Notice of Pre-AIA or AIA Status
1. 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
2. This Office Action is in response to the filing with the office dated 05/18/2026.
Claims 1-4, 6-11, 13 and 14 have been amended. Claims 5 and 15 have been cancelled. Claims 1 and 11 are independent claims. Claims 1-4, 6-14, 16-20 are presented in this office action.
Response to amendment/arguments
3. Applicant’s arguments with respect to the rejection of claims under 35 U.S.C. § 101 as the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more, have been fully considered. However, Examiner respectfully disagrees with the applicant’s argument. See response to arguments section. The rejection has been maintained
Applicant’s arguments with respect to the rejection of claims under 35 U.S.C. § 102 (a)(i) and 103(a) have been fully considered and are not persuasive, thus necessitated the rejection as presented in this Office action. Please see the response to arguments below. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
Response to 103 Arguments
4. Applicant’s arguments on page 14 regarding claim 1 states “Applicant respectfully submits that Rafidi does not teach, suggest or motivate at least the limitations of “wherein the LLM is isolated from the one or more entries of a medical database of interest, wherein the at least a processor is interfaceable with the medical database; receive from the LLM a feature set comprising a requirement set of hyper-logical nodes, wherein the feature set comprises at least a combination feature and wherein the hyper-logical nodes are synthesized by: extracting a plurality of atomic elements from the first natural language database query; and merging the plurality of atomic elements using logical operators and temporal relations identified by the LLM within the first natural language database query; utilize a medical database query map as an intermediary to map a nodal output of the isolated LLM to one or more correct elements within a medical database” as recited in part by amended claim 1”.
Examiner respectfully disagrees with the applicant, and maintains the rejection as Rafidi et al teaches the limitations of “wherein the LLM is isolated from the one or more entries of a medical database of interest, wherein the at least a processor is interfaceable with the medical database” (Paragraph [0048] discloses, LLM receiving a natural language query which is not directly connected to from medical database);
“receive from the LLM a feature set comprising a requirement set of hyper-logical nodes, wherein the feature set comprises at least a combination feature and wherein the hyper-logical nodes are synthesized by: extracting a plurality of atomic elements from the first natural language database query; and merging the plurality of atomic elements using logical operators and temporal relations identified by the LLM within the first natural language database query” (Paragraph [0056], [0060] discloses, extracting combination feature sets such as metrics and parameters that are in a pipeline from natural language query using logical operators and temporal relations and merging/ aggregating them based on the query (Examiner interprets atomic elements as elements in the pipeline). (Based on specification Paragraph [0072] hyper-logical nodes are created by merging feature set with a logical operator and temporal relations if any).
Rafidi et al fails to explicitly teach, “utilize a medical database query map as an intermediary to map a nodal output of the isolated LLM to one or more correct elements within a medical database”.
However, Belcher et al teaches the limitation “utilize a medical database query map as an intermediary to map a nodal output of the isolated LLM to one or more correct elements within a medical database” (Fig. 1 Paragraph [0100]-[0103] discloses, generating a query based on feature set, by filtering different properties based on lexical constraints containing relevant values as an intermediary step to retrieve only a dataset exactly corresponding to data indicated by an RLQL query/ output correct elements within the medical database. Also see [0094]-[0095]).
Therefore, Rafidi et al in view of Belcher et al teach the argued limitations.
Please see the rejection below.
Response to 101 Arguments
5. Applicant’s arguments regarding 101 rejection on pages 3 and 4, 5 recite “Applicant respectfully submits that Claim 1, as amended, does not recite a mental process because, at least, the synthesis of a “requirement set of hyper-logical nodes” and the automated translation of those nodes into a “structure executable by the medical database” involve complex data manipulations for which the human mind is not equipped. Specifically, the claim requires the LLM to extract a plurality of atomic elements and merge them using logical operators and temporal relations. This process is not a mere mental observation, but a programmatic construction of interdependent data structures. The human mind is not equipped to perform structured nodal synthesis and generate machine-readable queries. In addition, claim 1 is rooted in a specific computer architecture where the LLM is strictly isolated from the one or more entries of the medical database. Because the LLM application does not know where entries are in a database, the process requires a medical database query map to act as a technical intermediary to map the isolated nodal output to the correct elements within a medical database. The human mind cannot practically replicate this architectural isolation. Specifically, the claimed invention requires the use of a computer-related architecture to bridge two disconnected domains. The manipulation of combination features into a template to generate a specific, formatted query string is an inherently computer-related process that is fundamental to database architecture, not human thought. Because these limitations represent a technical solution to a computer related process (specifically one relating to privacy and database entries), they could not conceivably be performed in the human mind or with pencil and paper. Accordingly, Applicant submits claim 1 is patent-eligible under Step 2A Prong One of the Patent Eligibility Test under 35 U.S.C. § 101”. Applicant’s arguments on page 5 recites “Applicant respectfully submits that claim 1, as amended, integrates any alleged abstract idea into a practical application and is therefore not directed to an abstract idea under Step 2A, Prong Two. The claimed invention provides a technical transformation of the computer’s capability by enabling an isolated LLM to generate valid queries for a database that it strictly does not have access to, a process that generic generative models are not equipped to perform. This improvement is achieved through a specific technical transformation of the LLM’s role: the synthesis of a “requirement set of hyper-logical nodes” that serves as a bridge to a secure, blind environment. This requires the LLM to identify “atomic elements” and merge them using “logical operators and temporal relations” identified within the natural language query. Because the LLM remains strictly isolated from the one or more entries of the medical database and lacks knowledge of the database’s internal structure, the claim recites a medical database query map to function as a specialized technical intermediary. This intermediary specifically maps the nodal output of the isolated LLM to one or more correct elements within a medical database to generate a first medical database query. This process transforms the high-level nodal output into a structure executable by the medical database formatted according to specific syntax rules. This specific implementation provides a technical solution for translating natural language into complex machine code without requiring a generative model to have direct access to, or knowledge of, private data. Such a configuration is not a generic computer use but a specific architectural transformation that enables a new computer capability. That is, the generation of precise, executable queries for isolated data repositories. Similar to Enfish, The claim was not simply the addition of general-purpose computers added post-hoc to an abstract idea, but a specific implementation of a solution to a problem in the software arts. Accordingly, Applicant submits claim 1 integrates the purported abstract idea into a practical application and is patent-eligible under 35 U.S.C. § 101.
Examiner respectfully disagrees as the amended limitations “requirement set of hyper-logical nodes” is an additional element which is insignificant extra-solution activity of a data gathering process. Similarly using an LLM to translate those nodes into a “structure executable by the medical database” involves manipulation of queries is an abstract idea and using an LLM to manipulate the queries using a model to convert natural language to SQL queries amounts to nothing more than mere instructions to apply the recited abstract idea on a computer, under MPEP 2106.05(f). Combination of these additional elements is no more than mere instructions to apply the exception using series of steps and outputting the result of the mental process. Accordingly, even in combination, the additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The amended limitations “identify “atomic elements” and merge them using “logical operators and temporal relations” identified within the natural language query” involves identifying the features in a natural language and merging them based on the logical operators that under broadest reasonable interpretation, covers performance of the limitation in the mind. There is nothing in the claim element that precludes the steps from practically being performed by a human mentally or with pen and paper. These limitations, at the high level of generality as drafted, would encompass a user can look at a natural language query and identify features and logical operators. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
This judicial exception is not integrated into a practical application. In particular, the claim recites the additional element of, using a model is recited at a high level of generality as generic computer components. These additional elements amount to nothing more than mere instructions to apply the recited abstract idea on a computer, under MPEP 2106.05(f). The additional element of accessing the data from database amount to mere data outputting which is insignificant extra-solution activity. Combination of these additional elements is no more than mere instructions to apply the exception using series of steps and outputting the result of the mental process. Accordingly, even in combination, the additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea.
The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the recitation of generic computing components is still mere instructions to apply the exception under MPEP 2106.05(f) and does not provide significantly more. The “accessing the data from database” element that was identified as insignificant extra-solution activity as mere data outputting when re-evaluated still does not provide significantly more, since this generic data outputting on a user interface is well, understood, routine and conventional (WURC). Considering the additional elements in combination and the claim as a whole does not change the analysis and does not amount to significantly more. Thus, the claims are abstract.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whether he 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 therefore subject to the conditions and requirements of this title.
6. Determining whether claims are statutory under 35 U.S.C. 101 involves a two-step analysis. Step 1 requires a determination of whether the claims are directed to the statutory categories of invention. Step 2 requires a determination of whether the claims are directed to a judicial exception without significantly more. Step 2 is divided into two prongs, with the first prong having a part 1 and part 2. See MPEP 2106; See 2019 Revised Patent Subject Matter Eligibility Guidance (2019 PEG).
Step 1, Claims 1-11 recite an apparatus which are directed to the statutory category of a machine.
Regarding independent claims 1 and 11
Step 2A, Part 1, claims are analyzed to determine whether they are directed to an abstract idea. Under the 2019 PEG, claims are deemed to be directed to an abstract idea if they fall within one of the enumerated categories of (a) mathematical concepts, (b) certain methods of organizing human activity, and (c) mental processes. Here, claims 1 and 11 are directed to an abstract idea categorized under mental processes. Courts consider a mental process if it “can be performed in the human mind, or by a human using a pen and paper.” MPEP 2016(a)(2)(III). Courts also consider a mental process as one that can be performed in the human mind and is merely using a computer as a tool to perform the concept. MPEP 2016(a)(2)(III)(C)(3). Claims 1 and 11 recites a mental process because the recited steps recite “receiving a natural language query, wherein the LLM is isolated from one or more entries of a medical database of interest, wherein the at least a processor is interfaceable with the medical database”, “receive feature set… the hyper-logical nodes are synthesized by: extracting a plurality of atomic elements from the first natural language database query”, “merging….”, “generate a query to access a database using a database query map, and output the result”, “mapping….”. The processor and memory are recited at a high level of generality and do not place meaningful limits on the abstract idea. input the query into a model, receive feature set, generate a query to access a database, is a task that can be performed by a human with the use of the computer as a tool. These limitations are essentially steps of generating and manipulating data at a high level of generality, which can be performed by a person using a computer as a tool.
Step 2A, part 2, claims are analyzed to determine whether the recited abstract idea is integrated into a practical application. In this case, as explained above, claims 1, 11 merely recite a mental process. These limitations describe receiving a natural language query, inputting the query into a model, receive feature set, extracting plurality of elements, merging the extracted elements based on temporal relations, generate a query to access a database using a database query map, and output the result. While claims 1, 11 recites additional components in the form of processors, a memory, and a storage, these components are recited at a high level of generality, which do not add meaningful limits on the recited abstract idea to integrate it into a practical application by providing an improvement to the functioning of a computer or technology, implementing the abstract idea with a particular machine or manufacture that is integral to the claim, effecting a transformation or reduction of a particular article to a different state or thing, nor applying the abstract idea in some meaningful way beyond linking its use to computer technology. See 2019 PEG. The additional limitations receiving a natural language query, input the query into a model and output the result are insufficient to integrate the abstract idea into a practical application. receiving a natural language query, inputting the query into a model are insignificant extra-solution activity of a data gathering process. receive feature set, generate a query to access a database using a database query map is a mental process and output the results is post-solution insignificant activity as mere data outputting.
Since claims 1 and 11 are directed to an abstract idea categorized as a mental process and do not integrate the judicial exception into a practical application, claims 1 and 11 are directed to a judicial exception.
Step 2B, Claims are analyzed to determine whether they recite significantly more than the abstract idea. In other words, it is determined whether the claims provide an inventive concept. In this case, claims 1 and 11 does not recite limitations that amount to significantly more than the abstract idea. The additional elements of utilizing a medical database to map correct element from the query is mere data gathering and insignificant extra-solution activity of a data gathering process. These limitations are steps involving processes that can be practically performed by a human with the aid of pen and paper, or as explained above, using a computer as a tool to perform the concept. These limitations, at the high level of generality as drafted, would encompass a user to receive a query and analyze the query and uncover patterns, structures or relationships within the data without predefined outputs, so that input the query maps the extracted features to a medical concept and output the data is mentally performable as an evaluation or judgement. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the “Mental Processes” grouping of abstract ideas. Accordingly, the claim recites an abstract idea. These limitations do not improve the functioning of a computer, improve the technology, apply the abstract idea to a particular machine, effect a transformation, nor provide meaningful limitations beyond linking the abstract idea to computer technology. They do not recite specific details that amount to significantly more than the abstract idea or provide meaningful limits on the abstract idea. For at least these reasons, claim 1 is nonstatutory because they are directed to a judicial exception without significantly more.
Regarding dependent claims 2 and 12 depend on independent claims 1 and 11 and therefore recites the same abstract idea. claims 2 and 12 recites the additional limitations output the aggregated output to a user as a function of a medical database response subject count. This limitation seems to recite the feature of counting the number of results in the output. See specification Paragraph [0054]. These limitations do not integrate the abstract idea into a practical application because the additional limitation merely describes removing indices for data blocks that are no longer needed and will be deleted. Therefore, these additional limitations do not integrate the abstract idea into a practical application. Pursuant to step 2B, the additional limitations do not amount to significantly more than the abstract idea because the limitations are not recited in a manner that provides improvements to the functioning of a computer or any other technology or technical field.
Regarding dependent claims 3 and 13 depend on independent claims 1 and 11 and therefore recites the same abstract idea. claims 3 and 13 recites the additional limitations receive the first natural language database query as a function of a user input of a user. This limitation seems to recite the feature of receiving a natural language query. These limitations do not integrate the abstract idea into a practical application because the additional limitation merely describes removing indices for data blocks that are no longer needed and will be deleted. Therefore, these additional limitations do not integrate the abstract idea into a practical application. Pursuant to step 2B, the additional limitations do not amount to significantly more than the abstract idea because the limitations are not recited in a manner that provides improvements to the functioning of a computer or any other technology or technical field.
Regarding dependent claims 4 and 14 depend on independent claims 1 and 11 and therefore recites the same abstract idea. claims 4 and 14 recites the additional limitations creating a condensed feature set containing at least one combination feature as a function of the feature set. This limitation seems to recite the feature of having certain constraints on the desired data. See specification Paragraph [0050]. These limitations do not integrate the abstract idea into a practical application because the additional limitation merely describes removing indices for data blocks that are no longer needed and will be deleted. Therefore, these additional limitations do not integrate the abstract idea into a practical application. Pursuant to step 2B, the additional limitations do not amount to significantly more than the abstract idea because the limitations are not recited in a manner that provides improvements to the functioning of a computer or any other technology or technical field.
Regarding dependent claims 6 and 16 depend on independent claims 1 and 11 and therefore recites the same abstract idea. claims 6 and 16 recites the additional limitations output the feature set to a user. This limitation seems to recite the feature of presenting/ outputting the data to the user. This limitation seems to recite the feature of having certain constraints on the desired data. See specification Paragraph [0050]. These limitations do not integrate the abstract idea into a practical application because the additional limitation merely describes removing indices for data blocks that are no longer needed and will be deleted. Therefore, these additional limitations do not integrate the abstract idea into a practical application. Pursuant to step 2B, the additional limitations do not amount to significantly more than the abstract idea because the limitations are not recited in a manner that provides improvements to the functioning of a computer or any other technology or technical field.
Regarding dependent claims 7 and 17 depend on independent claims 1 and 11 and therefore recites the same abstract idea. claims 7 and 17 recites the additional limitations receive a second natural language database query; and modify the feature set as a function of the second natural language database query. This limitation seems to recite the feature of receiving another query from the user and modifying the first query based on the second query is a mental process. These limitations do not integrate the abstract idea into a practical application because the additional limitation merely describes removing indices for data blocks that are no longer needed and will be deleted. Therefore, these additional limitations do not integrate the abstract idea into a practical application. Pursuant to step 2B, the additional limitations do not amount to significantly more than the abstract idea because the limitations are not recited in a manner that provides improvements to the functioning of a computer or any other technology or technical field.
Regarding dependent claims 8 and 18 depend on independent claims 1 and 11 and therefore recites the same abstract idea. claims 8 and 18 recite the additional limitations train the LLM on a training dataset including a plurality of example natural language database queries as inputs correlated to a plurality of example feature sets as outputs. This limitation seems to recite the feature of training the dataset using a computer as a tool. These limitations do not integrate the abstract idea into a practical application because the additional limitation merely describes removing indices for data blocks that are no longer needed and will be deleted. Therefore, these additional limitations do not integrate the abstract idea into a practical application. Pursuant to step 2B, the additional limitations do not amount to significantly more than the abstract idea because the limitations are not recited in a manner that provides improvements to the functioning of a computer or any other technology or technical field.
Regarding dependent claims 9 and 19 depend on independent claims 1 and 11 and therefore recites the same abstract idea. claims 8 and 18 recite the additional limitations generate a second medical database query as a function of the feature set. This limitation seems to recite the feature of generating a query which is a post-solution insignificant activity. These limitations do not integrate the abstract idea into a practical application because the additional limitation merely describes removing indices for data blocks that are no longer needed and will be deleted. Therefore, these additional limitations do not integrate the abstract idea into a practical application. Pursuant to step 2B, the additional limitations do not amount to significantly more than the abstract idea because the limitations are not recited in a manner that provides improvements to the functioning of a computer or any other technology or technical field.
Regarding dependent claims 10 and 20 depend on independent claims 1 and 11 and therefore recite the same abstract idea. claims 8 and 18 recite the additional limitations generate the aggregated output as a function of a first medical database response responsive to the first medical database query and a second medical database response responsive to the second medical database query. This limitation seems to recite the feature of aggregated output which is a post-solution insignificant activity.
Claim Rejections - 35 U.S.C. § 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 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.
7. Claims 1-4, 6-14, 16-20 are rejected under 35 U.S.C. 103 as being unpatentable over Rafidi; Joseph (US 20240045863 A1) in view of Belcher; Thomas (US 20180137177 A1) and in further view of SHAHRIAR; Muneem ( US 20230161763 A1).
Regarding independent claim 1, Rafidi; Joseph (US 20240045863 A1) teaches, an apparatus for generating a medical database query (Paragraph [0047] In some embodiments, the process 120 includes generating a model query based on the NL query. Also see Paragraph [0055]), the apparatus comprising: at least a processor; and a memory communicatively connected to the at least a processor, wherein the memory contains instructions configuring the at least a processor (Fig. 4 elements 612, 604, Paragraph [0061]) to: receive a first natural language database query (Paragraph [0025] In certain embodiments, the process 110 includes receiving an NL query, one or more input datasets (e.g., including one or more tables), and optionally one or more target datasets (e.g., including one or more tables). The NL query may be a query indicating some desired information, or one or more desired datasets. The NL query may include one or more strings. The NL query may include language that indicates certain constraints on the desired data (e.g., may include language specifying a date range, or an age range));
input the first natural language database query into a large language model (LLM) (Paragraphs [0047], [0048] The model may be an NL processing model, such as a machine-learning NL processing model. For example, the model may be an autoregressive language model, such as a Generative Pre-trained Transformer 3 (GPT-3) model (i.e., Examiner interprets large language model (LLM) as machine learning models and GPT-3 is one of the model). Implementing the model may provide a query as an output (e.g., a structured query language (SQL) query). Also see Paragraph [0025]), wherein the LLM is isolated from one or more entries of a medical database of interest, wherein the at least a processor is interfaceable with the medical database (Paragraph [0048] discloses, LLM receiving a natural language query which is not directly connected to from medical database);
receive from the LLM a feature set comprising a requirement set of hyper-logical nodes, wherein the feature set comprises at least a combination feature and wherein the hyper-logical nodes are synthesized by: extracting a plurality of atomic elements from the first natural language database query; and merging the plurality of atomic elements using logical operators and temporal relations identified by the LLM within the first natural language database query (Paragraph [0056], [0060] discloses, extracting combination feature sets such as metrics and parameters that are in a pipeline from natural language query using logical operators and temporal relations and merging/ aggregating them based on the query (Examiner interprets atomic elements as elements in the pipeline). (Based on specification Paragraph [0072] hyper-logical nodes are created by merging feature set with a logical operator and temporal relations if any). Also see [0065] for temporal relations);
and generate an aggregated output by querying the medical database using the first medical database query (Paragraph [0060] FIG. 3A displays an example of input datasets (labeled “patients,” “hospitalization objects,” and “hospital objects” in the depicted image) and an example target dataset (labeled “hospitals_with_num_crit” in the depicted image) displayed via the GUI 300. The GUI 300 can provide for a user selecting the input datasets and, optionally, the target dataset, and selecting a button or other input mechanism to generate a pipeline based on those inputs. Responsive to the button or other input mechanism being activated, the GUI 300 may prompt the user to input an NL query (e.g., in textual format via a textbox, or in audio format). The computing system 600 may then use these inputs to implement process 100, thus generating a pipeline that may optionally be displayed or otherwise presented (e.g., in an audio format) by the GUI 300. FIG. 3B shows an example of such a pipeline. The depicted pipeline includes the three original input datasets, various transformations including two joins, a filter, and an aggregation, and an output dataset that matches certain parameters of the target dataset (e.g., matches the schema of the target dataset) (i.e., generating an aggregated output based on the input data sets/ feature sets/ input parameters). Also see Paragraph [0049], [0056]),
Rafidi et al fails to explicitly teach, utilize a medical database query map as an intermediary to map a nodal output of the isolated LLM to one or more correct elements within the medical database by generating a first medical database query as a function of the feature set; wherein generating the first medical database query comprises: inputting the at least a combination feature of a condensed feature set into a template; mapping the condensed feature set to a database-specific query syntax using the medical database query map and the template; and formatting the first medical database query into a structure executable by the medical database as a function of the mapping.
Belcher; Thomas (US 20180137177 A1) teaches, utilize a medical database query map as an intermediary to map a nodal output of the isolated LLM to one or more correct elements within the medical database by generating a first medical database query as a function of the feature set (Fig. 1 Paragraph [0100]-[0103] discloses, generating a query based on feature set, by filtering different properties based on lexical constraints containing relevant values as an intermediary step to retrieve only a dataset exactly corresponding to data indicated by an RLQL query/ output correct elements within the medical database. Also see [0094]-[0095]).
wherein generating the first medical database query comprises: inputting the at least a combination feature of a condensed feature set into a template (Paragraph [0197] In some embodiments, system 700 generates SQL queries using a template-driven approach, where each SQL filter 400 is generated based on the corresponding filter 300 and instantiated with properties identifying the corresponding database table and columns);
Belcher also further teaches, the feature set comprises at least a combination feature and wherein the hyper-logical nodes are synthesized by: extracting a plurality of atomic elements from the first natural language database query; and merging the plurality of atomic elements using logical operators and temporal relations identified by the LLM within the first natural language database query (Paragraphs [0100], [0103], discloses, feature set comprising hyper-logical nodes by extracting elements from the query and generating/ creating SQL filters and merging the elements using the logical operators and temporal relations identified from the query);
and generate an aggregated output by querying the medical database using the first medical database query (Paragraphs [0111]-[0115] the SQL builder 600 can generate an SQL query that, when executed over a dataset in a data repository, returns a larger dataset than the one indicating by the RLQL query 100. Subsequent filtering can be performed on the larger dataset to extract only the data corresponding to data indicated by the RLQL query 100 (i.e., an aggregated output is returned to the user);
Therefore it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention, to have modified the teachings of Rafidi et al by providing, utilize a medical database query map as an intermediary to map a nodal output of the isolated LLM to one or more correct elements within the medical database by generating a first medical database query as a function of the feature set; wherein generating the first medical database query comprises: inputting the at least a combination feature of a condensed feature set into a template, as taught by Belcher et al (Paragraph [0100]).
One of the ordinary skill in the art would have been motivated to make this modification, by doing so, improves computer performance, for example, as it can be more efficient to execute a single SQL query rather than execute multiple queries and then chain the results together as taught by Belcher et al (Paragraph [0100]).
Rafidi et al and Belcher et al fails to explicitly teach, mapping the condensed feature set to a database-specific query syntax using the medical database query map and the template; and formatting the first medical database query into a structure executable by the medical database as a function of the mapping.
SHAHRIAR et al further teaches, mapping the condensed feature set to a database-specific query syntax using the medical database query map and the template (Paragraph [0035] discloses, mapping the feature set to database specific syntax/ data service using the query map and the template (medical database query is taught by Rafidi et al (Paragraph [0056]);
and formatting the first medical database query into a structure executable by the medical database as a function of the mapping database (Paragraphs [0039], [0040] the combined results (e.g., the combination of the parsing results and the results of the one or more models) may be converted to a query having the query format associated with the data service 102. The query format may be a format associated with a corresponding database language. The combined results may be formatted according to one or more rules, constraints, and/or the like associated with the databases language. The resulting query may be a valid query in the database language).
SHAHRIAR et al also teaches, wherein generating the first medical database query comprises: inputting the at least a combination feature of a condensed feature set into a template (Paragraph [0022], [0023] A data store, such as a database, typically stores specific information related to a specialized field. A user that wants to retrieve data in the specialized field may ask questions for very specific types of data from such a data store. A medical researcher may want to retrieve blood pressure data from patients within a certain age range or a sales person may want to determine the amount of sales for one or more products over the last year. In such cases, conventional algorithms (e.g., WikiSQL based algorithms) for translating natural language queries into database queries fail to fulfill the exclusive needs of those specialists. The disclosed techniques addresses this problem at least in part by using a supervised template-driven process that can be trained and validated for specialized scenarios. The disclosed approach may be generally referred to herein as Text2SQL, which may comprise a natural language to database query translation process that may be trained for specific use cases. [0023] The disclosed techniques may comprise the use of templates. A template may comprise a sample question (e.g., in a natural language) associated with a corresponding database query (e.g., SQL queries). Templates may be determined and/or stored for frequently requested data searches. Based on the question templates, the disclosed techniques may adjust to an individual use case by learning the domain questions and filtering out those questions that are outside of its domain of knowledge. The table schema as well as the stored data types and values may be determined. Using the stored data information, the text of a question may be preprocessed by utilizing a heuristic search process to extract detectable conditions from the question and to reduce the question to a basic form. The basic form may be later input into multiple machine learning models configured to determine query language, such as query modifiers. The results of the heuristic search process may be combined with the results of the by multiple machine learning models to form a complete query. Also see Paragraphs [0039], [0109]).
Therefore it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention, to have modified the teachings of Rafidi et al by mapping the condensed feature set to a database-specific query syntax using the medical database query map and the template; and formatting the first medical database query into a structure executable by the medical database as a function of the mapping., as taught by SHAHRIAR et al (Paragraphs [0035], [0039], [0040])
One of the ordinary skill in the art would have been motivated to make this modification, by doing so, the base question may be input into one or more machine learning models. Since the base question is a simplified version of the original question, the machine learning models may yield more accurate results, be more efficient, and otherwise improve upon traditional techniques as taught by SHAHRIAR et al (Paragraph [0004]).
Regarding dependent claim 2, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the apparatus of claim 1.
Rafidi et al further teaches, wherein the memory contains instructions configuring the at least a processor (Fig. 4 elements 612, 604, Paragraph [0061]) to output the aggregated output to a user as a function of a medical database response subject count (Paragraph [0056] In some embodiments, the process 230 includes applying the model to the first query, thus generating a NL description of the input data pipeline. The NL description of the input data pipeline can be in any appropriate format (e.g., textual or audio). The NL description may include an explanation of one or more metrics or parameters that the pipeline can be used to determine, and may include a description of conditions defined by the pipeline. For example, one NL description may be a string that states “This pipeline is counting the number of patients who have recovered from COVID-19 and were in critical condition in Seattle” (As best understood by the examiner, with instant specification (Fig, Paragraph [0054] As used herein, a “medical database response subject count” is the number of subjects whose data is included in medical database response as guidance Examiner interprets counting number of patients included in the response as subject count), where the number of patients is a metric that the pipeline can be used to determine, and the conditions include having recovered from covid, having been in critical condition, and having been a patient in Seattle. The model may translate conditions defined in the pipeline (e.g., by referencing and translating corresponding conditions defined in the SQL query) into NL (e.g., into at least a portion of the NL description).
Belcher et al also teaches, wherein the memory contains instructions configuring the at least processor to output the aggregated output to a user as a function of a medical database response subject count (Paragraphs [0111]-[0115] the SQL builder 600 can generate an SQL query that, when executed over a dataset in a data repository, returns a larger dataset than the one indicating by the RLQL query 100. Subsequent filtering can be performed on the larger dataset to extract only the data corresponding to data indicated by the RLQL query 100 (i.e., an aggregated output is returned to the user based on the response of the query)).
Regarding dependent claim 3, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the apparatus of claim 1.
Rafidi et al further teaches, wherein the memory contains instructions configuring the at least a processor (Fig. 4 elements 612, 604, Paragraph [0061]) to receive the first natural language database query as a function of a user input of a user (Paragraph [0025] the process 110 includes receiving an NL query, one or more input datasets. Also see Paragraph [0065]).
Regarding dependent claim 4, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the apparatus of claim 1.
Rafidi et al further teaches, wherein generating the medical database query comprises creating the condensed feature set containing at least one combination feature as a function of the feature set (Paragraph [0025] The NL query may include language that indicates certain constraints on the desired data (e.g., may include language specifying a date range, or an age range. (As best understood by the examiner, with instant specification Paragraph [0050] As used herein, a “condensed feature set” is a feature set including a combination feature. As used herein, a “combination feature” is a data structure including 2 or more features and a logical operator….a combination feature may indicate that subjects must be above 40 and under 50. In another non-limiting example, a combination feature may indicate that subjects must be on a first drug but not on a second drug. A combination feature may describe a range of valid values (such as age between 40 and 50), a list of valid categorical items (such as on a first drug or on a second drug), or the like) as guidance Examiner interprets a condensed feature set containing at least one combination feature as a function of the feature set as a certain constraints on the desired data, which includes age range)).
Regarding dependent claim 6, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the apparatus of claim 1.
Rafidi et al further teaches, wherein the memory contains instructions configuring the at least processor (Fig. 4 elements 612, 604, Paragraph [0061]) to output the feature set to a user (Fig. 5 Paragraph [0070]- [0083] In certain embodiments, the computing system is configured to make sure data pipelines are accurate. In some embodiments, the computing system interacts with the model solution to figure out if, given the NL query, the model solution has the right level of understanding of the concepts in the NL query. If not, the model solution, via the computing system, prompt the user for an explanation (i.e., Examiner interprets output the feature set as prompting the user for explanation/ unmatched column regarding the dataset), and the computing system can feed the explanation back to the model to ensure the most accurate pipeline is generated. In certain embodiments, the computing system can tie the explanation back to the datasets (e.g., the input datasets, the target dataset), to make sure that that the explanation (e.g., context) is stored for the pipelining attempts. [0072] “Who is the CTO?” can be generated).
Regarding dependent claim 7, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the apparatus of claim 1.
Rafidi et al further teaches, wherein the memory contains instructions configuring the at least a processor (Fig. 4 elements 612, 604, Paragraph [0061]) to: receive a second natural language database query (Fig. 5 Paragraph [0070]- [0083] In certain embodiments, the computing system is configured to make sure data pipelines are accurate. In some embodiments, the computing system interacts with the model solution to figure out if, given the NL query, the model solution has the right level of understanding of the concepts in the NL query. If not, the model solution, via the computing system, prompt the user for an explanation, and the computing system can feed the explanation back to the model to ensure the most accurate pipeline is generated (i.e., Examiner interprets the explanation provided by the user as a second query). In certain embodiments, the computing system can tie the explanation back to the datasets (e.g., the input datasets, the target dataset), to make sure that that the explanation (e.g., context) is stored for the pipelining attempts. [0073] According to certain embodiments, at process 540, the computing system presents or transmits (e.g., to another computing device) the one or more additional NL queries. In some embodiments, at process 545, the computing system receives one or more explanations corresponding to the one or more additional NL queries. In certain embodiments, at process 515, the computing system can incorporate the one or more explanations to the model query. In some embodiments, the computing system can incorporate the one or more explanations into the one or more input datasets and/or the target dataset. In the previous example, the computing system may receive an explanation of “CTO is Joe Doe” and incorporate it to the model query (i.e., receiving second natural language query associating the first natural language query)).
and modify the feature set as a function of the second natural language database query (Paragraph [0073] In some embodiments, at process 545, the computing system receives one or more explanations corresponding to the one or more additional NL queries. In certain embodiments, at process 515, the computing system can incorporate the one or more explanations to the model query. In some embodiments, the computing system can incorporate the one or more explanations into the one or more input datasets and/or the target dataset. In the previous example, the computing system may receive an explanation of “CTO is Joe Doe” and incorporate it to the model query (i.e., modifying the feature set by incorporating the second natural language query/ explanation of a feature/ term)).
Regarding dependent claim 8, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the apparatus of claim 1.
SHAHRIAR et al further teaches, wherein the memory contains instructions configuring the at least a processor to train the LLM on a training dataset including a plurality of example natural language database queries as inputs correlated to a plurality of example feature sets as outputs (Paragraph [0133] Data Preparation for the example Text2SQL engine is described as follows. The example Text2SQL engine was trained with question-query templates based on the use case. Several base questions (e.g., basic questions) which are frequently asked on the supply chain dataset were determined. By determining the common traits among the questions, question templates were determined).
Therefore it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention, to have modified the teachings of Rafidi et al and by providing wherein the memory contains instructions configuring the at least processor to train the LLM on a training dataset including a plurality of example natural language database queries as inputs correlated to a plurality of example feature sets as outputs, as taught by SHAHRIAR et al (Paragraph [0133])
One of the ordinary skill in the art would have been motivated to make this modification, by doing so, the base question may be input into one or more machine learning models. Since the base question is a simplified version of the original question, the machine learning models may yield more accurate results, be more efficient, and otherwise improve upon traditional techniques as taught by SHAHRIAR et al (Paragraph [0004], [0022]).
Regarding dependent claim 9, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the apparatus of claim 1.
Rafidi et al further teaches, wherein the memory contains instructions configuring the at least a processor (Fig. 4 elements 612, 604, Paragraph [0061]) to generate a second medical database query as a function of the feature set (Paragraph [0072] According to some embodiments, at process 530, the computing system can determine whether the confidence score associated with the model result and/or the query in the standard language is higher than a predetermined threshold. In certain embodiments, if the confidence score is lower than a predetermined threshold, at process 535, the computing system and/or the model solution can generate one or more additional NL queries. In the previous example, the additional NL query of “Who is the CTO?” can be generated (i.e., generating the second database query)).
Regarding dependent claim 10, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the apparatus of claim 9.
Rafidi et al further teaches, wherein the memory contains instructions configuring the at least a processor (Fig. 4 elements 612, 604, Paragraph [0061]) to generate the aggregated output as a function of a first medical database response responsive to the first medical database query and a second medical database response responsive to the second medical database query (Paragraph [0074] According to some embodiments, the computing system may receive or generate the model result including an SQL query, and optionally a confidence score. In the previous example, the generated SQL query can be: [0075] SELECT first_name, last_name, salary_payment_in_us_dollars [0076] FROM Employees [0077] JOIN Payments ON Employees.employee_id=Payments.employee_id [0078] WHERE salary_payment_in_us_dollars >(SELECT salary_payment_in_us_dollars [0079] FROM Employees [0080] JOIN Payments ON Employees.employee_id=Payments.employee_id [0081] WHERE first_name=‘John’ AND last_name=‘Doe’ AND payment_year=2020. Paragraph [0088] According to certain embodiments, at process 560, the computing system can apply the data pipeline to the one or more input datasets to generate an output dataset (i.e., aggregated output is generated based on the first and the second natural language queries).
Regarding independent claim 11, Rafidi; Joseph (US 20240045863 A1) teaches, a method of generating a medical database query (Paragraph [0047] In some embodiments, the process 120 includes generating a model query based on the NL query. Also see Paragraph [0055]), the method comprising: receiving, using at least a processor, a first natural language database query for a medical database comprising one or more entries (Paragraph [0025] In certain embodiments, the process 110 includes receiving an NL query, one or more input datasets (e.g., including one or more tables), and optionally one or more target datasets (e.g., including one or more tables). The NL query may be a query indicating some desired information, or one or more desired datasets. The NL query may include one or more strings. The NL query may include language that indicates certain constraints on the desired data (e.g., may include language specifying a date range, or an age range));
inputting, using the at least a processor, the first natural language database query into a large language model (LLM) ) (Paragraphs [0047], [0048] The model may be an NL processing model, such as a machine-learning NL processing model. For example, the model may be an autoregressive language model, such as a Generative Pre-trained Transformer 3 (GPT-3) model (i.e., Examiner interprets large language model (LLM) as machine learning models and GPT-3 is one of the model). Implementing the model may provide a query as an output (e.g., a structured query language (SQL) query). Also see Paragraph [0025]), wherein the LLM is isolated from the one or more entries of the medical database (Paragraph [0048] discloses, LLM receiving a natural language query which is not directly connected to from medical database);
receiving, using the at least a processor, from the LLM a feature set comprising a requirement set of hyper-logical nodes, wherein the feature set comprises at least a combination feature and wherein the hyper-logical nodes are synthesized by: extracting a plurality of atomic elements from the first natural language database query; and merging the plurality of atomic elements using logical operators and temporal relations identified by the LLM within the first natural language database query (Paragraph [0056], [0060] discloses, extracting combination feature sets such as metrics and parameters that are in a pipeline from natural language query using logical operators and temporal relations and merging/ aggregating them based on the query (Examiner interprets atomic elements as elements in the pipeline). (Based on specification Paragraph [0072] hyper-logical nodes are created by merging feature set with a logical operator and temporal relations if any). Also see [0065] for temporal relations);
and generating, using the at least a processor,(Paragraph [0060] FIG. 3A displays an example of input datasets (labeled “patients,” “hospitalization objects,” and “hospital objects” in the depicted image) and an example target dataset (labeled “hospitals_with_num_crit” in the depicted image) displayed via the GUI 300. The GUI 300 can provide for a user selecting the input datasets and, optionally, the target dataset, and selecting a button or other input mechanism to generate a pipeline based on those inputs. Responsive to the button or other input mechanism being activated, the GUI 300 may prompt the user to input an NL query (e.g., in textual format via a textbox, or in audio format). The computing system 600 may then use these inputs to implement process 100, thus generating a pipeline that may optionally be displayed or otherwise presented (e.g., in an audio format) by the GUI 300. FIG. 3B shows an example of such a pipeline. The depicted pipeline includes the three original input datasets, various transformations including two joins, a filter, and an aggregation, and an output dataset that matches certain parameters of the target dataset (e.g., matches the schema of the target dataset) (i.e., generating an aggregated output based on the input data sets/ feature sets/ input parameters). Also see Paragraph [0049], [0056]),
Rafidi et al fails to explicitly teach, utilizing, using the at least a processor a medical database query map as an intermediary to map a nodal output of the isolated LLM to one or more correct elements within the medical database by the condensed feature set to a database-specific query syntax using the medical database query map and the template; and formatting the first medical database query into a structure executable by the medical database as a function of the mapping.
Belcher; Thomas (US 20180137177 A1 ) teaches, utilizing, using the at least a processor a medical database query map as an intermediary to map a nodal output of the isolated LLM to one or more correct elements within the medical database by (Fig. 1 Paragraph [0100]-[0103] discloses, generating a query based on feature set, by filtering different properties based on lexical constraints containing relevant values as an intermediary step to retrieve only a dataset exactly corresponding to data indicated by an RLQL query/ output correct elements within the medical database. Also see [0094]-[0095]).
wherein generating the first medical database query comprises: inputting the at least a combination feature of a condensed feature set into a template (Paragraph [0197] In some embodiments, system 700 generates SQL queries using a template-driven approach, where each SQL filter 400 is generated based on the corresponding filter 300 and instantiated with properties identifying the corresponding database table and columns);
Belcher also further teaches, the feature set comprises at least a combination feature and wherein the hyper-logical nodes are synthesized by: extracting a plurality of atomic elements from the first natural language database query; and merging the plurality of atomic elements using logical operators and temporal relations identified by the LLM within the first natural language database query (Paragraphs [0100], [0103], discloses, feature set comprising hyper-logical nodes by extracting elements from the query and generating/ creating SQL filters and merging the elements using the logical operators and temporal relations identified from the query).
and generate an aggregated output by querying the medical database using the first medical database query (Paragraphs [0111]-[0115] the SQL builder 600 can generate an SQL query that, when executed over a dataset in a data repository, returns a larger dataset than the one indicating by the RLQL query 100. Subsequent filtering can be performed on the larger dataset to extract only the data corresponding to data indicated by the RLQL query 100 (i.e., an aggregated output is returned to the user);
Rafidi et al and Belcher et al fails to explicitly teach, mapping the condensed feature set to a database-specific query syntax using the medical database query map and the template; and formatting the first medical database query into a structure executable by the medical database as a function of the mapping.
SHAHRIAR et al further teaches, mapping the condensed feature set to a database-specific query syntax using the medical database query map and the template (Paragraph [0035] discloses, mapping the feature set to database specific syntax using the query map and the template (medical database query is taught by Rafidi et al (Paragraph [0056]);
and formatting the first medical database query into a structure executable by the medical database as a function of the mapping (Paragraphs [0039], [0040] the combined results (e.g., the combination of the parsing results and the results of the one or more models) may be converted to a query having the query format associated with the data service 102. The query format may be a format associated with a corresponding database language. The combined results may be formatted according to one or more rules, constraints, and/or the like associated with the databases language. The resulting query may be a valid query in the database language).
SHAHRIAR et al also teaches, wherein generating the first medical database query comprises: inputting the at least a combination feature of a condensed feature set into a template (Paragraph [0022], [0023] A data store, such as a database, typically stores specific information related to a specialized field. A user that wants to retrieve data in the specialized field may ask questions for very specific types of data from such a data store. A medical researcher may want to retrieve blood pressure data from patients within a certain age range or a sales person may want to determine the amount of sales for one or more products over the last year. In such cases, conventional algorithms (e.g., WikiSQL based algorithms) for translating natural language queries into database queries fail to fulfill the exclusive needs of those specialists. The disclosed techniques addresses this problem at least in part by using a supervised template-driven process that can be trained and validated for specialized scenarios. The disclosed approach may be generally referred to herein as Text2SQL, which may comprise a natural language to database query translation process that may be trained for specific use cases. [0023] The disclosed techniques may comprise the use of templates. A template may comprise a sample question (e.g., in a natural language) associated with a corresponding database query (e.g., SQL queries). Templates may be determined and/or stored for frequently requested data searches. Based on the question templates, the disclosed techniques may adjust to an individual use case by learning the domain questions and filtering out those questions that are outside of its domain of knowledge. The table schema as well as the stored data types and values may be determined. Using the stored data information, the text of a question may be preprocessed by utilizing a heuristic search process to extract detectable conditions from the question and to reduce the question to a basic form. The basic form may be later input into multiple machine learning models configured to determine query language, such as query modifiers. The results of the heuristic search process may be combined with the results of the by multiple machine learning models to form a complete query. Also see Paragraphs [0039], [0109]);
Therefore it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention, to have modified the teachings of Rafidi et al by mapping the condensed feature set to a database-specific query syntax using the medical database query map and the template; and formatting the first medical database query into a structure executable by the medical database as a function of the mapping., as taught by SHAHRIAR et al (Paragraphs [0035], [0039], [0040])
One of the ordinary skill in the art would have been motivated to make this modification, by doing so, the base question may be input into one or more machine learning models. Since the base question is a simplified version of the original question, the machine learning models may yield more accurate results, be more efficient, and otherwise improve upon traditional techniques as taught by SHAHRIAR et al (Paragraph [0004]).
Regarding dependent claim 12, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the method of claim 11.
Rafidi et al further teaches, wherein the method further comprises outputting the aggregated output to a user as a function of a medical database response subject count (Paragraph [0056] In some embodiments, the process 230 includes applying the model to the first query, thus generating a NL description of the input data pipeline. The NL description of the input data pipeline can be in any appropriate format (e.g., textual or audio). The NL description may include an explanation of one or more metrics or parameters that the pipeline can be used to determine and may include a description of conditions defined by the pipeline. For example, one NL description may be a string that states “This pipeline is counting the number of patients who have recovered from COVID-19 and were in critical condition in Seattle” (As best understood by the examiner, with instant specification (Fig, Paragraph [0054] As used herein, a “medical database response subject count” is the number of subjects whose data is included in medical database response as guidance Examiner interprets counting number of patients included in the response as subject count), where the number of patients is a metric that the pipeline can be used to determine, and the conditions include having recovered from covid, having been in critical condition, and having been a patient in Seattle. The model may translate conditions defined in the pipeline (e.g., by referencing and translating corresponding conditions defined in the SQL query) into NL (e.g., into at least a portion of the NL description).
Belcher et al also teaches, wherein the memory contains instructions configuring the at least processor to output the aggregated output to a user as a function of a medical database response subject count (Paragraphs [0111]-[0115] the SQL builder 600 can generate an SQL query that, when executed over a dataset in a data repository, returns a larger dataset than the one indicating by the RLQL query 100. Subsequent filtering can be performed on the larger dataset to extract only the data corresponding to data indicated by the RLQL query 100 (i.e., an aggregated output is returned to the user based on the response of the query)).
Regarding dependent claim 13, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the method of claim 11.
Rafidi et al further teaches, wherein the first natural language database query is received as a function of a user input of a user (Paragraph [0025] the process 110 includes receiving an NL query, one or more input datasets. Also see Paragraph [0065]).
Regarding dependent claim 14, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the method of claim 11.
Rafidi et al further teaches, wherein generating the medical database query comprises creating the condensed feature set containing at least one combination feature as a function of the feature set (Paragraph [0025] The NL query may include language that indicates certain constraints on the desired data (e.g., may include language specifying a date range, or an age range. (As best understood by the examiner, with instant specification Paragraph [0050] As used herein, a “condensed feature set” is a feature set including a combination feature. As used herein, a “combination feature” is a data structure including 2 or more features and a logical operator….a combination feature may indicate that subjects must be above 40 and under 50. In another non-limiting example, a combination feature may indicate that subjects must be on a first drug but not on a second drug. A combination feature may describe a range of valid values (such as age between 40 and 50), a list of valid categorical items (such as on a first drug or on a second drug), or the like) as guidance Examiner interprets a condensed feature set containing at least one combination feature as a function of the feature set as a certain constraints on the desired data, which includes age range)).
Regarding dependent claim 16, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the method of claim 11.
Rafidi et al further teaches, wherein the method further comprises, using the at least a processor, outputting the feature set to a user (Fig. 5 Paragraph [0070]- [0083] In certain embodiments, the computing system is configured to make sure data pipelines are accurate. In some embodiments, the computing system interacts with the model solution to figure out if, given the NL query, the model solution has the right level of understanding of the concepts in the NL query. If not, the model solution, via the computing system, prompt the user for an explanation (i.e., Examiner interprets output the feature set as prompting the user for explanation/ unmatched column regarding the dataset), and the computing system can feed the explanation back to the model to ensure the most accurate pipeline is generated. In certain embodiments, the computing system can tie the explanation back to the datasets (e.g., the input datasets, the target dataset), to make sure that that the explanation (e.g., context) is stored for the pipelining attempts. [0072] “Who is the CTO?” can be generated).
Regarding dependent claim 17, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the method of claim 11.
Rafidi et al further teaches, wherein the method further comprises: using the at least a processor, receiving a second natural language database query (Fig. 5 Paragraph [0070]- [0083] In certain embodiments, the computing system is configured to make sure data pipelines are accurate. In some embodiments, the computing system interacts with the model solution to figure out if, given the NL query, the model solution has the right level of understanding of the concepts in the NL query. If not, the model solution, via the computing system, prompt the user for an explanation, and the computing system can feed the explanation back to the model to ensure the most accurate pipeline is generated (i.e., Examiner interprets the explanation provided by the user as a second query). In certain embodiments, the computing system can tie the explanation back to the datasets (e.g., the input datasets, the target dataset), to make sure that that the explanation (e.g., context) is stored for the pipelining attempts. [0073] According to certain embodiments, at process 540, the computing system presents or transmits (e.g., to another computing device) the one or more additional NL queries. In some embodiments, at process 545, the computing system receives one or more explanations corresponding to the one or more additional NL queries. In certain embodiments, at process 515, the computing system can incorporate the one or more explanations to the model query. In some embodiments, the computing system can incorporate the one or more explanations into the one or more input datasets and/or the target dataset. In the previous example, the computing system may receive an explanation of “CTO is Joe Doe” and incorporate it to the model query (i.e., receiving second natural language query associating the first natural language query));
and using the at least a processor, modifying the feature set as a function of the second natural language database query (Paragraph [0073] In some embodiments, at process 545, the computing system receives one or more explanations corresponding to the one or more additional NL queries. In certain embodiments, at process 515, the computing system can incorporate the one or more explanations to the model query. In some embodiments, the computing system can incorporate the one or more explanations into the one or more input datasets and/or the target dataset. In the previous example, the computing system may receive an explanation of “CTO is Joe Doe” and incorporate it to the model query (i.e., modifying the feature set by incorporating the second natural language query/ explanation of a feature/ term)).
Regarding dependent claim 18, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the method of claim 11.
SHAHRIAR et al further teaches, wherein the memory contains instructions configuring the at least processor to train the LLM on a training dataset including a plurality of example natural language database queries as inputs correlated to a plurality of example feature sets as outputs (Paragraph [0133] Data Preparation for the example Text2SQL engine is described as follows. The example Text2SQL engine was trained with question-query templates based on the use case. Several base questions (e.g., basic questions) which are frequently asked on the supply chain dataset were determined. By determining the common traits among the questions, question templates were determined).
Therefore it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention, to have modified the teachings of Rafidi et al by providing wherein the memory contains instructions configuring the at least processor to train the LLM on a training dataset including a plurality of example natural language database queries as inputs correlated to a plurality of example feature sets as outputs, as taught by SHAHRIAR et al (Paragraph [0133])
One of the ordinary skill in the art would have been motivated to make this modification, by doing so, the base question may be input into one or more machine learning models. Since the base question is a simplified version of the original question, the machine learning models may yield more accurate results, be more efficient, and otherwise improve upon traditional techniques as taught by SHAHRIAR et al (Paragraph [0004], [0022]).
Regarding dependent claim 19, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the method of claim 11.
Rafidi et al further teaches, wherein the method further comprises generating a second medical database query as a function of the feature set (Paragraph [0072] According to some embodiments, at process 530, the computing system can determine whether the confidence score associated with the model result and/or the query in the standard language is higher than a predetermined threshold. In certain embodiments, if the confidence score is lower than a predetermined threshold, at process 535, the computing system and/or the model solution can generate one or more additional NL queries. In the previous example, the additional NL query of “Who is the CTO?” can be generated (i.e., generating the second database query)).
Regarding dependent claim 20, Rafidi et al, Belcher et al and SHAHRIAR et al teach, the method of claim 19.
Rafidi et al further teaches, wherein the aggregated output is generated as a function of a first medical database response responsive to the first medical database query and a second medical database response responsive to the second medical database query (Paragraph [0074] According to some embodiments, the computing system may receive or generate the model result including an SQL query, and optionally a confidence score. In the previous example, the generated SQL query can be: [0075] SELECT first_name, last_name, salary_payment_in_us_dollars [0076] FROM Employees [0077] JOIN Payments ON Employees.employee_id=Payments.employee_id [0078] WHERE salary_payment_in_us_dollars >(SELECT salary_payment_in_us_dollars [0079] FROM Employees [0080] JOIN Payments ON Employees.employee_id=Payments.employee_id [0081] WHERE first_name=‘John’ AND last_name=‘Doe’ AND payment_year=2020. Paragraph [0088] According to certain embodiments, at process 560, the computing system can apply the data pipeline to the one or more input datasets to generate an output dataset (i.e., aggregated output is generated based on the first and the second natural language queries).
Closest Prior Art
8. The prior art made of record and not relied upon is considered pertinent to the applicant’s disclosure.
Nagaraju; Divija (US 20240185001 A1) teaches, Disclosed are systems and techniques that may generate datasets for training task-oriented dialogue systems. The techniques include generating natural language queries by selecting a template query, sampling one or more tokens from a data store of domain-specific tokens, modifying the selected template query using the one or more sampled tokens to generate a query prompt, and using a natural language generative machine-learning model to generate, based on the query prompt, a respective natural language query of the subset of the plurality of natural language queries, and causing the generated plurality of natural language queries to be provided to a machine-learning model training engine configured to train, using the generated plurality of natural language queries, a conversational machine-learning model to perform a domain-specific conversational task (Abstract).
9. Examiner has pointed out particular references contained in the prior arts of record in the body of this action for the convenience of the applicant. Although the specified citations are representative of the teachings in the art and are applied to the specific limitations within the individual claim, other passages and Figures may apply as well. It is respectfully requested from the applicant, in preparing the response, to consider fully the entire references as potentially teaching all or part of the claimed invention, as well as the context of the passage as taught by the prior arts or disclosed by the examiner. It is noted that any citation to specific pages, columns, figures, or lines in the prior art references any interpretation of the references should not be considered to be limiting in any way. A reference is relevant for all it contains and may be relied upon for all that it would have reasonably suggested to one having ordinary skill in the art. In re Heck, 699 F.2d 1331-33, 216 USPQ 1038-39 (Fed. Cir. 1983) (quoting In re Lemelson, 397 F.2d 1006, 1009, 158 USPQ 275, 277 (CCPA 1968))).
Conclusion
Applicant’s amendments/Arguments necessitated the rejection as presented in this office action. THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any extension fee pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SUMAN RAJAPUTRA whose telephone number is (571) 272-4669. The examiner can normally be reached between 8:00 AM - 5:00 PM.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Tony Mahmoudi (571) 272-4078 can be reached. 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.
/S. R./
Examiner, Art Unit 2163
/ALEX GOFMAN/Primary Examiner, Art Unit 2163
Prosecution Insights