DETAILED ACTION
This action is responsive to Remarks and Claim amendments filed on August 17, 2026.
Claims 1, 3-8 and 10-20 have been amended.
Claims 1-20 are pending and presented for examination.
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Examiner Notes
Examiner cites particular columns, paragraphs, figures and line numbers in the references as applied to the claims below 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 that, in preparing responses, the applicant fully consider the references in their entirety as potentially teaching all or part of the claimed invention, as well as the context of the passage as taught by the prior art or disclosed by the examiner.
Response to Amendment
The objection to the abstract of the disclosure is maintained, as set forth under Specification below.
The objections to claims 10 and 13-14 set forth in the prior action are withdrawn in view of Applicant's amendments. A new objection to claim 16 is set forth below, necessitated by Applicant's amendment.
The rejection of claims 7-8 and 17-19 under 35 U.S.C. 112(b) is withdrawn. Applicant has amended claim 7 to depend from claim 5 rather than claim 6, and claim 17 to depend from claim 15 rather than claim 16, and has added to each an affirmative recitation that the task-specific data element comprises a task-specific feature value and a feature time point. The contingent antecedent basis defect identified in the prior action is thereby resolved, and the rejection of dependent claims 8 and 18-19 predicated on that defect is likewise withdrawn.
Specification
Applicant is reminded of the proper language and format for an abstract of the disclosure. The abstract should be in narrative form and generally limited to a single paragraph on a separate sheet within the range of 50 to 150 words in length. The abstract should describe the disclosure sufficiently to assist readers in deciding whether there is a need for consulting the full patent text for details.
The language should be clear and concise and should not repeat information given in the title. It should avoid using phrases which can be implied, such as, "The disclosure concerns," "The disclosure defined by this invention," "The disclosure describes," etc. In addition, the form and legal phraseology often used in patent claims, such as "means" and "said," should be avoided.
The abstract of the disclosure is objected to because the substitute abstract filed with the Amendment begins with "The present disclosure provides…". The prior action objected to the abstract because it began with "Various embodiments of the present disclosure provide…". The substituted language is a phrase which can be implied, of the same character as the language previously objected to, and the objection is therefore maintained. Applicant is invited to begin the abstract with a direct recitation of the subject matter, for example, --A contextualized task-specific graphical visualization is generated from third-party data…--. A corrected abstract of the disclosure is required and must be presented on a separate sheet, apart from any other text. See MPEP § 608.01(b).
Claim Objections
Claim 16 is objected to because of the following informalities. This objection is newly presented and is necessitated by Applicant's amendment.
Claim 16 recites "wherein the task-specific data element of the task-specific data element comprises a task-specific feature value, a feature time point, the correlation indicator, or a relevancy indicator" (emphasis on the duplicated recitation). The Amendment deleted the phrase "one or more" from the recitation "the task-specific data element of the one or more task-specific data elements" but did not delete the remainder of the prepositional phrase. By comparison, corresponding claim 6 deleted the entire phrase "of the one or more task-specific data elements." As presently recited, claim 16 refers the task-specific data element to itself. Examiner suggests deleting "of the task-specific data element" from claim 16 so that the claim reads "wherein the task-specific data element comprises a task-specific feature value, a feature time point, the correlation indicator, or a relevancy indicator," consistent with claim 6.
Appropriate correction is required.
Claim Rejections — 35 U.S.C. § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention recites a judicial exception, is directed to that judicial exception, an abstract idea, as it has not been integrated into practical application and the claims further do not recite significantly more than the judicial exception. Examiner has evaluated the claims under the framework provided in the 2019 Patent Eligibility Guidance published in the Federal Register 01/07/2019 and has provided such analysis below.
Step 1: Claims 1-10 are directed to methods and fall within the statutory category of processes; Claims 11-19 are directed to systems and fall within the statutory category of machines; and Claim 20 is directed to a medium and falls within the statutory category of manufactures. Therefore, "Are the claims to a process, machine, manufacture or composition of matter?" Yes.
In order to evaluate the Step 2A inquiry "Is the claim directed to a law of nature, a natural phenomenon or an abstract idea?" we must determine, at Step 2A Prong 1, whether the claim recites a law of nature, a natural phenomenon or an abstract idea and further whether the claim recites additional elements that integrate the judicial exception into a practical application.
Step 2A Prong 1:
Claims 1, 11 and 20 as drafted, recite a process that, under its broadest reasonable interpretation, covers steps that could reasonably be performed in the mind, including with the aid of pen and paper, but for the recitation of generic computer components. That is, the limitations:
a) "generating, by one or more processors and using a plurality of first-party data ingestion protocols, a defined data object by transforming a plurality of third-party data elements from a third-party data source to a defined first-party format;" – Mental processes (see MPEP 2106.04(a)(2), III). For example, a person may gather patient records from a provider and transcribe them into a uniform format.
b) "generating, by the one or more processors and using a machine learning formatting model, a structured data object from the defined data object based on a data structure format comprising a set of format features defined by a machine learning formatting prompt that comprises a first prompt data object generated based on a first interactive user interface session associated with the machine learning formatting model;" – Mental processes (see MPEP 2106.04(a)(2), III). For example, a person may organize gathered information into structured categories according to instructions that person received during a conversation.
c) "generating, by the one or more processors and using a machine learning relevancy model, a task-specific data object from the structured data object and corresponding to a defined domain task by filtering a plurality of task-agnostic features of the structured data object using a task-specific prompt that (i) defines a set of task-related features and (ii) comprises a second prompt data object generated based on a second interactive user interface session associated with the machine learning relevancy model;" – Mental processes (see MPEP 2106.04(a)(2), III). For example, a person may identify which records are relevant to a particular clinical question specified in a conversation and disregard the rest; this is pure observation, evaluation and judgment.
d) "initiating, by the one or more processors and via a user interface, a rendering of a contextualized task-specific graphical visualization that is based on the task-specific data object and comprises a set of interactive graphical elements for the defined domain task." – Mental processes (see MPEP 2106.04(a)(2), III). For example, a person may present the curated information for review by providing a summary of findings on a chart or whiteboard.
That is, nothing in the claim elements precludes the steps from practically being performed in the mind or with a pen and paper (i.e., "generating", "initiating") through observation, evaluation, judgment and opinion with the aid of pen and paper.
The addition, by amendment, of the recitations that each prompt "comprises a … prompt data object generated based on a … interactive user interface session" does not remove the limitations from the Mental Processes grouping. The recitation specifies only the source from which the instruction is obtained, namely a user interaction. Instructions describing how information is to be organized and which information is relevant to a task have long been communicated between persons; a clinician who tells a colleague to arrange a chart by date and to retain only records bearing on a particular workup has supplied the same instruction. Reciting that the instruction is obtained by way of a user interface addresses the additional element considered at Prong 2 below and does not alter the character of the recited exception.
Thus, these limitations fall within the "Mental Processes" grouping of abstract ideas.
Therefore, Yes, claims 1, 11 and 20 recite judicial exceptions.
The claims have been identified to recite judicial exceptions, Step 2A Prong 2 will evaluate whether the claims are directed to the judicial exception.
Step 2A Prong 2:
This judicial exception is not integrated into a practical application. The claims recite the following additional elements: "one or more processors", "user interface", "machine learning formatting model", "machine learning relevancy model", "machine learning formatting prompt", "task-specific prompt", "first interactive user interface session", "second interactive user interface session" and "interactive graphical elements". The additional elements are merely instructions to implement an abstract idea on a computer, or merely using a generic computer or computer components as a tool to perform the abstract idea (see MPEP 2106.05(f)).
– "one or more processors" – generic computing hardware (specification paragraphs [0031]-[0032] describe generic CPUs and memory).
– "user interface" – generic display/input mechanism (specification paragraph [0046] describes a generic browser/keypad/touch display).
– "machine learning formatting model" – characterized at the highest level of generality. Specification paragraph [0055] expressly admits: "the machine learning formatting model may be configured as a LLM, a generative pre-trained transformer (GPT) model, or another type of generative machine learning model".
– "machine learning relevancy model" – generic characterization. Specification paragraph [0057] expressly admits: "the machine learning relevancy model may be configured as a LLM, a GPT model, or another type of generative machine learning model".
– "machine learning formatting prompt" and "task-specific prompt" – generic prompt inputs (specification paragraphs [0056], [0058]).
– "first interactive user interface session" and "second interactive user interface session" – generic pre-solution data gathering by which the prompt is obtained from a user (see MPEP 2106.05(g)), implemented using a generic interface. Specification paragraphs [0056] and [0058] expressly admit that the prompt "is configured for communication via a network, an API, a machine learning model plug-in, and/or another type of interface" between a user device and the respective model.
– "interactive graphical elements" – generic UI features.
Examiner notes: the specification at paragraphs [0002]-[0003] frames the "improvement" as efficiency of data ingestion and the production of a "contextualized task-specific graphical visualization" — i.e., better information presentation for human users. This is an improvement to the abstract data-organization process, not to how the computer itself operates. The claims do not improve the machine learning models themselves, GUI rendering technology, or computer architecture.
The claims are directly analogous to Electric Power Group, LLC v. Alstom S.A., 830 F.3d 1350 (Fed. Cir. 2016), where the Federal Circuit held that "collecting information, analyzing it, and displaying certain results of the collection and analysis" using conventional computer technology was patent-ineligible. The claims are also analogous to In re TLI Communications LLC Patent Litig., 823 F.3d 607 (Fed. Cir. 2016), where using a generic computer to organize and classify data was held abstract; and SAP America, Inc. v. InvestPic, LLC, 898 F.3d 1161 (Fed. Cir. 2018), where mathematical analysis displayed on a screen was held abstract.
The claims are further controlled by Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1205 (Fed. Cir. 2025). There the Federal Circuit addressed as a question of first impression whether claims that do no more than apply established methods of machine learning to a new data environment are patent eligible, and held that patents which do no more than claim the application of generic machine learning to new data environments, without disclosing improvements to the machine learning models to be applied, are patent ineligible under 35 U.S.C. 101. The present claims apply generic machine learning models — which the specification at paragraphs [0055] and [0057] expressly permits to be any LLM, GPT model, or other generative model — to the data environment of multi-source patient records. Neither the claims nor the specification discloses any improvement to the models themselves.
Accordingly, the additional elements recited in the claims do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea, thus failing to integrate the abstract idea into a practical application.
Therefore, "Do the claims recite additional elements that integrate the judicial exception into a practical application?" No, these additional elements do not integrate the abstract idea into a practical application and they do not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea.
After having evaluated the inquiries set forth in Step 2A Prong 1 and Prong 2, it has been concluded that claims 1, 11 and 20 not only recite a judicial exception but that the claims are directed to the judicial exception as the judicial exception has not been integrated into practical application.
Step 2B:
As discussed above with respect to integration of the abstract idea into a practical application, the additional elements "one or more processors", "user interface", "machine learning formatting model", "machine learning relevancy model", "machine learning formatting prompt", "task-specific prompt", "first interactive user interface session", "second interactive user interface session" and "interactive graphical elements" are generic computer components used as tools to perform the abstract idea.
Further, the specification at paragraphs [0028], [0031]-[0032], [0044]-[0048], [0050], [0055]-[0058] expressly admits that the claim elements are well-understood, routine, conventional activity in the field, as evidenced by the express admissions quoted above (e.g., machine learning models configured as any LLM or GPT model; prompts communicated via a network, an API, a machine learning model plug-in, or another type of interface; first-party platforms; network protocols; and AI computing entities). Elements that a specification describes as interchangeable with any conventional alternative do not amount to significantly more. See MPEP 2106.05(d)(II).
Accordingly, the additional elements recited in the claims cannot provide an inventive concept. In addition, after further evaluation the claims as a whole do not improve any function of a computer or any other technology or technical field. Thus, the claims are not patent eligible.
Therefore, "Do the claims recite additional elements that amount to significantly more than the judicial exception?" No, these additional elements, alone or in combination, do not amount to significantly more than the judicial exception.
Having concluded analysis within the provided framework, Claims 1, 11 and 20 do not recite patent eligible subject matter under 35 U.S.C. § 101.
With regards to claim 2 (and similar for claim 12), it recites "receiving, at a particular time point and via the user interface, a task query for an entity that identifies the defined domain task and the entity corresponding to the structured data object; and responsive to the task query, receiving the structured data object corresponding to the particular time point, generating the task-specific data object based on the structured data object, and initiating the rendering of the contextualized task-specific graphical visualization based on the task-specific data object." as drafted, is mere instructions to apply an exception (see MPEP 2106.05(f)), which is mere automation using generic computing. Also, insignificant extra-solution activity (see MPEP 2106.05(g)) as data-gathering. Moreover, claim 2 does not recite any other additional elements and for the same reasons as above with regard to integration into practical application and whether additional elements amount to significantly more, claim 2 also fails both Step 2A Prong 2, thus the claim is directed to the judicial exception as it has not been integrated into practical application, and fails Step 2B as not amounting to significantly more. Therefore, Claim 2 does not recite patent eligible subject matter under 35 U.S.C. § 101.
With regards to claim 3 (and similar for claim 13), it recites "filtering the plurality of task-agnostic features of the structured data object based on a set of duplication rules that define a multi-tiered time-based constraint for the plurality of third-party data elements." as drafted, is a process that, but for the recitation of generic computing components, under its broadest reasonable interpretation, covers performance of the limitation in the mind and/or using pen and paper (Mental processes, see MPEP 2106.04(a)(2), III). Moreover, claim 3 does not recite any other additional elements and for the same reasons as above with regard to integration into practical application and whether additional elements amount to significantly more, claim 3 also fails both Step 2A Prong 2, thus the claim is directed to the judicial exception as it has not been integrated into practical application, and fails Step 2B as not amounting to significantly more. Therefore, Claim 3 does not recite patent eligible subject matter under 35 U.S.C. § 101.
With regards to claim 4 (and similar for claim 14), it recites "filtering the plurality of task-agnostic features of the structured data object based on a set of data quality rules that define a quality constraint for the third-party data source." as drafted, is a process that, but for the recitation of generic computing components, under its broadest reasonable interpretation, covers performance of the limitation in the mind and/or using pen and paper (Mental processes, see MPEP 2106.04(a)(2), III). Moreover, claim 4 does not recite any other additional elements and for the same reasons as above with regard to integration into practical application and whether additional elements amount to significantly more, claim 4 also fails both Step 2A Prong 2, thus the claim is directed to the judicial exception as it has not been integrated into practical application, and fails Step 2B as not amounting to significantly more. Therefore, Claim 4 does not recite patent eligible subject matter under 35 U.S.C. § 101.
With regards to claim 5 (and similar for claim 15), it recites "(i) the task-specific data object comprises a task-specific feature from the plurality of task-agnostic features, and (ii) the task-specific data element comprises a correlation indicator that maps the task-specific feature to the third-party data source." as drafted, is a process that, but for the recitation of generic computing components, under its broadest reasonable interpretation, covers performance of the limitation in the mind and/or using pen and paper (Mental processes, see MPEP 2106.04(a)(2), III). Annotating an item of information with its source is a clerical act long performed by hand. Moreover, claim 5 does not recite any other additional elements and for the same reasons as above with regard to integration into practical application and whether additional elements amount to significantly more, claim 5 also fails both Step 2A Prong 2, thus the claim is directed to the judicial exception as it has not been integrated into practical application, and fails Step 2B as not amounting to significantly more. Therefore, Claim 5 does not recite patent eligible subject matter under 35 U.S.C. § 101.
With regards to claim 6 (and similar for claim 16), it recites "wherein the task-specific data element comprises a task-specific feature value, a feature time point, the correlation indicator, or a relevancy indicator." as drafted, is a process that, but for the recitation of generic computing components, under its broadest reasonable interpretation, covers performance of the limitation in the mind and/or using pen and paper (Mental processes, see MPEP 2106.04(a)(2), III). Moreover, claim 6 does not recite any other additional elements and for the same reasons as above with regard to integration into practical application and whether additional elements amount to significantly more, claim 6 also fails both Step 2A Prong 2, thus the claim is directed to the judicial exception as it has not been integrated into practical application, and fails Step 2B as not amounting to significantly more. Therefore, Claim 6 does not recite patent eligible subject matter under 35 U.S.C. § 101.
With regards to claim 7 (and similar for claim 17), it recites "(a) the task-specific data element comprises a task-specific feature value and a feature time point, and (b) the set of interactive graphical elements corresponds to the task-specific data element, and an interactive graphical element corresponding to the task-specific data element: (i) is positioned within an interactive timeline chart based on the feature time point, (ii) presents the task-specific feature value, and (iii) comprises interactive presentation features that are presented responsive to a user selection or a detected selection intent of the interactive graphical element." as drafted, is mere instructions to apply an exception (see MPEP 2106.05(f)), generic instructions using generic computing and GUI elements. Also, insignificant extra-solution activity (see MPEP 2106.05(g)) as presenting results. Moreover, claim 7 does not recite any other additional elements and for the same reasons as above with regard to integration into practical application and whether additional elements amount to significantly more, claim 7 also fails both Step 2A Prong 2, thus the claim is directed to the judicial exception as it has not been integrated into practical application, and fails Step 2B as not amounting to significantly more. Therefore, Claim 7 does not recite patent eligible subject matter under 35 U.S.C. § 101.
With regards to claim 8 (and similar for claim 18), it recites "wherein the correlation indicator is an interactive presentation feature and comprises an interactive link that routes an entity from the user interface to the third-party data source." as drafted, is mere instructions to apply an exception (see MPEP 2106.05(f)), generic instructions using generic computing and GUI elements. Moreover, claim 8 does not recite any other additional elements and for the same reasons as above with regard to integration into practical application and whether additional elements amount to significantly more, claim 8 also fails both Step 2A Prong 2, thus the claim is directed to the judicial exception as it has not been integrated into practical application, and fails Step 2B as not amounting to significantly more. Therefore, Claim 8 does not recite patent eligible subject matter under 35 U.S.C. § 101.
With regards to claim 9 (and similar for claim 19), it recites "wherein the structured data object comprises a tabular format associated with respective timepoints for respective data elements of the structured data object." as drafted, is a process that, but for the recitation of generic computing components, under its broadest reasonable interpretation, covers performance of the limitation in the mind and/or using pen and paper (Mental processes, see MPEP 2106.04(a)(2), III). Arranging information in a table ordered by time is a clerical practice that predates computing. Moreover, claim 9 does not recite any other additional elements and for the same reasons as above with regard to integration into practical application and whether additional elements amount to significantly more, claim 9 also fails both Step 2A Prong 2, thus the claim is directed to the judicial exception as it has not been integrated into practical application, and fails Step 2B as not amounting to significantly more. Therefore, Claim 9 does not recite patent eligible subject matter under 35 U.S.C. § 101.
With regards to claim 10, it recites "generating, using the machine learning relevancy model, a plurality of relevancy scores for the plurality of task-agnostic features based on a semantic comparison between the plurality of task-agnostic features and the set of task-related features defined by the task-specific prompt; and identifying a task-specific feature from the plurality of task-agnostic features based on the plurality of relevancy scores." as drafted, is a process that, but for the recitation of generic computing components, under its broadest reasonable interpretation, covers performance of the limitation in the mind and/or using pen and paper (Mental processes, see MPEP 2106.04(a)(2), III). Further, it recites mathematical concepts and calculations in order to compute the relevancy score (e.g., cosine-distance computation), which is a separate judicial exception (Mathematical concepts, see MPEP 2106.04(a)(2), I). Moreover, claim 10 does not recite any other additional elements and for the same reasons as above with regard to integration into practical application and whether additional elements amount to significantly more, claim 10 also fails both Step 2A Prong 2, thus the claim is directed to the judicial exception as it has not been integrated into practical application, and fails Step 2B as not amounting to significantly more. Therefore, Claim 10 does not recite patent eligible subject matter under 35 U.S.C. § 101.
Therefore, Claims 1-20 do not recite patent eligible subject matter under 35 U.S.C. § 101.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-20 are rejected under 35 U.S.C. 103 as being unpatentable over Dhillon et al. (US Pub. No. 2018/0113945, hereinafter "Dhillon" – previously presented) in view of Li et al. (US Pub. No. 2023/0367799, hereinafter "Li" – previously presented), and further in view of Hawes et al. (US Pub. No. 2024/0403290, hereinafter "Hawes").
With respect to claim 1 (Currently Amended), Dhillon teaches A computer-implemented method comprising: (Dhillon discloses a computer-implemented method for processing electronic record data and generating a graphical user interface. See Dhillon paragraph [0031] ("System 100 may include one or more central processing units (CPUs, or processors) 120 and system memory 121.").)
generating, by one or more processors and using a plurality of first-party data ingestion protocols, a defined data object by transforming a plurality of third-party data elements from a third-party data source to a defined first-party format (Dhillon discloses receiving electronic record data from a remote third-party data source and processing that data through a plurality of ingestion operations that transform it into a defined format stored in a database. As to the recited processors, see Dhillon paragraph [0031] ("CPUs 120 may be single or multiple microprocessors, field-programmable gate arrays, or digital signal processors capable of execution sets of instructions stored in a memory"). As to the third-party data source and third-party data elements, see Dhillon paragraph [0036] and paragraph [0038], which disclose that the system receives data from sources including physician documents 201, laboratory testing 202, ambulatory charts 203, and hospital records 204. As to the plurality of first-party data ingestion protocols and the transformation to a defined first-party format, see Dhillon paragraph [0040], which discloses that the input filter 210 processes the data provided by each data source before storing the data in database 215, and that such processing may include recognizing the source and type of data, performing optical character recognition to make data searchable, extracting dates and keywords, generating metadata, and reorganizing the data as related to pre-determined categories or classifications. The processed data so stored constitutes the recited defined data object.)
generating, by the one or more processors [[and using a machine learning formatting model]], a structured data object from the defined data object based on a data structure format comprising a set of format features [[defined by a machine learning formatting prompt that comprises a first prompt data object generated based on a first interactive user interface session associated with the machine learning formatting model]] (Dhillon discloses generating, by its processors, a structured representation of the ingested data according to a defined format having a set of format features. See Dhillon paragraph [0042] (data selector 218 receives a request and retrieves data from database 215) and paragraph [0043] (index generator 220 and timeline generator 225 analyze the data, determine the most effective organization of the data based on the request, and display the data to the requestor). As to the data structure format comprising a set of format features, see Dhillon paragraphs [0049]-[0057], which disclose that the resulting structured representation is organized according to defined metadata fields including patient information 331, source 332, type 333, date stamp 334, permissions 335, and test results 336, and paragraph [0065], which discloses that the timeline-element format includes an event date, an event name, an event description, a reference number, and an image URL.)
generating, by the one or more processors [[and using a machine learning relevancy model, a task-specific data object from the structured data object and corresponding to a defined domain task by filtering a plurality of task-agnostic features of the structured data object using a task-specific prompt that (i) defines a set of task-related features and (ii) comprises a second prompt data object generated based on a second interactive user interface session associated with the machine learning relevancy model]] (Dhillon teaches that the recited generating operation is performed by one or more processors, as set forth above with respect to the preceding limitation (Dhillon paragraph [0031]).)
initiating, by the one or more processors and via a user interface, a rendering of a contextualized task-specific graphical visualization that is based on the task-specific data object and comprises a set of interactive graphical elements for the defined domain task (Dhillon discloses initiating, via a user interface, the rendering of an interactive graphical visualization that is contextualized to a particular requestor and condition and that comprises interactive graphical elements. See Dhillon paragraph [0064] ("System 100 may also build and display (step 560) the timeline as exemplified in FIG. 8. In some embodiments, system 100 may generate interface information for displaying an interactive graphical user interface that includes the timeline. The timeline may provide a chronological listing of data extracted from the records. For example, the timeline may include medical events, procedures, and consultations. The timeline may also be dynamically generated from the records based on the request, the requestor, permissions of the requestor, the patient, and the condition. The timeline may include hyperlinks, allowing quick access to the records."). The hyperlinks so disclosed constitute the recited set of interactive graphical elements, and the dynamic generation of the timeline based on the request, the requestor, the patient and the condition constitutes the recited contextualization for the defined domain task.)
Dhillon does not explicitly disclose: (i) generating the structured data object using a machine learning formatting model;
(ii) using a machine learning relevancy model, a task-specific data object from the structured data object and corresponding to a defined domain task by filtering a plurality of task-agnostic features of the structured data object;
(iii) that the data structure format is defined by a machine learning formatting prompt that comprises a first prompt data object generated based on a first interactive user interface session associated with the machine learning formatting model; and
(iv) that the filtering is performed using a task-specific prompt that (i) defines a set of task-related features and (ii) comprises a second prompt data object generated based on a second interactive user interface session associated with the machine learning relevancy model.
However, in an analogous art, Li teaches:
generating, by the one or more processors using a machine learning formatting model, a structured data object from the defined data object based on a data structure format comprising a set of format features (Li discloses an artificial intelligence model that processes input clinical text and produces a structured output. See Li paragraph [0012] ("The method includes accessing unstructured clinical documentation, converting the unstructured clinical documentation to vector representations, using an artificial intelligence (AI) model to encode the vector representations to obtain embeddings and processing the embeddings.") and paragraph [0026] ("A model 106 connects to the input unit 102 through a text converter 104. A processor 108 is connected to the model 106 to receive output from the model 106 and provide output for the system 100."). As to the model being a machine learning model, see Li paragraph [0058] ("For illustration purposes, the known BERT model is used in the following for the model 106 of FIG. 1."). The output of model 106 is a structured subject journal. See Li paragraph [0023] ("A "journal" may be understood to contain "entries" arranged by "facet" through time.").)
generating, by the one or more processors and using a machine learning relevancy model, a task-specific data object from the structured data object and corresponding to a defined domain task by filtering a plurality of task-agnostic features of the structured data object using a task-specific prompt (Li discloses that its model determines, for each candidate sentence in the input documentation, whether that sentence is relevant to a particular facet, and retains only those sentences that are, thereby filtering out the sentences that are not. See Li paragraph [0025] ("An "entry" is a sentence from the notes that is determined to be both relevant and related to the facet. A facet is expected to contain multiple entries."). The facet constitutes the recited defined domain task, the retained entries constitute the recited task-specific data object, and the sentences that are not determined to be relevant and related to the facet constitute the recited plurality of task-agnostic features that are filtered. As to the relevance determination being performed by the model, see Li paragraph [0040] ("The difficulty of determining the facet may be resolved, in computer science, at inference time with standard techniques. For one example, an unknown similarity and distance matrix property for a given facet may be determined using cosine distance as a runtime parameter.").)
It would have been obvious to one of ordinary skill in the art at the time the invention was made before the effective filing date of the claimed invention to modify Dhillon's data analysis and timeline generation system to incorporate Li's AI-model-based generation of a structured patient journal organized by facet, and to use Li's facet-relevance determination to filter task-agnostic features from the structured data. The motivation to do so includes reducing manual data abstraction effort, as expressly recognized by Li paragraph [0010], and enabling clinicians to rapidly understand a patient's experience through time without spending substantial time reviewing previous notes, as expressly recognized by Li paragraph [0028] ("Indeed, subject journals may be shown to allow clinicians to rapidly understand a patient's experience through time. It is known for clinicians to rely on the patient to remember details of patient's experience through time. Alternatively, the clinician may be expected to spend substantial time reviewing previous notes."). The combination is further supported because Dhillon and Li both operate on the same underlying data, namely multi-source unstructured and structured medical records, and Li's AI-model-based facet extraction provides a known and predictable improvement to Dhillon's metadata-based data selection, yielding the predictable benefit of automated, task-relevant content selection.
Dhillon in view of Li is silent to disclose defined by a machine learning formatting prompt that comprises a first prompt data object generated based on a first interactive user interface session associated with the machine learning formatting model and using a task-specific prompt that (i) defines a set of task-related features and (ii) comprises a second prompt data object generated based on a second interactive user interface session associated with the machine learning relevancy model; however, in an analogous art, Hawes teaches:
a data structure format comprising a set of format features defined by a machine learning formatting prompt that comprises a first prompt data object generated based on a first interactive user interface session associated with the machine learning formatting model (Hawes discloses a graphical user interface through which a user selects a prompt that is sent to a machine learning model and that carries the instructions defining the format of the model's output. See Hawes paragraph [0059] ("FIG. 4 shows an example user interface 400 in which a user can provide a prompt. The user interface 400 can include a system prompt selector 404 that allows a user to select a prompt to be sent to the LLM.") and, in the same paragraph ("The system prompt selector 404 may be used to create a long prompt comprising multiple paragraphs of specific instructions, examples of acceptable responses, logical explanations, instructions for formatting the response, and/or other aspects described herein."). Hawes further discloses that those formatting instructions define the set of format features of the output. See Hawes paragraph [0059] ("The instructions can include examples of desired or required formatting (e.g., grammatical structure, syntax, terminology, etc.).") and paragraph [0061] ("The output schema selector 416 may allow a user to select a target output of the LLM by stipulating a kind of one or more variables, a type of variable, and/or an acceptable format (e.g., string) of the output."). As to the prompt being generated based on an interactive user interface session associated with the model, see Hawes paragraph [0037] ("The system may comprise one or more interactive graphical user interfaces ("GUIs") configured to allow a user to create and/or input prompts to the computer model and/or to receive a response to the prompts.") and paragraph [0045] ("User interface service 123 may also receive data entered by a user into a client device, such as user device(s) 130, and may store and/or forward it to the other various components of the system 120.").)
filtering a plurality of task-agnostic features of the structured data object using a task-specific prompt that (i) defines a set of task-related features and (ii) comprises a second prompt data object generated based on a second interactive user interface session associated with the machine learning relevancy model (Hawes discloses a second, separately selected prompt, directed to the task to be performed, likewise obtained through the user interface. See Hawes paragraph [0061] ("A user may additionally or alternatively select a task prompt using a task prompt selector 412 of the user interface 400.") and, in the same paragraph ("The task prompt may be concatenated to or otherwise included with the system prompt to create the final prompt that is sent to the LLM."). As to the task prompt defining the set of task-related features and being supplied in a session distinct from that of the first prompt, see Hawes paragraph [0060] ("It may be possible for the user to create a multiple-prompt "dialogue" with the LLM where a first task is addressed at a first step, a second task is addressed at a second step, and so forth. Such multi-step dialogue can serve as a kind of fine-tuning of responses from the LLM, where each step gets closer to a desired or target response."). As to the two prompts being associated with respective models, see Hawes paragraph [0042] ("In some implementations, the example computing environment may include multiple large language models 140, which may be included internal and/or external to the model communication system 120, in any combination.").)
It would have been obvious to one of ordinary skill in the art at the time the invention was made before the effective filing date of the claimed invention to further modify Dhillon in view of Li so that the formatting operation and the relevance-filtering operation are each directed by a prompt obtained from a user through an interactive graphical user interface session associated with the respective model, as taught by Hawes. The motivation to do so is expressly supplied by Hawes paragraph [0026], which describes the disclosed system as "for providing improved (including more accurate and more efficient) interactions with LLMs" and as including "various processes, functionality, and interactive graphical user interfaces related to the system," and by Hawes paragraph [0038], which discloses that the system "may further include one or more Large Language Models ("LLMs") to allow for more flexible interaction, such as through receiving natural language instructions, with a user of the system." One of ordinary skill applying machine learning models to the structuring and relevance-filtering operations of Dhillon in view of Li would have been led to obtain the instructions governing those operations from the user through an interface, rather than fixing them in advance, in order to obtain the flexibility and improved accuracy Hawes describes. Applying Hawes' prompt-selection interface separately to each of the two models is no more than the use of a known technique to improve similar devices in the same way, and Hawes paragraph [0060] expressly contemplates addressing a first task at a first step and a second task at a second step, yielding the predictable result of each model being directed by its own user-supplied prompt.
With respect to claim 2 (Original), Dhillon further teaches wherein the structured data object is regenerated at a defined time interval (Dhillon discloses that the ingested data is continually reprocessed and updated on a recurring basis. See Dhillon paragraph [0039], which discloses that the data sources and the data provided by those sources may be dynamic, that additional data sources may continually be added and data from those sources may be continually processed and updated, and that the system may continuously or intermittently request data pertaining to a patient.), and initiating the rendering of the contextualized task-specific graphical visualization comprises: receiving, at a particular time point and via the user interface, a task query for an entity that identifies the defined domain task and the entity corresponding to the structured data object (Dhillon discloses receiving, through the interface, a request that identifies both the condition being queried and the patient to whom the records pertain. See Dhillon paragraph [0059], which discloses that the system receives a request from a third party via the network interface, and paragraph [0060], which discloses that the system may be configured to determine the requestor, the patient, and any specific conditions that are being queried. The queried condition constitutes the recited defined domain task and the patient constitutes the recited entity corresponding to the structured data object.)
and responsive to the task query, receiving the structured data object corresponding to the particular time point, generating the task-specific data object based on the structured data object, and initiating the rendering of the contextualized task-specific graphical visualization based on the task-specific data object (Dhillon discloses that, upon receipt of the request, the system retrieves the stored records, selects from them the data pertinent to the request, and builds and displays the index and timeline from the data so selected. See Dhillon paragraphs [0062]-[0063] and paragraph [0064] ("System 100 may also build and display (step 560) the timeline as exemplified in FIG. 8."). Dhillon further discloses that the timeline so built is generated from the records as they stand at the time of the request. See Dhillon paragraph [0064] ("The timeline may also be dynamically generated from the records based on the request, the requestor, permissions of the requestor, the patient, and the condition.").)
With respect to claim 3 (Currently Amended), Dhillon further teaches wherein generating the task-specific data object further comprises: filtering the plurality of task-agnostic features of the structured data object based on a set of duplication rules that define a multi-tiered time-based constraint for the plurality of third-party data elements (As to the set of duplication rules, see Dhillon paragraph [0045], which discloses that the system may be further configured to compare received records and delete duplicate records, and that when a request to view the medical records of a patient is received the system may automatically filter and discard repeated content. As to the multi-tiered time-based constraint, see Dhillon paragraph [0053], which discloses that the system may be configured to determine the date the record was created, updated, recorded, and any other dates pertaining to the record, such as previous consultations or future scheduled appointments, and that the date stamp may also include dates pertaining to any condition. Applying the duplication rules of Dhillon paragraph [0045] across the several temporal tiers disclosed at Dhillon paragraph [0053] constitutes the recited multi-tiered time-based constraint.)
With respect to claim 4 (Currently Amended), Dhillon further teaches wherein generating the task-specific data object further comprises: filtering the plurality of task-agnostic features of the structured data object based on a set of data quality rules that define a quality constraint for the third-party data source (See Dhillon paragraph [0040] (input filter 210 processes the data provided by each data source before storing it, including recognizing the source and type of data and reorganizing the data as related to pre-determined categories or classifications) and paragraph [0056], which discloses that a medical specialist associated with the system may analyze the received records and provide a ranking of the records indicating the perceived level of importance of the records based on the relevant patient condition, with some records ranked as critical and others as tangential. The source-based ranking so disclosed constitutes a data quality rule defining a quality constraint for the third-party data source.)
With respect to claim 5 (Currently Amended), Dhillon further teaches (ii) the task-specific data element comprises a correlation indicator that maps the task-specific feature to the third-party data source (See Dhillon paragraph [0043], which discloses that the index generator and timeline generator may also provide references and direct access to the full version of the records through interactive hyperlinks in the portions of the timeline associated with each health event, and paragraph [0064] ("The timeline may include hyperlinks, allowing quick access to the records."). The hyperlink linking a displayed element to the underlying source record constitutes the recited correlation indicator.)
Dhillon in view of Hawes is silent to disclose (i) the task-specific data object comprises a task-specific feature from the plurality of task-agnostic features; however, in an analogous art, Li teaches this limitation, as set forth in the mapping of claim 1 above. See Li paragraph [0025] ("An "entry" is a sentence from the notes that is determined to be both relevant and related to the facet."). The retained entries are the recited task-specific features drawn from the broader pool of candidate sentences constituting the task-agnostic features. The motivation to combine Dhillon with Li set forth with respect to claim 1 applies equally here.
With respect to claim 6 (Currently Amended), the claim recites four alternatives joined by "or," and disclosure of any one alternative is sufficient. Dhillon further teaches wherein the task-specific data element comprises a task-specific feature value, a feature time point, the correlation indicator, or a relevancy indicator (As to the correlation indicator alternative, see Dhillon paragraphs [0043] and [0064] as set forth for claim 5. As to the feature time point alternative, see Dhillon paragraph [0053] (the system determines the date the record was created, updated, recorded, and any other dates pertaining to the record) and paragraph [0064] (the timeline provides a chronological listing of data extracted from the records).)
With respect to claim 7 (Currently Amended), Dhillon further teaches:
(b) the set of interactive graphical elements corresponds to the task-specific data element, and an interactive graphical element corresponding to the task-specific data element: (i) is positioned within an interactive timeline chart based on the feature time point (Dhillon discloses that each displayed element corresponds to an extracted record item and is placed along the timeline according to the date associated with that item. See Dhillon paragraph [0043], which discloses that the timeline generator may display data from the records associated with health events in a chronological order in the form of the timeline, and paragraph [0064] ("The timeline may provide a chronological listing of data extracted from the records."). As to the date by which each element is positioned, see Dhillon paragraph [0053], which discloses that the system may be configured to determine the date the record was created, updated, recorded, and any other dates pertaining to the record.)
(ii) presents the task-specific feature value (Dhillon discloses that the substantive content of each record item is displayed in connection with its timeline element. See Dhillon paragraph [0065], which discloses that the data extracted from the records and displayed in connection with each timeline element includes an event date, an event name, an event description, a reference number, and an image URL.)
(iii) comprises interactive presentation features that are presented responsive to a user selection or a detected selection intent of the interactive graphical element (Dhillon discloses interface features that are surfaced upon user selection of a displayed element. See Dhillon paragraph [0064] ("The timeline may include hyperlinks, allowing quick access to the records.") and paragraph [0070], which discloses interactive option menus 1001-1003 presenting additional information and editing functionality after receiving the user's selection of the option menu. See also Dhillon paragraph [0066], which discloses that the user interfaces may include tabs to enable access to different aspects of the patient's collective records.)
Dhillon in view of Hawes is silent to disclose (a) the task-specific data element comprises a task-specific feature value and a feature time point; however, in an analogous art, Li teaches this limitation. As to the task-specific feature value, Li teaches that each entry is "a sentence from the notes that is determined to be both relevant and related to the facet" (Li paragraph [0025]), the sentence constituting the substantive value of the element, and as to the feature time point, Li teaches that the entries are "arranged by "facet" through time" (Li paragraph [0023]). The motivation to combine Dhillon with Li set forth with respect to claim 1 applies equally here.
With respect to claim 8 (Currently Amended), Dhillon further teaches wherein the correlation indicator is an interactive presentation feature and comprises an interactive link that routes an entity from the user interface to the third-party data source (See Dhillon paragraph [0043], which discloses that the index generator and timeline generator may provide references and direct access to the full version of the records through interactive hyperlinks in the portions of the timeline associated with each health event, and paragraph [0069], which discloses that the records may be linked from the timeline 801 to provide an expert with additional information when desired.)
With respect to claim 9 (Original), Dhillon in view of Hawes is silent to disclose wherein the structured data object comprises a tabular format associated with respective timepoints for respective data elements of the structured data object however, in an analogous art, Li teaches this limitation. See Li paragraph [0023] ("A "journal" may be understood to contain "entries" arranged by "facet" through time.") and paragraph [0027] ("A top portion 202 of the screenshot 200 includes a journal summary. The journal summary includes indications of clinical facets discovered by the model 106 and a summary of a number of entries per facet through time."). The arrangement of entries by facet along one dimension and by time along another constitutes the recited tabular format associated with respective timepoints. See also Li paragraph [0029] ("A lower portion 204 of the screenshot 200 may be considered to allow the user to examine details of each event."). The motivation to combine Dhillon with Li set forth with respect to claim 1 applies equally here.
With respect to claim 10 (Currently Amended), Dhillon in view of Hawes is silent to disclose the recited relevancy scoring; however, in an analogous art, Li teaches:
wherein the task-specific data object is generated by: generating, using the machine learning relevancy model, a plurality of relevancy scores for the plurality of task-agnostic features based on a semantic comparison between the plurality of task-agnostic features and the set of task-related features defined by the task-specific prompt (Li discloses that the model computes, for each candidate sentence, a measure of semantic proximity between that sentence and the facet under consideration. See Li paragraph [0040] ("The difficulty of determining the facet may be resolved, in computer science, at inference time with standard techniques. For one example, an unknown similarity and distance matrix property for a given facet may be determined using cosine distance as a runtime parameter."). The cosine distance so computed constitutes the recited relevancy score, and it is computed between representations of the candidate sentences and of the facet in the embedding space produced by the model, which constitutes the recited semantic comparison. See also Li paragraph [0058] ("For illustration purposes, the known BERT model is used in the following for the model 106 of FIG. 1.").)
and identifying a task-specific feature from the plurality of task-agnostic features based on the plurality of relevancy scores (Li discloses that the sentences retained as entries are those the model determines to be relevant to the facet, the remainder being excluded. See Li paragraph [0025] ("An "entry" is a sentence from the notes that is determined to be both relevant and related to the facet. A facet is expected to contain multiple entries."). The motivation to combine Dhillon with Li set forth with respect to claim 1 applies equally here.)
With respect to claim 11 (Currently Amended), the claim recites limitations similar to claim 1 in system form, differs only in the recitation of A system comprising: one or more processors; and one or more memories storing processor-executable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising: and is otherwise rejected for the same reasons set forth for claim 1. As to the recited hardware, Dhillon teaches this structure. See Dhillon paragraph [0031] ("System 100 may include one or more central processing units (CPUs, or processors) 120 and system memory 121.") and, in the same paragraph, that the CPUs are "capable of execution sets of instructions stored in a memory (e.g., system memory 121), a cache, or a register".
With respect to claim 12, the claim recites limitations similar to claim 2 and is rejected for the same reasons set forth for claim 2 above.
With respect to claim 13, the claim recites limitations similar to claim 3 and is rejected for the same reasons set forth for claim 3 above.
With respect to claim 14, the claim recites limitations similar to claim 4 and is rejected for the same reasons set forth for claim 4 above.
With respect to claim 15, the claim recites limitations similar to claim 5 and is rejected for the same reasons set forth for claim 5 above.
With respect to claim 16, the claim recites limitations similar to claim 6 and is rejected for the same reasons set forth for claim 6 above.
With respect to claim 17, the claim recites limitations similar to claim 7 and is rejected for the same reasons set forth for claim 7 above.
With respect to claim 18, the claim recites limitations similar to claim 8 and is rejected for the same reasons set forth for claim 8 above.
With respect to claim 19, the claim recites limitations similar to claim 9 and is rejected for the same reasons set forth for claim 9 above.
With respect to claim 20, the claim recites limitations similar to claim 1 in medium form, differs only in the recitation of One or more non-transitory computer-readable media storing processor-executable instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising: and is otherwise rejected for the same reasons set forth for claim 1. As to the recited medium, Dhillon teaches this structure. See Dhillon paragraph [0031] ("System memory 121 may include a tangible and/or non-transitory computer-readable medium, such as a flexible disk, a hard disk, a compact disk read-only memory (CD-ROM), magneto-optical (MO) drive, digital versatile disk random-access memory (DVD-RAM), a solid-state disk (SSD), a flash drive and/or flash memory, processor cache, memory register, or a semiconductor memory.").
Claims 1-20 are rejected, in the alternative to the rejection set forth above, under 35 U.S.C. 103 as being unpatentable over Dhillon et al. (US Pub. No. 2018/0113945, hereinafter "Dhillon" – previously presented) in view of Li et al. (US Pub. No. 2023/0367799, hereinafter "Li" – previously presented), and further in view of Maddigan et al. ("Chat2VIS: Generating Data Visualisations via Natural Language using ChatGPT, Codex and GPT-3 Large Language Models", hereinafter "Maddigan").
With respect to claim 1 (Currently Amended), the mappings of the preamble and of the limitations credited to Dhillon and to Li in the first ground of rejection set forth above are adopted in full, as is the motivation to combine Dhillon with Li.
Dhillon in view of Li is silent to disclose defined by a machine learning formatting prompt that comprises a first prompt data object generated based on a first interactive user interface session associated with the machine learning formatting model and using a task-specific prompt that (i) defines a set of task-related features and (ii) comprises a second prompt data object generated based on a second interactive user interface session associated with the machine learning relevancy model; however, in an analogous art, Maddigan teaches:
a data structure format comprising a set of format features defined by a machine learning formatting prompt that comprises a first prompt data object generated based on a first interactive user interface session associated with the machine learning formatting model (Maddigan discloses a natural language interface through which a user enters free-form text, from which a prompt is constructed and forwarded to a machine learning model, the prompt specifying the format of the structured object the model is to produce. As to the interactive user interface session from which the prompt is generated, see Maddigan section 3 ("A user enters a NL query via a Streamlit NLI app which is an open-source Python framework for web-based dashboards. The query is combined with a prompt script which engineers a suitable prompt for a selected dataset. The prompt is forwarded to selected LLMs, which return a Python script that is subsequently rendered within the Streamlit NLI.") and section 3.1 ("The interface enables users to select a dataset and enter free-form text describing their data visualisation intent."). As to the prompt defining the data structure format and its set of format features, see Maddigan section 3.2, which discloses that the Description Prompt "consists of one entry for each column indicating its data type" and that where a column has fewer than twenty distinct values it "is deemed as a categorical type, and its values are enumerated in the prompt", and further discloses instructing the model "to use a dataframe (a tablular data object) with the name df" [sic].)
filtering a plurality of task-agnostic features of the structured data object using a task-specific prompt that (i) defines a set of task-related features and (ii) comprises a second prompt data object generated based on a second interactive user interface session associated with the machine learning relevancy model (Maddigan discloses a second prompt element, directed to the task the model is to perform, and further discloses that the user supplies successive queries through the interface in order to refine the result. As to the second prompt element, see Maddigan section 3.2, which discloses that the Description Prompt "concludes with an instruction to create a plotting script with the supplied user query", that "The Code Prompt begins with import statements for the required Python packages we would like the LLMs to use", and that "Once formulated, the two prompt elements are amalgamated, with the resulting string submitted to the LLMs via the text completion endpoint API". As to the successive interface sessions, see Maddigan section 3.3 ("To further enhance the visualisations, users can refine their NL query, including requesting alternative chart types, plot colours, labels etc.").)
It would have been obvious to one of ordinary skill in the art at the time the invention was made before the effective filing date of the claimed invention to further modify Dhillon in view of Li so that the formatting operation and the relevance-filtering operation are each directed by a prompt constructed from free-form text a user enters through an interactive interface session, as taught by Maddigan. The motivation to do so is expressly supplied by Maddigan, which describes the resulting system as offering "simpler and more accurate end-to-end solutions than prior approaches" and as presenting "a significant reduction in costs for the development of NLI systems, while attaining greater visualisation inference abilities compared to traditional NLP approaches that use hand-crafted grammar rules and tailored models" (Maddigan, Abstract). Maddigan further teaches that the approach is reliable "even when queries are highly misspecified and underspecified" (Maddigan, Abstract), which supplies a direct reason to obtain the governing instructions from the user at run time rather than fixing them in advance. Applying Maddigan's prompt-construction interface separately to each of the two models of Dhillon in view of Li is no more than the use of a known technique to improve similar devices in the same way, yielding the predictable result of each model being directed by its own user-supplied prompt.
With respect to claims 2-20, the mappings and motivations set forth for those claims in the first ground of rejection above are adopted in full, except that the prompt and the interactive user interface session limitations inherited from claim 1, claim 11 and claim 20 are taught by Maddigan as set forth above. For the avoidance of doubt as to the scope of the combination applied in this alternative ground: wherever a mapping incorporated from the first ground states that the combination is silent as to a limitation thereafter credited to Li, that statement is read, for purposes of this alternative ground, as "Dhillon in view of Maddigan is silent to disclose." No reference other than Dhillon, Li and Maddigan is applied in this alternative ground.
Response to Arguments
Applicant's arguments filed August 17, 2026 have been fully considered. The arguments directed to the objections and to the rejection under 35 U.S.C. 112(b) are persuasive and those grounds are withdrawn as set forth above. The arguments directed to the prior art are persuasive in part, as set forth below, and the rejection has been restructured accordingly. The arguments directed to 35 U.S.C. 101 are not persuasive.
A. Characterization of the interview.
Applicant states at pages 12 and 13 of the Remarks that "Applicant's representative understood the Examiner to agree to these distinctions during the Interview." That characterization is not consistent with the Examiner's Interview Summary of record. As reflected in that summary, the Examiner indicated only that the proposed amendment appeared to overcome the specific combination of references then applied, expressly declined to indicate allowable subject matter, and expressly stated that entry of the amendment would require an updated search directed to the newly added limitations. No agreement as to patentability was reached during the interview.
B. Arguments directed to Amarasingham.
Applicant argues at page 12 of the Remarks that Amarasingham "discloses only that summaries are generated by separately trained NLG models, not by models directed through prompt data objects generated from interactive user interface sessions," and that Amarasingham's summary task "is addressed through model training, not through a prompt input provided via an interactive user interface session associated with a machine learning model."
Applicant's argument is persuasive. The rejection over Dhillon in view of Li and further in view of Amarasingham is withdrawn, and Amarasingham is no longer applied. New grounds of rejection, necessitated by Applicant's amendment, are set forth above over Dhillon in view of Li and further in view of Hawes, and in the alternative over Dhillon in view of Li and further in view of Maddigan. Hawes discloses the prompt limitations Applicant identifies as absent, including a user interface 400 through which a user selects a system prompt carrying instructions for formatting the response (Hawes paragraph [0059]), a separately selected task prompt (Hawes paragraph [0061]), and a multiple-prompt dialogue in which a first task is addressed at a first step and a second task at a second step (Hawes paragraph [0060]).
C. Arguments directed to 35 U.S.C. 101 — Step 2A Prong One.
Applicant argues at page 16 of the Remarks that the claims do not recite a mental process because "[a] human cannot, for example, generate a prompt data object from an interactive user interface session with a machine learning model—an operation that is inherently computational and has no mental-process analog."
This argument is not persuasive because it is directed to an additional element rather than to the recited exception. The abstract idea recited by claims 1, 11 and 20 is the gathering of records, their organization into a structured form, the selection of those records relevant to a task, and the presentation of the result — each of which is observation, evaluation and judgment within the Mental Processes grouping. The recitation that the governing instruction is obtained from a user interface session identifies the source of the instruction and the mechanism by which it is conveyed; it is evaluated as an additional element at Step 2A Prong Two and Step 2B, as set forth above. A claim does not cease to recite a judicial exception merely because one additional element is computer-implemented. Were it otherwise, the recitation of any generic computing component would remove a claim from Prong One, which is contrary to MPEP 2106.04(a).
D. Arguments directed to 35 U.S.C. 101 — Ex parte Desjardins.
Applicant relies at pages 14 and 15 of the Remarks on Ex parte Desjardins, Appeal No. 2024-000567 (ARP Sept. 26, 2025) (precedential), for the propositions that examiners should not evaluate claims at a high level of generality and that categorically excluding AI innovations from patent protection jeopardizes United States leadership in this technology.
The decision is precedential and has been considered. It does not compel a different result here. The claims at issue in Desjardins were directed to training a machine learning model, and the Appeals Review Panel found them eligible because they improved the operation of the machine learning model itself, including by reducing storage requirements, lowering system complexity, and preventing catastrophic forgetting across sequentially trained tasks. The panel characterized the improvement as one "in training the machine learning model itself," and applied Enfish, LLC v. Microsoft Corp., 822 F.3d 1327 (Fed. Cir. 2016), on that basis.
The present claims improve neither model. The machine learning formatting model and the machine learning relevancy model are unchanged, off-the-shelf components which the specification at paragraphs [0055] and [0057] expressly permits to be any LLM, any GPT model, or any other generative machine learning model. What the amendment adds is the source from which each model's instruction is obtained. Desjardins supports eligibility for improvements to machine learning; it does not support eligibility for applications of machine learning that leave the models untouched. That distinction is the one drawn by the Federal Circuit in Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1205 (Fed. Cir. 2025), which controls and which is addressed in the Step 2A Prong 2 analysis above.
The guidance implementing Desjardins further directs that the specification be evaluated for a detailed, non-conclusory disclosure of a technological improvement and that the claim as a whole be confirmed to reflect that improvement. As set forth in subsection E below, the claims as amended do not reflect the improvement the specification describes.
E. Arguments directed to 35 U.S.C. 101 — Step 2A Prong Two.
Applicant argues at pages 17 through 19 of the Remarks that the claims provide an improved data processing pipeline, relying on specification paragraph [0003] (technical challenges within the realm of traditional data ingestion engines, such as time-consuming ingestion of data and resource intensive transformation of data), paragraph [0066], and paragraph [0068] (improved data visualizations for contextually visualizing data stored in disparate data sources).
This argument is not persuasive for two independent reasons.
First, the improvements Applicant identifies are not reflected in the claims. The Amendment replaced the recitation of "one or more third-party data sources" with "a third-party data source" in each of claims 1, 11 and 20, and made the corresponding change in claims 4, 5, 8, 14, 15 and 18. As amended, the claims require that the plurality of third-party data elements be obtained from a single third-party data source. Whatever improvement the specification attributes at paragraph [0050] to the aggregation of data across multiple disparate datasets, and at paragraph [0068] to the contextual visualization of data stored in disparate data sources, the claims as a whole no longer require disparate sources and therefore do not reflect that improvement. An improvement described only in the specification and not captured by the claim cannot supply integration into a practical application.
Second, and independently, the improvement the specification describes at paragraphs [0002]-[0003] is to the efficiency of an admittedly abstract data-organization process and to the presentation of information to human users. The specification identifies no improvement to the operation of the processors, to the user interface, or to either machine learning model. As set forth in the Step 2A Prong 2 analysis above, the additional elements are recited at the highest level of generality and are expressly described in the specification as interchangeable with any conventional alternative.
F. Arguments directed to 35 U.S.C. 101 — Ollnova.
Applicant relies at page 19 of the Remarks on Ollnova Technologies for the proposition that claims are patent-eligible where they recite a technological solution to a technological problem. Ollnova Technologies Ltd. v. ecobee Technologies ULC, Nos. 25-1045, 25-1046 (Fed. Cir. June 4, 2026), concerned patents directed to wireless communications for building automation systems. The eligibility determinations affirmed there rested on the specific dual-network architecture recited by those claims and on evidence that the claimed modes of wireless network control prevented loss of functionality during communication failure. Nothing in that decision addresses the application of generic machine learning models to a new data environment, and nothing in it displaces Recentive, which does.
G. Arguments directed to 35 U.S.C. 101 — Step 2B.
Applicant argues at page 20 of the Remarks that "the prior art rejections are overcome in light of the claim amendments" and that the rejection is therefore improper.
This argument is not persuasive. Eligibility under 35 U.S.C. 101 and novelty and non-obviousness under 35 U.S.C. 102 and 103 are separate inquiries, and the novelty of an abstract idea does not render it eligible. See MPEP 2106.05, subsection I. Applicant has not identified any additional element, considered individually or as an ordered combination, said to amount to significantly more than the judicial exception. The Berkheimer showing set forth in the Step 2B analysis above rests on Applicant's own specification, which states at paragraphs [0056] and [0058] that the prompt is configured for communication via a network, an API, a machine learning model plug-in, and/or another type of interface, and at paragraphs [0055] and [0057] that each model may be any LLM or GPT model. Elements the specification describes as interchangeable with any conventional alternative do not supply an inventive concept.
H. Arguments directed to the dependent claims.
Applicant argues at page 13 of the Remarks that the dependent claims are patentable by virtue of their dependency from the independent claims. Because the independent claims are not allowable for the reasons set forth above, the dependent claims are not allowable by virtue of their dependency. Applicant has not identified any particular additional feature of any dependent claim said to distinguish over the applied art, and accordingly no further response is possible.
Conclusion
The prior art made of record and not relied upon is considered pertinent to Applicant’s disclosure: Dingliwal et al. (US Pub. No. 2025/0005298) discloses large language models fine-tuned to generate summaries of doctor-patient conversations together with evidence mappings, each evidence mapping indicating, for a sentence of the generated summary, the sentence of the source transcript that provides evidence for it, and further discloses instruction prompts that request creation of an annotated summary and that may specify keywords the generated summary should include (see paragraphs [0019], [0024] and [0026]). THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ANIBAL RIVERACRUZ whose telephone number is (571)270-1200. The examiner can normally be reached Monday-Friday 9:30 AM-6:00 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, Hyung S Sough can be reached at 5712726799. 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.
/ANIBAL RIVERACRUZ/Primary Examiner, Art Unit 2192
Prosecution Insights