DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application is being examined under the pre-AIA first to invent provisions.
Response to Amendment
In response to the amendment filed on October 30, 2025:
The abstract is amended.
Claims 1, and 13 are amended.
Claims 1-24 are pending.
Response to Arguments
In response to the remarks filed on October 30, 2025:
a. Objection to the abstract is withdrawn in view of Applicant’s amendment.
b. Applicant’s remarks regarding the 35 U.S.C. 101 rejections of the pending claims have been fully considered but are not persuasive.
At least independent claims 1, and 13 are amended to recite “the modular domain heuristic data structure guarding against hallucinations in the response data.” However, Applicant is noted that this limitation, as currently drafted, describes a functional result or an intended outcome rather than a specific technical process. It is a known mental process of ensuring the accuracy of retrieved information and preventing errors, i.e. hallucinations, that can be performed by a person comparing result data to an answer template. Each claim merely recites the performance of this mental process on a general-purpose computer. Having no explicit definition in the claim language, the recitation of modular domain heuristics is interpreted as a high-level software component without reciting how its internal structure improves the functioning of the computer itself beyond the generic ability to retrieve and compare data. Thus, the claims each fails to recite a practical application per step 2A – prong 2 of the abstract idea analysis. The additional elements in each claim, i.e., event trigger agent, query listener agent, partial information retrieval agent, are recited at a high level of generality as these agents perform well-understood, routine, and conventional (WURC) activities in the field of computer-based Q&A systems. Merely combining these generic components to achieve the functional goal of guarding against hallucinations does not transform the abstract idea into a patent-eligible invention. Each claim is drafted at a relatively high level of generality and is directed to the result or effect of the invention rather than the specific means of achieving it. Thus, the claims also fail per step 2B of the analysis. As such, 35 U.S.C. 101 rejections are maintained for claims 1, 13 and all respective dependent claims.
c. Applicant’s remarks regarding the 35 U.S.C. 103 rejections of the pending claims have been fully considered but are moot in view of a new ground of rejections presented hereon.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
The claimed invention in claims 1-24 are directed to a judicial exception (i.e., an abstract idea) without significantly more.
Claims 1-24 pass step 1 of the 35 U.S.C. 101 analysis since each claim is either directed to a method, or a system comprising a partial information retrieval processor (i.e., a physical component in Figure 2 and [0020] of instant specification).
Claims 1, and 13 recite each, in part, steps that are directed to an abstract idea (“Courts have examined claims that required the use of a computer and still found that the underlying, patent-ineligible invention could be performed via pen and paper or in a person’s mind.” Versata Dev. Group v. SAP Am., Inc., 793 F.3d 1306, 1335, 115 USPQ2d 1681, 1702 (Fed. Cir. 2015)). Each claim recites the limitations of …identifying [the occurrence of] an event; …generating a query in response to the identified event; processing the query…, and generating response data that includes provenance information.” The limitations, as drafted, is a process that, under its broadest reasonable interpretation, covers performance of the limitations in the mind but for the recitation of generic computer components (e.g., mentally or visually identifying an occurring event; mentally generating a query and/or writing down the query on a piece of paper; mentally processing the query; and writing down result data based on the processed query). That is, other than reciting generic components (e.g., partial information retrieval processor, event trigger agent, query listener agent, partial information retrieval agent which are software-implemented), nothing in the claim precludes the limitations from being performed in the human mind and/or with the aid of pen/paper per step 2A – prong 1 of the Abstract Idea Analysis. Thus, the limitations are parts of a mental process.
Each claim further recites an additional element of providing a partial information retrieval processor in communication with a data server which is an extra-solution activity (per step 2A – prong 2 of the Abstract Idea Analysis) that cannot be integrated into a practical application (e.g., the elements recite trivial elements that occurred or would occur after the mental process(s)). Each of the additional limitation(s) is no more than mere instructions to apply the exception using a generic computer component (e.g., processor, memory, and computer-executable instructions).
The extra-solution activity in step 2A - prong 2 are reevaluated in step 2B to determining if each limitation is more than what is well-understood, routine, conventional (WURC) activity in the field. The background of the limitations does not provide any indication that the computer components (e.g., processor, memory, and computer-executable instructions) are not off-the-shelf computer components. The Symantec, TLI, and OOP Techs court decisions cited in MPEP 2106.05(d)(II) indicate that mere receiving, generating, storing, determining, identifying, and transmitting of data over a network are a well-understood, routine, and conventional functions when claimed in a merely generic manner (as it is here). Accordingly, a conclusion that the claims are well-understood, routine, conventional activity is supported under Berkheimer Option 2. For these reasons, there is no inventive concept in each claim, thus, the claims are ineligible.
Applicant is noted that the recitation of “the modular domain heuristic data structure guarding against hallucinations in the response data” describes a functional result or an intended outcome rather than a specific technical process. It is a known mental process of ensuring the accuracy of retrieved information and preventing errors, i.e. hallucinations, that can be performed by a person comparing result data to an answer template. Each claim merely recites the performance of this mental process on a general-purpose computer. Having no explicit definition in the claim language, the recitation of modular domain heuristics is interpreted as a high-level software component without reciting how its internal structure improves the functioning of the computer itself beyond the generic ability to retrieve and compare data. Thus, the claims each fails to recite a practical application per step 2A – prong 2 of the abstract idea analysis. The additional elements in each claim, i.e., event trigger agent, query listener agent, partial information retrieval agent, are recited at a high level of generality as these agents perform well-understood, routine, and conventional (WURC) activities in the field of computer-based Q&A systems. Merely combining these generic components to achieve the functional goal of guarding against hallucinations does not transform the abstract idea into a patent-eligible invention. Each claim is drafted at a relatively high level of generality and is directed to the result or effect of the invention rather than the specific means of achieving it. Thus, the claims also fail per step 2B of the analysis.
Claims 2, and 14 each further recites an additional step of …updating a knowledge base using the response data which can be done mentally and/or with the aid of pen/paper (e.g., writing down new data to a piece of paper or mentally noting data changes in mind). At best, this additional step is an extra-solution activity and/or a WURC activity, similar to the above analysis, of updating a database with new result data. Thus, the claims are ineligible.
Claims 3, and 15 each further recites an additional step of …generating the modular domain heuristic data structure…includes domain knowledge… which can be done mentally and/or with the aid of pen/paper (e.g., writing down the data structure to include certain information to a piece of paper). At best, this additional step is an extra-solution activity and/or a WURC activity, similar to the above analysis, of generating a data structure comprising certain types of data or information. Thus, the claims are ineligible.
Claims 4, and 16 each further recites a definition for the domain heuristic. Thus, the claims are ineligible.
Claims 5, and 17 each further recites a definition for the domain-specific heuristic. Thus, the claims are ineligible.
Claims 6, and 18 each further recites an additional step of …generating at least one human-readable response based on the response data… which can be done mentally and/or with the aid of pen/paper (e.g., writing down the query result on a piece of paper). At best, this additional step is an extra-solution activity and/or a WURC activity, similar to the above analysis, of generating an human-readable query result. Thus, the claims are ineligible.
Claims 7, and 19 each further recites an additional step of …triggering an event… which can be done mentally and/or with the aid of pen/paper (e.g., noting or highlighting the query result on a piece of paper). At best, this additional step is an extra-solution activity and/or a WURC activity, similar to the above analysis, of generating a notification of query result. Thus, the claims are ineligible.
Claims 8, and 20 each further recites a definition for the response comprises a natural language answer. Thus, the claims are ineligible.
Claims 9, and 21 each further recites a definition for a response data structure having at least one data provenance chain. Thus, the claims are ineligible.
Claims 10, and 22 each further recites an additional step of …generating a query data structure including a question type… which can be done mentally and/or with the aid of pen/paper (e.g., writing down the query on a piece of paper with a certain types of data). Thus, the claims are ineligible.
Claims 11, and 23 each further recites an additional step of …generating and displaying a visualization interface…that visualizes the response data which is an extra-solution activity and/or a WURC activity, similar to the above analysis, of generating a GUI to display query result. Thus, the claims are ineligible.
Claims 12, and 24 each further defines the at least one visualization piece which is an extra-solution activity and/or a WURC activity, similar to the above analysis, of generating a GUI with multiple sections to display query result. Thus, the claims are ineligible.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-4, 6-11, 13-16, and 18-23 are rejected under 35 U.S.C. 103 as being unpatentable over Adler et al. (Pub. No. US 2010/0114630, published on May 6, 2010; hereinafter Adler) in view of Bayless et al. (Pub. No. US 2025/0112878, filed on September 29, 2023; hereinafter Bayless).
Regarding claims 1, and 13, Adler clearly shows and discloses a system for partial information retrieval comprising: an information retrieval processor in communication with a data source; an event trigger agent executed by the processor, a query listener agent executed by the processor, and an information retrieval agent executed by the processor to implement a method for information retrieval (Figure 1 and Abstract) comprising:
providing an information retrieval processor in communication with a data source (Figure 8 shows graph event analysis and notification subsystem 830 in communication with enterprise data 860 and/or practice library 840);
identifying the occurrence of an event using an event trigger agent executed by the processor (Graph event listener 820 is a component that listens and reports all the new activity on the provenance graph, [0070]. A graph event may show that an employee performing a read on some financial data is a violation, if the practice library 840 shows that the connection of the employee to financial data as a reader is a violation, [0072]);
generating a query by a query listener agent executed by the processor in response to the identified event (When a new graph event arrives, the graph event analysis and notification subsystem 830 sends a set of queries to the provenance graph 895 via graph query interface 810. The queries are built based on the entries of the practice library 840 to determine if the new graph event generates one of the practices listed in the library, [0072]); and
processing the query in accordance with a modular domain heuristic data structure using an information retrieval agent executed by the processor (As soon as the graph event about employee performing the task is received, the query that seeks for an approver relation between the manager and the task is fired. The result of the query is retrieved by result analyzer 940 and it contains the graph properties sought around the new graph event. If the retrieved result is about unstructured text and requires further processing, text analysis engine 850 is used. Enterprise data 860, such as employee records, is also available to be used for result analysis. Result analyzer 940 has a list of actions to take based on the result of the analysis, [0073]); and
generating response data that includes provenance information (More execution information around the event is collected by sending a query to the graph based on the practices stored in the library (1030). Since practices are stored as sub-graphs, queries can be built to extract the nodes and edges that are used to express the practice. The query returns the actual practice from the provenance graph and it is compared with the practice in the library (1040), [0074]-[0075]).
Bayless then discloses:
the information being partial information (the query engine 204 may determine which portion of the knowledge graph 122 is semantically most relevant, or has answers that are relevant for the formal-language query 128. The prompt-engineering component 208 may then provide the most relevant information along with the formal-language query 128 to the LLM 118 in one or more prompts 132, [0069]); and
the modular domain heuristic data structure (The query engine 204 may utilize various statistical methods to estimate the likelihood of certain patterns occurring in the knowledge graphs 122/202, as well as heuristic approaches to guide the search for relevant information. The heuristics are rules of thumb that can help the query engine 204 prune the search space and focus on potentially relevant parts of the graph, [0066]) guarding against hallucinations in the response data (providing the answer to the user as it is in the knowledge graph, or providing the answer from the knowledge graph to the language model for use in generating an output that will be provided to the user. Retrieval augmented generation (RAG) retrieves data from outside the language model and augments the provided prompts by adding the relevant retrieved data in context. RAG can help reduce model hallucinations by guiding the output to be similar to or based on the retrieved information. Accordingly, information from one or more nodes in the knowledge graph may be retrieved from the knowledge graph and added to the context window of the language model, [0131]).
It would have been obvious to an ordinary person skilled in the art at the time of the invention was effectively filed to incorporate the teachings of Bayless with the teachings of Adler for the purpose of implementing an anti-hallucination and attribution architecture for enterprise generative AI systems to increase the accuracy and reliability of generative artificial intelligence content by detecting, preventing, and mitigating inaccuracy, model overfitting and data bias associated with data sources used to generate AI responses.
Regarding claims 2, and 14, Bayless further discloses updating a knowledge base by a knowledge base updating agent executed by the processor and using the response data (remote answer validators 214 (e.g., external LLMs, external subject matter experts, etc.), and/or remote data sources 216 (e.g., web-based encyclopedias, user forums, etc.), may be used to determine whether the answer 134 is accurate, either modify or remove the answer 134 if it is determined to be inaccurate or confusing. In this example, the knowledge-graph component 120 may determine, or receive input from the other sources, that the answer 134 is inaccurate or confusing, and may perform a graph modification 416 to change, modify, or remove the answer 134 and improve the knowledge graph 122, [0091]).
Regarding claims 3, and 15, Adler further discloses the knowledge base updating agent generates the modular domain heuristic data structure, and the modular domain heuristic data structure includes domain knowledge (The practice library is composed of practices expressed in terms of provenance graph elements. These practices may be categorized as best, recommended, anomalies or violations, [0071]), a domain heuristic (FIG. 2 illustrates a comprehensive, generic data model that can be extended to meet the domain specific needs. As shown, the data of enterprise artifacts stored in the provenance store, depicted as Provenance Record 210, falls into five dimensions or classes including data record, task record, process record, resource record, and custom records, [0041]-[0047]), and at least one partial information retrieval heuristic (These five classes of records represent the nodes of the provenance graph. To define the correlation between two records, Relation Records 260 represent the edges. These are the records generally produced as a result of relation analysis among the collected records. Relations between relation records are possible and such higher degree relation. Some relations are rather basic on the IT (information technology) level, such as the read and write between tasks and data. Other relations are derived from the context, such as that between manager and achieved challenge, [0047]).
Regarding claims 4, and 16, Adler further discloses the domain heuristic comprises a domain-specific heuristic (Custom Records 250: Custom records provide the extension point to capture domain specific, mostly virtual artifacts such as compliance goals, alerts, checkpoints, etc., [0046]).
Regarding claims 6, and 18, Adler further discloses generating at least one human-readable response based on the response data using a response generator agent executed by the processor (the type of violation is identified (1055) and an appropriate action command is fired to the process modification subsystem 875 through notification component 830. Such "warnings" or "alerts" can be sent by the dynamic process-behavior influencing system to a system administrator and/or another automated subsystem, depending on the nature of the violation, [0075]).
Regarding claims 7, and 19, Adler further discloses the response generator agent causes the event trigger agent to trigger an event (Based on the type of violation, different actions are taken: (i) the workflow is stopped and execution steps are taken back to a prior point (1060); (ii) the process continues with warning 1070; (iii) the process continues without warning (1080); (iv) the process stops with warning (1090); or (v) a new process is started (1095), [0075]).
Regarding claims 8, and 20, Bayless further discloses the response comprises a natural language answer (the answer may, at least partially, be a natural-language answer 136 that is presented to the user 106, [0040]).
Regarding claims 9, and 21, Bayless then discloses the response comprises a response data structure having at least one data provenance chain (the provenance data that indicates sources of the answers may be attached to the nodes and/or edges that are included in the expression of the answer. In this way, when an answer is selected for a query, the provenance information indicating source(s) of that answer may be provided to a user 106 to indicate the source of the answers, [0065]).
Regarding claims 10, and 22, Bayless further discloses the query listener agent generates a query data structure including a question type, a question context, and provenance data (The query may be input as a natural-language query 126 in a plain or natural language of a user 106, such as a human language that is used by humans to communicate. At 2, the chatbot interface 114 may receive the natural-language query 126, and the chatbot component 112 may translate the natural-language query 126 into a formal-language query 128. The formal language may be any type of language that is usable to query a knowledge graph 122. For instance, the formal language may be a structured query language (SQL), a SPARQL Protocol and RDF Query Language (SPARQL), a resource description framework (RDF) language, or any other formal query language, [0037]-[0041]).
Regarding claims 11, and 23, Adler further discloses generating and displaying a visualization interface, the visualization interface displaying at least one visualization piece that visualizes the response data (the type of violation is identified (1055) and an appropriate action command is fired to the process modification subsystem 875 through notification component 830. Such "warnings" or "alerts" can be sent by the dynamic process-behavior influencing system to a system administrator and/or another automated subsystem, depending on the nature of the violation, [0075]).
Claims 12, and 24 are rejected under 35 U.S.C. 103 as being unpatentable over Adler in view of Bayless and further in view of Kien (Pub. No. US 2008/0204831, published on August 28, 2008).
Regarding claims 12, and 24, Kien then discloses the at least one visualization piece includes a first section which graphically illustrates an actual value, a second section which graphically illustrates a recommended value, and a difference section which graphically illustrates a difference between the actual value and the recommended value (Figure 4A a table with 3 columns, each column represents an output value, an ideal output value, and a difference output value for each row).
It would have been obvious to an ordinary person skilled in the art at the time of the invention was effectively filed to incorporate the teachings of Kien with the teachings of Adler, as modified by Bayless, for the purpose of visualizing result data such that difference between actual and ideal values are readily available for enhancing result analysis.
Allowable Subject Matter
Claims 5, and 17 are objected for being dependent on a base rejected claim but would be allowable over the prior art if rewritten in independent form to incorporate the limitations of the base claim and all intervening claim(s).
Relevant Prior Art
The following references are considered relevant to the claims:
Siebel et al. (Pub. No. US 2024/0370709) teaches the anti-hallucination and attribution module has generated and quantified the relationship among the response and source segments. All the different scores that define the relationships among the response and source segments can in turn be used to quantify a singular score associated with each response segment (e.g., quantifying its groundedness in sources) and a singular score associated with each source segment (e.g., in the context of the corresponding response segment). This process can accommodate user-defined heuristics for computing some or all of these scores and for computing the source credibility score based on the available metadata of the source, either during inference or upon ingestion of the sources.
Choi et al. (Pub. No. US 2025/0117583) teaches a model-agnostic natural language generation method comprises providing an information-containing plain text prompt, generating tokens based on the information-containing plain text prompt on an automated basis, re-ordering the importance of each generated token and selecting the token with the highest knowledge groundedness in each scenario through a heuristic search process, determining token-level hallucinations with a specifically trained knowledge classifier by identifying the inflection point of hallucination and replacing the hallucinated tokens with further generated tokens, and generating a plain text response with the tokens with the least hallucination and highest knowledge groundedness.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. 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).
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.
Contact Information
Any inquiry concerning this communication or earlier communications from the Examiner should be directed to Son T. Hoang whose telephone number is (571) 270-1752. The Examiner can normally be reached on Monday – Friday (7:00 AM – 4:00 PM).
If attempts to reach the Examiner by telephone are unsuccessful, the Examiner’s supervisor, Sherief Badawi can be reached on (571) 272-9782. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/SON T HOANG/Primary Examiner, Art Unit 2169 December 27, 2025