprise a predefined, partially complete portion that outlines the basic structure of the query, while the dynamic portion allows for the inclusion of specific details that are only available at the time of the query, such as the exact data fields to be retrieved…..The predefined portion of the prompt could look like this: “Retrieve total sales from Q_for the following products: “. Here, “Q_” represents a placeholder where the specific quarter needs to be specified, and the product list is left dynamic for user input. As the financial analyst interacts with the system, they specify that they are interested in “Q2” and want to see sales data for “Product A, Product B, and Product C.” The dynamic portion of the prompt is then completed in real-time, filling in the specified quarter and product names”).
Claim 6, Brende teaches A system, comprising: one or more hardware processors; a communication interface; and a memory storing a plurality of instructions, wherein the plurality of instructions cause the one or more hardware processors to(Brende, Fig. 9 discloses a system comprising storage medium and processors): receive a) at least one query in natural language, b) a database schema, and c) a database meta data from domain specific data source (Brende, para 0116 discloses receiving a natural language query having databases schema (tables) and metadata (date field for date range) “abstraction language interface 213 obtains a natural language query from a user, which could be a simple question or a complex instruction pertaining to data retrieval or manipulation….abstraction language interface 213 determines the main data subjects by linking nouns and noun phrases to corresponding entities in the database schema, such as mapping “sales” to a “Sales” database table….”);
pre-process the at least one query to obtain a pre- processed at least one query(Brende, para 0110 discloses pre-processing the queries by outlining the basic structure of the query while leaving the place holder for dynamic values of the queries “the prompt may comprise a predefined, partially complete portion that outlines the basic structure of the query, while the dynamic portion allows for the inclusion of specific details that are only available at the time of the query, such as the exact data fields to be retrieved…..”); segment the pre-processed at least one query into one or more sets, each comprising one or more sub-queries(Brende, para 0180 discloses generation of sub-queries “the generator might split the query into smaller sub-queries that can be processed in parallel”);
But Brende does not explicitly teach compute an embedding for each of the one or more sub- queries in the one or more sets; compute a similarity metric between the embedding of the one or more sub-queries in the one or more sets, and an embedding of a natural language part of a plurality of templatized domain statements; compute a weighted set score for each of the one or more sets using the one or more sub-queries in the one or more sets and the similarity metric, wherein the weighted set score represents extent of similarity of the at least one query with each of the plurality of templatized domain statements; retrieve one or more of the plurality of templatized domain statements based on the weighted set score; and generate a SQL query for the at least one query in natural language, using the database schema, the database meta data, and the retrieved one or more templatized domain statements.
However, in the same field of endeavor of embedding feature matching Ghosh teaches compute an embedding for each of the one or more sub- queries in the one or more sets (Ghosh, para 0009 discloses embedding for natural language queries/sub-queries “in response to receiving a user question, e.g., in natural language format, using an embedding model to obtain an embedded user question” );
compute a similarity metric between the embedding of the one or more sub-queries in the one or more sets, and an embedding of a natural language part of a plurality of templatized domain statements (Ghosh, para 0055 discloses finding similarity matching of query embedding with template statement or template questions “performing a first similarity matching process to match the embedded user question with a standard question template in a question template vector database”); and generate a SQL query for the at least one query in natural language, using the database schema, the database meta data, and the retrieved one or more templatized domain statements(Ghosh, para 0009 discloses matching a query template by similarity matching for SQL query generation “performing a first similarity matching process to match the embedded user question with a standard question template in a question template vector database; extracting parameters from the user question to populate the standard question template; populating the standard question template with the extracted parameters to generate a final question; and providing the final question and the standardized context information to a generative AI model, e.g., a transformer-based model that has been trained for SQL query generation, that converts the final question into a query code”; where Brende in para 0116 discloses receiving a natural language query having databases schema (tables) and metadata (date field for date range)).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of matching query template based on embedding similarity of Ghosh into natural language query conversion to SQL statement of Brende to produce an expected result of consistent SQL generation from natural language queries. The modification would be obvious because one of ordinary skill in the art would be motivated to present consistent output results by using similarity matching between embedded question and query template (Ghosh, para 0030).
But Brende and Ghosh do not explicitly teach compute a weighted set score for each of the one or more sets using the one or more sub-queries in the one or more sets and the similarity metric, wherein the weighted set score represents extent of similarity of the at least one query with each of the plurality of templatized domain statements; retrieve one or more of the plurality of templatized domain statements based on the weighted set score;
However, in the same field of endeavor of feature matching Rizk teaches compute a weighted set score for each of the one or more sets using the one or more sub-queries in the one or more sets and the similarity metric, wherein the weighted set score represents extent of similarity of the at least one query with each of the plurality of templatized domain statements (Rizk, para 0075 discloses query and template matching bases on weighed similarity, which can similarly be applied to sub-queries and statement template matching “If the input query requested a response from a recommended model, one implementation of module 350 uses the model-prompt template pair with the highest predicted or highest average predicted confidence value to generate a candidate response to the input query, and incorporates the candidate response into a response to the input query. …….. Another implementation of module 350 computes a weighted average of each candidate's similarity scores”);
retrieve one or more of the plurality of templatized domain statements based on the weighted set score (Rizk, para 0075 discloses response selection by highest weighted similarity score and which can similarly be applied for retrieving/selecting statement template “Module 350 incorporates the candidate response with the highest combined (e.g., the highest average) similarity score into a response to the input query”);
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of matching query template based on weighted similarity scores of Rizk into natural language query conversion to SQL statement of Brende and Ghosh to produce an expected result of consistent SQL generation from natural language queries. The modification would be obvious because one of ordinary skill in the art would be motivated to present relevant response by selecting templates based predicted confidence values(Rizk, abstract).
Regarding claim 7, Brende, Ghosh and Rizk teach all the limitation of claim 6 and Brende further teaches wherein the pre-processing comprises replacing a numerical value and a date to a predefined fixed integer (Brende, para 0110 discloses pre-processing the queries by outlining the basic structure of the query while leaving the place holder for dynamic values of the queries and further replacing the place holders “the prompt may comprise a predefined, partially complete portion that outlines the basic structure of the query, while the dynamic portion allows for the inclusion of specific details that are only available at the time of the query, such as the exact data fields to be retrieved…..The predefined portion of the prompt could look like this: “Retrieve total sales from Q_for the following products: “. Here, “Q_” represents a placeholder where the specific quarter needs to be specified, and the product list is left dynamic for user input. As the financial analyst interacts with the system, they specify that they are interested in “Q2” and want to see sales data for “Product A, Product B, and Product C.” The dynamic portion of the prompt is then completed in real-time, filling in the specified quarter and product names”).
Claim 11, Brende teaches One or more non-transitory machine-readable information storage mediums comprising one or more instructions which when executed by one or more hardware processors cause (Brende, Fig. 9 discloses storage medium and processor for executing codes): receiving a) at least one query in natural language, b) a database schema, and c) a database meta data from domain specific data source(Brende, para 0116 discloses receiving a natural language query having databases schema (tables) and metadata (date field for date range) “abstraction language interface 213 obtains a natural language query from a user, which could be a simple question or a complex instruction pertaining to data retrieval or manipulation….abstraction language interface 213 determines the main data subjects by linking nouns and noun phrases to corresponding entities in the database schema, such as mapping “sales” to a “Sales” database table….”);
pre-processing the at least one query to obtain a pre-processed at least one query(Brende, para 0110 discloses pre-processing the queries by outlining the basic structure of the query while leaving the place holder for dynamic values of the queries “the prompt may comprise a predefined, partially complete portion that outlines the basic structure of the query, while the dynamic portion allows for the inclusion of specific details that are only available at the time of the query, such as the exact data fields to be retrieved…..”); segmenting the pre-processed at least one query into one or more sets, each comprising one or more sub-queries (Brende, para 0180 discloses generation of sub-queries “the generator might split the query into smaller sub-queries that can be processed in parallel”);
But Brende does not explicitly teach computing an embedding for each of the one or more sub-queries in the one or more sets; computing a similarity metric between the embedding of the one or more sub-queries in the one or more sets, and an embedding of a natural language part of a plurality of templatized domain
statements; computing a weighted set score for each of the one or more sets using the one or more sub-queries in the one or more sets and the similarity metric, wherein the weighted set score represents extent of similarity of the at least one query with each of the plurality of templatized domain statements;
retrieving one or more of the plurality of templatized domain statements based on the weighted set score; and generating a SQL query for the at least one query in natural language, using the database schema, the database meta data, and the retrieved one or more templatized domain statements.
However, in the same field of endeavor of embedding feature matching Ghosh teaches computing an embedding for each of the one or more sub-queries in the one or more sets(Ghosh, para 0009 discloses embedding for natural language queries/sub-queries “in response to receiving a user question, e.g., in natural language format, using an embedding model to obtain an embedded user question” );
computing a similarity metric between the embedding of the one or more sub-queries in the one or more sets, and an embedding of a natural language part of a plurality of templatized domain statements (Ghosh, para 0055 discloses finding similarity matching of query embedding with template statement or template questions “performing a first similarity matching process to match the embedded user question with a standard question template in a question template vector database”); and generating a SQL query for the at least one query in natural language, using the database schema, the database meta data, and the retrieved one or more templatized domain statements (Ghosh, para 0009 discloses matching a query template by similarity matching for SQL query generation “performing a first similarity matching process to match the embedded user question with a standard question template in a question template vector database; extracting parameters from the user question to populate the standard question template; populating the standard question template with the extracted parameters to generate a final question; and providing the final question and the standardized context information to a generative AI model, e.g., a transformer-based model that has been trained for SQL query generation, that converts the final question into a query code”; where Brende in para 0116 discloses receiving a natural language query having databases schema (tables) and metadata (date field for date range)).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of matching query template based on embedding similarity of Ghosh into natural language query conversion to SQL statement of Brende to produce an expected result of consistent SQL generation from natural language queries. The modification would be obvious because one of ordinary skill in the art would be motivated to present consistent output results by using similarity matching between embedded question and query template (Ghosh, para 0030).
But Brende and Ghosh do not explicitly teach computing a weighted set score for each of the one or more sets using the one or more sub-queries in the one or more sets and the similarity metric, wherein the weighted set score represents extent of similarity of the at least one query with each of the plurality of templatized domain statements; retrieving one or more of the plurality of templatized domain statements based on the weighted set score;
However, in the same field of endeavor of feature matching Rizk teaches computing a weighted set score for each of the one or more sets using the one or more sub-queries in the one or more sets and the similarity metric, wherein the weighted set score represents extent of similarity of the at least one query with each of the plurality of templatized domain statements (Rizk, para 0075 discloses query and template matching bases on weighed similarity, which can similarly be applied to sub-queries and statement template matching “If the input query requested a response from a recommended model, one implementation of module 350 uses the model-prompt template pair with the highest predicted or highest average predicted confidence value to generate a candidate response to the input query, and incorporates the candidate response into a response to the input query. …….. Another implementation of module 350 computes a weighted average of each candidate's similarity scores”);
retrieving one or more of the plurality of templatized domain statements based on the weighted set score (Rizk, para 0075 discloses response selection by highest weighted similarity score and which can similarly be applied for retrieving/selecting statement template “Module 350 incorporates the candidate response with the highest combined (e.g., the highest average) similarity score into a response to the input query”);
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of matching query template based on weighted similarity scores of Rizk into natural language query conversion to SQL statement of Brende and Ghosh to produce an expected result of consistent SQL generation from natural language queries. The modification would be obvious because one of ordinary skill in the art would be motivated to present relevant response by selecting templates based predicted confidence values(Rizk, abstract).
Regarding claim 12, Brende, Ghosh and Rizk teach all the limitation of claim 11 and Brende further teaches wherein the pre-processing comprises replacing a numerical value and a date to a predefined fixed integer (Brende, para 0110 discloses pre-processing the queries by outlining the basic structure of the query while leaving the place holder for dynamic values of the queries and further replacing the place holders “the prompt may comprise a predefined, partially complete portion that outlines the basic structure of the query, while the dynamic portion allows for the inclusion of specific details that are only available at the time of the query, such as the exact data fields to be retrieved…..The predefined portion of the prompt could look like this: “Retrieve total sales from Q_for the following products: “. Here, “Q_” represents a placeholder where the specific quarter needs to be specified, and the product list is left dynamic for user input. As the financial analyst interacts with the system, they specify that they are interested in “Q2” and want to see sales data for “Product A, Product B, and Product C.” The dynamic portion of the prompt is then completed in real-time, filling in the specified quarter and product names”).
Claim 3, 8 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Brende, Hans et al (PGPUB Document No. 20240394251), hereafter referred as to “Brende”, in view of Ghosh, Dipanjan et al (PGPUB Document No. 20260056945), hereafter, referred to as “Ghosh”, in view of Rizk, Yara et al (PGPUB Document No. 20250265489), hereafter, referred to as “Rizk”, in further view of Nair, Vinay et al (PGPUB Document No. 20220147515), hereafter, referred to as “Nair”.
Regarding claim 3, Brende, Ghosh and Rizk teach all the limitation of claim 1 Brende further teaches wherein segmenting the pre-processed at least one query into the one or more sets comprising the one or more sub-queries comprises(Brende, para 0180 discloses generation of sub-queries “the generator might split the query into smaller sub-queries that can be processed in parallel”):
But Brende, Ghosh and Rizk don’t explicitly teach creating, via the one or more hardware processors, a list of a plurality of individual words in the pre-processed at least one query; generating, via the one or more hardware processors, the one or more sets comprising a plurality of combinations of the one or more sub-queries in the one or more sets by iterating through the list of a plurality of individual words, and joining two or more of the plurality of individual words;
iteratively generating, via the one or more hardware processors, the one or more sub-queries until all of the plurality of individual words are used in at least one of the one or more sets, wherein each of the one or more sets comprises one or more sub-queries of same or different length of plurality of individual words, and wherein each of the one or more sub-queries among the one or more sets and each of the one or more set among the one or more sets are unique; and
generating, via the one or more hardware processors, the one or more sub-queries in the one or more sets until the one or more sub-queries are matching with the at least one query in natural language.
However, in the same field of endeavor of SQL query formation from natural language queries Nair teaches creating, via the one or more hardware processors, a list of a plurality of individual words in the pre-processed at least one query(Nair, para 0070 discloses dividing of queries into list of words “In accordance with one or more semantic parsing techniques known in the art, parsing module 222 divides the query into constituent parts, whereby different parts of speech are identified”); ;generating, via the one or more hardware processors, the one or more sets comprising a plurality of combinations of the one or more sub-queries in the one or more sets by iterating through the list of a plurality of individual words, and joining two or more of the plurality of individual words; iteratively generating, via the one or more hardware processors, the one or more sub-queries until all of the plurality of individual words are used in at least one of the one or more sets, wherein each of the one or more sets comprises one or more sub-queries of same or different length of plurality of individual words, and wherein each of the one or more sub-queries among the one or more sets and each of the one or more set among the one or more sets are unique(Nair, para 0120 discloses iterating/looping through a list/node of words for generating queries by joining or concatenating until the all words are considered “<Pseudo Code>
// Single SQL Generation
Initialize base SQL with placeholders for columns and tables
Set SQL = “Select s.id, s.ticker, s.name <column_place_holder> From security_master s
<table_place_holder>”
// For each node add column and table needed to SQL
FOR node in query: do
DB_Column = node // attribute/cost
Table = Lookup node ONTOLOGY node Factor (node) // attribute
Add DB_Column to <column_place_holder> in SQL
Add Table to <table_place_holder> in SQL
Add JOIN Clause for table with security_master An exemplary SQL”);
and generating, via the one or more hardware processors, the one or more sub-queries in the one or more sets until the one or more sub-queries are matching with the at least one query in natural language(Nair, further in para 0120-0121 discloses matching natural language query to generated SQL query “An exemplary SQL query generated by mapping module 223 that corresponds to the natural language input query “low cost emerging markets,” and which selects from the nodes “cost” and “emerging markets,” may be as follows:”).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of forming queries by iterating words of Nair into natural language query conversion to SQL statement of Brende, Ghosh and Rizk to produce an expected result of composing sub-queries form words. The modification would be obvious because one of ordinary skill in the art would be motivated to improve the matching of natural language query by implementing depiction of database table column names in a list or node (Nair, para 0093).
Regarding claim 8, Brende, Ghosh and Rizk teach all the limitation of claim 5 Brende further teaches wherein the one or more hardware processors are configured for segmenting the pre-processed at least one query into the one or more sets comprising the one or more sub-queries, comprises (Brende, para 0180 discloses generation of sub-queries “the generator might split the query into smaller sub-queries that can be processed in parallel”):
But Brende, Ghosh and Rizk don’t explicitly teach creating a list of a plurality of individual words in the pre- processed at least one query; generating the one or more sets comprising a plurality of combinations of the one or more sub-queries in the one or more sets by iterating through the list of a plurality of individual words, and joining two or more of the plurality of individual words;
iteratively generating the one or more sub-queries until all of the plurality of individual words are used in at least one of the one or more sets, wherein each of the one or more sets comprises one or more sub-queries of same or different length of plurality of individual words, and wherein each of the one or more sub- queries among the one or more sets, and each of the one or more set among the one or more sets are unique; and generating the one or more sub-queries in the one or more sets until the one or more sub-queries are matching with the at least one query in natural language.
However, in the same field of endeavor of SQL query formation from natural language queries Nair teaches creating a list of a plurality of individual words in the pre- processed at least one query (Nair, para 0070 discloses dividing of queries into list of words “In accordance with one or more semantic parsing techniques known in the art, parsing module 222 divides the query into constituent parts, whereby different parts of speech are identified”); generating the one or more sets comprising a plurality of combinations of the one or more sub-queries in the one or more sets by iterating through the list of a plurality of individual words, and joining two or more of the plurality of individual words;
iteratively generating the one or more sub-queries until all of the plurality of individual words are used in at least one of the one or more sets, wherein each of the one or more sets comprises one or more sub-queries of same or different length of plurality of individual words, and wherein each of the one or more sub- queries among the one or more sets, and each of the one or more set among the one or more sets are unique (Nair, para 0120 discloses iterating/looping through a list/node of words for generating queries by joining or concatenating “
<Pseudo Code>
// Single SQL Generation
Initialize base SQL with placeholders for columns and tables
Set SQL = “Select s.id, s.ticker, s.name <column_place_holder> From security_master s
<table_place_holder>”
// For each node add column and table needed to SQL
FOR node in query: do
DB_Column = node // attribute/cost
Table = Lookup node ONTOLOGY node Factor (node) // attribute
Add DB_Column to <column_place_holder> in SQL
Add Table to <table_place_holder> in SQL
Add JOIN Clause for table with security_master An exemplary SQL”);
and generating the one or more sub-queries in the one or more sets until the one or more sub-queries are matching with the at least one query in natural language (Nair, further in para 0120-0121 discloses matching natural language query to generated SQL query “An exemplary SQL query generated by mapping module 223 that corresponds to the natural language input query “low cost emerging markets,” and which selects from the nodes “cost” and “emerging markets,” may be as follows:”).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of forming queries by iterating words of Nair into natural language query conversion to SQL statement of Brende, Ghosh and Rizk to produce an expected result of composing sub-queries form words. The modification would be obvious because one of ordinary skill in the art would be motivated to improve the matching of natural language query by implementing depiction of database table column names in a list or node (Nair, para 0093).
Regarding claim 13, Brende, Ghosh and Rizk teach all the limitation of claim 11 Brende further teaches wherein segmenting the pre-processed at least one query into the one or more sets comprising the one or more sub- queries comprises (Brende, para 0180 discloses generation of sub-queries “the generator might split the query into smaller sub-queries that can be processed in parallel”):
But Brende, Ghosh and Rizk don’t explicitly teach creating a list of a plurality of individual words in the pre- processed at least one query; generating the one or more sets comprising a plurality of combinations of the one or more sub-queries in the one or more sets by iterating through the list of a plurality of individual words, and joining two or more of the plurality of individual words;
iteratively generating, the one or more sub-queries until all of the plurality of individual words are used in at least one of the one or more sets wherein each of the one or more sets comprises one or more sub-queries of same or different length of plurality of individual words and wherein each of the one or more sub- queries among the one or more sets and each of the one or more set among the one or more sets are unique; and generating the one or more sub-queries in the one or more sets until the one or more sub-queries are matching with the at least one query in natural language.
However, in the same field of endeavor of SQL query formation from natural language queries Nair teaches creating a list of a plurality of individual words in the pre- processed at least one query (Nair, para 0070 discloses dividing of queries into list of words “In accordance with one or more semantic parsing techniques known in the art, parsing module 222 divides the query into constituent parts, whereby different parts of speech are identified”); generating the one or more sets comprising a plurality of combinations of the one or more sub-queries in the one or more sets by iterating through the list of a plurality of individual words, and joining two or more of the plurality of individual words; iteratively generating, the one or more sub-queries until all of the plurality of individual words are used in at least one of the one or more sets wherein each of the one or more sets comprises one or more sub-queries of same or different length of plurality of individual words and wherein each of the one or more sub- queries among the one or more sets and each of the one or more set among the one or more sets are unique (Nair, para 0120 discloses iterating/looping through a list/node of words for generating queries by joining or concatenating “<Pseudo Code>
// Single SQL Generation
Initialize base SQL with placeholders for columns and tables
Set SQL = “Select s.id, s.ticker, s.name <column_place_holder> From security_master s
<table_place_holder>”
// For each node add column and table needed to SQL
FOR node in query: do
DB_Column = node // attribute/cost
Table = Lookup node ONTOLOGY node Factor (node) // attribute
Add DB_Column to <column_place_holder> in SQL
Add Table to <table_place_holder> in SQL
Add JOIN Clause for table with security_master An exemplary SQL”);
and generating the one or more sub-queries in the one or more sets until the one or more sub-queries are matching with the at least one query in natural language (Nair, further in para 0120-0121 discloses matching natural language query to generated SQL query “An exemplary SQL query generated by mapping module 223 that corresponds to the natural language input query “low cost emerging markets,” and which selects from the nodes “cost” and “emerging markets,” may be as follows:”).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of forming queries by iterating words of Nair into natural language query conversion to SQL statement of Brende, Ghosh and Rizk to produce an expected result of composing sub-queries form words. The modification would be obvious because one of ordinary skill in the art would be motivated to improve the matching of natural language query by implementing depiction of database table column names in a list or node (Nair, para 0093).
Claim 4, 9 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Brende, Hans et al (PGPUB Document No. 20240394251), hereafter referred as to “Brende”, in view of Ghosh, Dipanjan et al (PGPUB Document No. 20260056945), hereafter, referred to as “Ghosh”, in view of Rizk, Yara et al (PGPUB Document No. 20250265489), hereafter, referred to as “Rizk”, in further view of Yano, Taro et al (PGPUB Document No. 20240119079 ), hereafter, referred to as “Yano”.
Regarding claim 4, Brende, Ghosh and Rizk teach all the limitation of claim 1 but don’t explicitly teach wherein computing the weighted set score by using the length of sub-query representing number of words in the sub- query and the one or more similarity metric is represented as: Set_Score = ∑ (length of sub-string (number of words)) * (Max. similarity score with a templatized domain statement),
where, the length of sub-string is the number of words in the one or more sub-queries, and the Max. similarity score with a templatized domain statement is the similarity metric of the one or more sub-queries in the one or more sets.
However, in the same field of endeavor of feature matching Yano teaches wherein computing the weighted set score by using the length of sub-query representing number of words in the sub- query and the one or more similarity metric is represented as: Set_Score = ∑(length of sub-string (number of words)) * (Max. similarity score with a templatized domain statement),
where, the length of sub-string is the number of words in the one or more sub-queries, and the Max. similarity score with a templatized domain statement is the similarity metric of the one or more sub-queries in the one or more sets (Yano, para 0100 discloses considering weighted score sum for the length of the string and, this disclosure can similarly be applied to similarity determination between sub-queries to query statement templates “the matching score calculation means 183 may calculate the matching score by applying the class label of each class and the test data to a matcher which calculates the matching score using a sigmoid function that takes as an argument a weighted linear sum of at least one of similarities in either or both semantic similarities (for example, cosine similarity, Euclidean distance, etc.) and similarities (for example, the longest common substring length, an edit distance, etc.) of included character strings”);
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of matching using sum of similarity scores of Yano into natural language query conversion to SQL statement of Brende, Ghosh and Rizk to produce an expected result of matching between queries and templates. The modification would be obvious because one of ordinary skill in the art would be motivated to improve the accuracy of estimation calculation by using probability approach(Yano, para 0008).
Regarding claim 9, Brende, Ghosh and Rizk teach all the limitation of claim 6 but don’t explicitly teach wherein the one or more hardware processors are configured for computing the weighted set score by using the length of sub-query representing number of words in the sub-query and the one or more similarity metric is represented as: Set_Score = ∑ (length of sub-string (number of words)) * (Max. similarity score with a templatized domain statement), where, the length of sub-string is the number of words in the one or more sub-queries, and the Max. similarity score with a templatized domain statement is the similarity metric of the one or more sub-queries in one or more sets.
However, in the same field of endeavor of feature matching Yano teaches wherein the one or more hardware processors are configured for computing the weighted set score by using the length of sub-query representing number of words in the sub-query and the one or more similarity metric is represented as: Set_Score = ∑ (length of sub-string (number of words)) * (Max. similarity score with a templatized domain statement), where, the length of sub-string is the number of words in the one or more sub-queries, and the Max. similarity score with a templatized domain statement is the similarity metric of the one or more sub-queries in one or more sets(Yano, para 0100 discloses considering weighted score sum for the length of the string and, this disclosure can similarly be applied to similarity determination between sub-queries to query statement templates “the matching score calculation means 183 may calculate the matching score by applying the class label of each class and the test data to a matcher which calculates the matching score using a sigmoid function that takes as an argument a weighted linear sum of at least one of similarities in either or both semantic similarities (for example, cosine similarity, Euclidean distance, etc.) and similarities (for example, the longest common substring length, an edit distance, etc.) of included character strings”);
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of matching using sum of similarity scores of Yano into natural language query conversion to SQL statement of Brende, Ghosh and Rizk to produce an expected result of matching between queries and templates. The modification would be obvious because one of ordinary skill in the art would be motivated to improve the accuracy of estimation calculation by using probability approach(Yano, para 0008).
Regarding claim 14, Brende, Ghosh and Rizk teach all the limitation of claim 6 but don’t explicitly teach wherein computing the weighted set score by using the length of sub-query representing number of words in the sub- query and the one or more similarity metric is represented as: Set_Score = ∑ (length of sub-string (number of words)) * (Max. similarity score with a templatized domain statement),
where, the length of sub-string is the number of words in the one or more sub-queries, and the Max. similarity score with a templatized domain statement is the similarity metric of the one or more sub-queries in the one or more sets.
However, in the same field of endeavor of feature matching Yano teaches wherein computing the weighted set score by using the length of sub-query representing number of words in the sub- query and the one or more similarity metric is represented as: Set_Score = ∑ (length of sub-string (number of words)) * (Max. similarity score with a templatized domain statement), where, the length of sub-string is the number of words in the one or more sub-queries, and the Max. similarity score with a templatized domain statement is the similarity metric of the one or more sub-queries in the one or more sets (Yano, para 0100 discloses considering weighted score sum for the length of the string and, this disclosure can similarly be applied to similarity determination between sub-queries to query statement templates “the matching score calculation means 183 may calculate the matching score by applying the class label of each class and the test data to a matcher which calculates the matching score using a sigmoid function that takes as an argument a weighted linear sum of at least one of similarities in either or both semantic similarities (for example, cosine similarity, Euclidean distance, etc.) and similarities (for example, the longest common substring length, an edit distance, etc.) of included character strings”);
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of matching using sum of similarity scores of Yano into natural language query conversion to SQL statement of Brende, Ghosh and Rizk to produce an expected result of matching between queries and templates. The modification would be obvious because one of ordinary skill in the art would be motivated to improve the accuracy of estimation calculation by using probability approach(Yano, para 0008).
Claim 5, 10 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Brende, Hans et al (PGPUB Document No. 20240394251), hereafter referred as to “Brende”, in view of Ghosh, Dipanjan et al (PGPUB Document No. 20260056945), hereafter, referred to as “Ghosh”, in view of Rizk, Yara et al (PGPUB Document No. 20250265489), hereafter, referred to as “Rizk”, in further view of Balasubramaniyan, Rajarajeswari et al (PGPUB Document No. 20250316105), hereafter, referred to as “Balasubramaniyan”.
Regarding claim 5, Brende, Ghosh and Rizk teach all the limitation of claim 1 Brende further teaches wherein generating the plurality of templatized domain statements comprises: receiving, via the one or more hardware processors, a) one or more domain statements for a given domain, in natural language, and b) a domain specific database schema comprising name of a plurality of tables, columns, data types, and representative strings corresponding to one or more entities stored in each column(Brende, para 0116 discloses receiving a natural language query having databases schema (tables) and metadata (date field for date range) “abstraction language interface 213 obtains a natural language query from a user, which could be a simple question or a complex instruction pertaining to data retrieval or manipulation….abstraction language interface 213 determines the main data subjects by linking nouns and noun phrases to corresponding entities in the database schema, such as mapping “sales” to a “Sales” database table….”);
Ghosh teaches selecting, via the one or more hardware processors, one or more few shot exemplars from a training dataset comprising a natural language query and a plurality of associated SQL queries(Ghosh, in para 0009 discloses providing a particular instance (exemplar) to the learning model “performing a first similarity matching process to match the embedded user question with a standard question template in a question template vector database; extracting parameters from the user question to populate the standard question template; populating the standard question template with the extracted parameters to generate a final question; and providing the final question and the standardized context information to a generative AI model, e.g., a transformer-based model that has been trained for SQL query generation, that converts the final question into a query code”);
generating, via the one or more hardware processors, the natural language part of each of the plurality of templatized domain statement by applying a trained model on the one or more few shot exemplars, the one or more domain statements, and the domain specific database schema(Ghosh, further in para 0010 discloses generation templatized statement using trained model “The embedding model may be a pre-trained embedding model that is configured to transform natural language into a vector representation. At least one of the first similarity matching process or the second similarity matching process comprises using a cosine similarity to find a closest standard question template in the question template vector database”);
generating, via the one or more hardware processors, a SQL logic for the natural language part of each of the plurality of the templatized domain statements by applying the trained model on the one or more few shot exemplars, the one or more domain statements, the database schema, and the generated natural language part of each of the plurality of templatized domain statements(Ghosh, para 0055 discloses finding similarity matching of query embedding with template statement or template questions “performing a first similarity matching process to match the embedded user question with a standard question template in a question template vector database; extracting parameters from the user question to populate the standard question template; populating the standard question template with the extracted parameters to generate a final question; and providing the final question and the standardized context information to a generative AI model that converts the final question into a query code”);
based on the one or more few shot exemplar; and computing, via the one or more hardware processors, the embedding of the natural language part of the plurality of templatized domain statements(Ghosh, further in para 0009-0010 disclose embedding of user query para 0009 discloses embedding for natural language queries/sub-queries “in response to receiving a user question, e.g., in natural language format, using an embedding model to obtain an embedded user question”), wherein the computed embedding is stored in a database(Ghosh, in para 0009 discloses embeddings ae stored in database “performing a first similarity matching process to match the embedded user question with a standard question template in a question template vector database).
But Brende, Ghosh and Rizk don’t explicitly teach validating, via the one or more hardware processors, the natural language part of each of the plurality of templatized domain statements and associated SQL logic for consistency using the trained model, wherein, if the validation is inconsistent, the SQL logic associated with the natural language part of each of the plurality of templatized domain statements is updated using the trained model, and wherein if validation is consistent, the templatized domain statement along with the SQL logic are generated by using the trained model for combining the natural language part of each of the plurality of templatized domain statements and the SQL logic,
However, in the same field of endeavor of model training of Balasubramaniyan teaches validating, via the one or more hardware processors, the natural language part of each of the plurality of templatized domain statements and associated SQL logic for consistency using the trained model(Balasubramaniyan, para 0132 discloses model training validation with any dataset “The model trainer 160 validates the trained models (e.g., the trained machine learning models 135 or the trained attention embedded transformer network models 140) using a test data set”),
wherein, if the validation is inconsistent, the SQL logic associated with the natural language part of each of the plurality of templatized domain statements is updated using the trained model, and wherein if validation is consistent, the templatized domain statement along with the SQL logic are generated by using the trained model for combining the natural language part of each of the plurality of templatized domain statements and the SQL logic(Balasubramaniyan, para 0135 discloses model training validation and update or re-train the model if validation fails “The model trainer 160 checks the retrained models (e.g., the retrained machine learning models 135 or the retrained attention embedded transformer network models 140) for validity. The model trainer 160 checks or tests the retrained models as described herein, by comparing an error score of each model with a threshold error for each model. Upon the model trainer 160 determining that one or more of the retrained models are invalid”),
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of validating training models of Balasubramaniyan into natural language query conversion to SQL statement of Brende, Ghosh and Rizk to produce an expected result of matching between queries and templates. The modification would be obvious because one of ordinary skill in the art would be motivated to improve the accuracy of the training model by retraining the model to a threshold of accuracy(Balasubramaniyan, para 0132).
Regarding claim 10, Brende, Ghosh and Rizk teach all the limitation of claim 6 Brende further teaches wherein the one or more hardware processors are configured for generating the plurality of templatized domain statements, comprises: receiving, a) one or more domain statements for a given domain, in natural language, and b) a domain specific database schema comprising a name of a plurality of tables, columns, data types, and representative strings corresponding to one or more entities stored in each column(Brende, para 0116 discloses receiving a natural language query having databases schema (tables) and metadata (date field for date range) “abstraction language interface 213 obtains a natural language query from a user, which could be a simple question or a complex instruction pertaining to data retrieval or manipulation….abstraction language interface 213 determines the main data subjects by linking nouns and noun phrases to corresponding entities in the database schema, such as mapping “sales” to a “Sales” database table….”);
Ghosh teaches selecting one or more few shot exemplars from a training dataset comprising a natural language query and a plurality of associated SQL queries (Ghosh, in para 0009 discloses providing a particular instance (exemplar) to the learning model “performing a first similarity matching process to match the embedded user question with a standard question template in a question template vector database; extracting parameters from the user question to populate the standard question template; populating the standard question template with the extracted parameters to generate a final question; and providing the final question and the standardized context information to a generative AI model, e.g., a transformer-based model that has been trained for SQL query generation, that converts the final question into a query code”);
generating the natural language part of each of the plurality of templatized domain statement by applying a trained model on the one or more few shot exemplars, the one or more domain statements, and the domain specific database schema (Ghosh, further in para 0010 discloses generation templatized statement using trained model “The embedding model may be a pre-trained embedding model that is configured to transform natural language into a vector representation. At least one of the first similarity matching process or the second similarity matching process comprises using a cosine similarity to find a closest standard question template in the question template vector database”);
generating a SQL logic for the natural language part of each of the plurality of the templatized domain statements by applying the trained model on the one or more few shot exemplars, the one or more domain statements, the database schema, and the generated natural language part of each of the plurality of templatized domain statements (Ghosh, para 0055 discloses finding similarity matching of query embedding with template statement or template questions “performing a first similarity matching process to match the embedded user question with a standard question template in a question template vector database; extracting parameters from the user question to populate the standard question template; populating the standard question template with the extracted parameters to generate a final question; and providing the final question and the standardized context information to a generative AI model that converts the final question into a query code”);
based on the one or more few shot exemplar; and computing the embedding of the natural language part of the plurality of templatized domain statements (Ghosh, further in para 0009-0010 disclose embedding of user query para 0009 discloses embedding for natural language queries/sub-queries “in response to receiving a user question, e.g., in natural language format, using an embedding model to obtain an embedded user question”), wherein the computed embedding is stored in a database (Ghosh, in para 0009 discloses embeddings ae stored in database “performing a first similarity matching process to match the embedded user question with a standard question template in a question template vector database).
But Brende, Ghosh and Rizk don’t explicitly teach validating the natural language part of each of the plurality of templatized domain statements and associated SQL logic for consistency using the trained model, wherein,
if validation is inconsistent, the SQL logic associated with the natural language part of each of the plurality of templatized domain statements is updated using the trained model, wherein if validation is consistent, the templatized domain statement along with the SQL logic are generated by using the trained model for combining the natural language part of each of the plurality of templatized domain statements and the SQL logic,
However, in the same field of endeavor of model training of Balasubramaniyan teaches validating the natural language part of each of the plurality of templatized domain statements and associated SQL logic for consistency using the trained model (Balasubramaniyan, para 0132 discloses model training validation with any dataset “The model trainer 160 validates the trained models (e.g., the trained machine learning models 135 or the trained attention embedded transformer network models 140) using a test data set”),
wherein, if validation is inconsistent, the SQL logic associated with the natural language part of each of the plurality of templatized domain statements is updated using the trained model, wherein if validation is consistent, the templatized domain statement along with the SQL logic are generated by using the trained model for combining the natural language part of each of the plurality of templatized domain statements and the SQL logic (Balasubramaniyan, para 0135 discloses model training validation and update or re-train the model if validation fails “The model trainer 160 checks the retrained models (e.g., the retrained machine learning models 135 or the retrained attention embedded transformer network models 140) for validity. The model trainer 160 checks or tests the retrained models as described herein, by comparing an error score of each model with a threshold error for each model. Upon the model trainer 160 determining that one or more of the retrained models are invalid”),
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of validating training models of Balasubramaniyan into natural language query conversion to SQL statement of Brende, Ghosh and Rizk to produce an expected result of matching between queries and templates. The modification would be obvious because one of ordinary skill in the art would be motivated to improve the accuracy of the training model by retraining the model to a threshold of accuracy(Balasubramaniyan, para 0132).
Regarding claim 15, Brende, Ghosh and Rizk teach all the limitation of claim 11 Brende further teaches wherein generating the plurality of templatized domain statements comprises: receiving a) one or more domain statements for a given domain, in natural language, and b) a domain specific database schema comprising name of a plurality of tables, columns, data types, and representative strings corresponding to one or more entities stored in each column(Brende, para 0116 discloses receiving a natural language query having databases schema (tables) and metadata (date field for date range) “abstraction language interface 213 obtains a natural language query from a user, which could be a simple question or a complex instruction pertaining to data retrieval or manipulation….abstraction language interface 213 determines the main data subjects by linking nouns and noun phrases to corresponding entities in the database schema, such as mapping “sales” to a “Sales” database table….”);
Ghosh teaches selecting one or more few shot exemplars from a training dataset comprising a natural language query and a plurality of associated SQL queries (Ghosh, in para 0009 discloses providing a particular instance (exemplar) to the learning model “performing a first similarity matching process to match the embedded user question with a standard question template in a question template vector database; extracting parameters from the user question to populate the standard question template; populating the standard question template with the extracted parameters to generate a final question; and providing the final question and the standardized context information to a generative AI model, e.g., a transformer-based model that has been trained for SQL query generation, that converts the final question into a query code”);
generating the natural language part of each of the plurality of templatized domain statement by applying a trained model on the one or more few shot exemplars, the one or more domain statements, and the domain specific database schema (Ghosh, further in para 0010 discloses generation templatized statement using trained model “The embedding model may be a pre-trained embedding model that is configured to transform natural language into a vector representation. At least one of the first similarity matching process or the second similarity matching process comprises using a cosine similarity to find a closest standard question template in the question template vector database”);
generating a SQL logic for the natural language part of each of the plurality of the templatized domain statements by applying the trained model on the one or more few shot exemplars, the one or more domain statements, the database schema, and the generated natural language part of each of the plurality of templatized domain statements (Ghosh, para 0055 discloses finding similarity matching of query embedding with template statement or template questions “performing a first similarity matching process to match the embedded user question with a standard question template in a question template vector database; extracting parameters from the user question to populate the standard question template; populating the standard question template with the extracted parameters to generate a final question; and providing the final question and the standardized context information to a generative AI model that converts the final question into a query code”);
based on the one or more few shot exemplar; and computing the embedding of the natural language part of the plurality of templatized domain statements (Ghosh, further in para 0009-0010 disclose embedding of user query para 0009 discloses embedding for natural language queries/sub-queries “in response to receiving a user question, e.g., in natural language format, using an embedding model to obtain an embedded user question”), wherein the computed embedding is stored in a database(Ghosh, in para 0009 discloses embeddings ae stored in database “performing a first similarity matching process to match the embedded user question with a standard question template in a question template vector database).
But Brende, Ghosh and Rizk don’t explicitly teach validating the natural language part of each of the plurality of templatized domain statements and associated SQL logic for consistency using the trained model, wherein, if the validation is inconsistent, the SQL logic associated with the natural language part of each of the plurality of templatized domain statements is updated using the trained model, and where in if validation is consistent, the templatized domain statement along with the SQL logic are generated by using the trained model for combining the natural language part of each of the plurality of templatized domain statements and the SQL logic,
However, in the same field of endeavor of model training of Balasubramaniyan teaches validating the natural language part of each of the plurality of templatized domain statements and associated SQL logic for consistency using the trained model (Balasubramaniyan, para 0132 discloses model training validation with any dataset “The model trainer 160 validates the trained models (e.g., the trained machine learning models 135 or the trained attention embedded transformer network models 140) using a test data set”),
wherein, if the validation is inconsistent, the SQL logic associated with the natural language part of each of the plurality of templatized domain statements is updated using the trained model, and where in if validation is consistent, the templatized domain statement along with the SQL logic are generated by using the trained model for combining the natural language part of each of the plurality of templatized domain statements and the SQL logic (Balasubramaniyan, para 0135 discloses model training validation and update or re-train the model if validation fails “The model trainer 160 checks the retrained models (e.g., the retrained machine learning models 135 or the retrained attention embedded transformer network models 140) for validity. The model trainer 160 checks or tests the retrained models as described herein, by comparing an error score of each model with a threshold error for each model. Upon the model trainer 160 determining that one or more of the retrained models are invalid”),
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature of validating training models of Balasubramaniyan into natural language query conversion to SQL statement of Brende, Ghosh and Rizk to produce an expected result of matching between queries and templates. The modification would be obvious because one of ordinary skill in the art would be motivated to improve the accuracy of the training model by retraining the model to a threshold of accuracy(Balasubramaniyan, para 0132).
Conclusion
Listed below are the prior arts made of record and not relied upon but are considered pertinent to applicant’s disclosure.
Arjun (US 2025011095) -text to SQL conversion and query preprocessing..
Govinda (US 20220405281) -teaches segmentation of query into sub-queries.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ABDULLAH A DAUD whose telephone number is (469)295-9283. The examiner can normally be reached M~F: 9:30 am~6:30 pm.
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, Amy Ng can be reached at 571-270-1698. 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.
/ABDULLAH A DAUD/Examiner, Art Unit 2164
/AMY NG/Supervisory Patent Examiner, Art Unit 2164
Prosecution Insights