DETAILED ACTION
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-4, 8, 10, 11-14, 18, and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Xiong et al (NPL: TA-Student VQA: Multi-Agents Training by Self-Questioning) in view of Latapie et al (US 20210390423).
Regarding claim 1, Xiong discloses a method comprising:
identifying, by a student agent, a topic of interest (Pg. 10064 Section 3.1 TA model is responsible for raising questions (QT ) for the given image Img, wherein image Img is interpreted as the topic of interest);
issuing, by the student agent, a set of one or more questions to a teacher agent regarding the topic of interest (Pg. 10064 Section 3.2 The Question Generating Agent (Qg) acts as a TA, which is designed to generate a set of questions with diversity in format and content under the condition that they are related to the given image Img);
receiving, at the student agent and from the teacher agent, answer data in response to the set of one or more questions (Pg. 10065 Section 3.3 Two Visual Question Answering Agents (Agts) act as two students to answer the questions QT generated by the TA, which is Qg. Their outputs, At and AT, are the results from two heterogeneous-structured Visual Question Answering models (the details of the two models will be provided in Section 3.5); and
using, by the student agent, the answer data to generate a metamodel (Pg. 10067 Section 4.2 we use VIS CNN and DMN as our Question-Answering Agents (Agt) because they obtain a heterogeneous structure).
Xiong fails to teach where Latapie teaches a student agent that is a first reasoning system metamodel agent (¶137 one or more DFRE agents 404, as shown in FIG. 6; ¶143 At the core of the techniques herein is a knowledge representation metamodel 700 for different levels of abstraction, as shown in FIG. 7 including troubleshooting agent 702), and where the teacher agent is a second reasoning system metamodel agent (¶137 one or more DFRE agents 404, as shown in FIG. 6; ¶143 At the core of the techniques herein is a knowledge representation metamodel 700 for different levels of abstraction, as shown in FIG. 7 including troubleshooting agent 702); and
using, by the student agent, the answer data to generate a neuro-symbolic metamodel (¶46 DFRE process 248 may employ deep learning, in some embodiments. Generally, deep learning is a subset of machine learning that employs ANNs with multiple layers, with a given layer extracting features or transforming the outputs of the prior layer; ¶48 & Fig. 3 hierarchy 300 for a deep fusion reasoning engine (DFRE); ¶58 output 314 may comprise a video feed/stream augmented with inferences or conclusions made by the DFRE, such as the locations of unstocked or under-stocked shelves, etc.) that comprises a semantic reasoner (¶49 a reasoning engine, also known as a ‘semantic reasoner,’ ‘reasoner,’ or ‘rules engine,’ is a specialized form of machine learning software that uses asserted facts or axioms to infer consequences, logically; DFRE is an enhanced form of reasoning engine that further leverages the power of sub-symbolic machine learning techniques, such as neural networks (e.g., deep learning), allowing the system to operate across the full spectrum of sub-symbolic data all the way to the symbolic level) and a plurality of layers ranging from a sub-symbolic layer to a symbolic layer (Fig. 3 FIG. 3 illustrates an example hierarchy 300 for a deep fusion reasoning engine (DFRE) & ¶50-51 At the lowest layer of hierarchy 300 is sub-symbolic layer 302 that processes the sensor data 312 collected from the network; At the opposing end of hierarchy 300 may be symbolic layer 306 that may leverage symbolic learning), wherein using the answer data to generate the neuro- symbolic metamodel comprises populating a knowledge graph (¶70 model may be represented as a DFRE knowledge graph and/or network simulation; ¶75-86 DFRE agent 404 may feed and interact with the AIKR reasoner so as to populate and leverage a DFRE knowledge graph with knowledge) that represents cross-layer mappings as relations between nodes of the sub-symbolic layer corresponding to learned features and nodes of the symbolic layer corresponding to concepts (¶48 FIG. 3 illustrates an example hierarchy 300 for a deep fusion reasoning engine (DFRE); ¶50 sub-symbolic layer 302 may perform sensor fusion on sensor data 312 to identify hidden relationships between the data; ¶51 At the opposing end of hierarchy 300 may be symbolic layer 306 that may leverage symbolic learning; ¶52 Symbolic learning models what are referred to as “concepts,”; ¶53 Linking sub-symbolic layer 302 and symbolic layer 306 may be conceptual layer 304 that leverages conceptual spaces. In general, conceptual spaces are a proposed framework for knowledge representation by a cognitive system on the conceptual level that provides a natural way of representing similarities. Conceptual spaces enable the interaction between different type of data representations as an intermediate level between sub-symbolic and symbolic representations; ¶55 a conceptual space is built up from geometrical representations based on a number of quality dimensions that complements the symbolic and deep learning models of symbolic layer 306 and sub-symbolic layer 302, representing an operational bridge between them; ¶83-86 The Knowledge graph also allows different reasoners to: Have their internal subgraphs; Share or coalesce knowledge; Work cooperatively), and wherein the semantic reasoner operates on the symbolic layer using the relations (¶58 the reasoning logic in symbolic layer 306 may be non-axiomatic and constructed around the assumption of insufficient knowledge and resources (AIKR). It may be implemented, for example, with a Non-Axiomatic Reasoning System (open-NARS) 310. However, other reasoning engines can also be used, such as Auto-catalytic Endogenous Reflective Architecture (AERA), OpenCog, and the like, in symbolic layer 306, in further embodiments. Even Prolog may be suitable, in some cases, to implement a reasoning engine in symbolic layer 306).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of a student agent that is a first reasoning system metamodel agent and where the teacher agent is a second reasoning system metamodel agent, and using, by the student agent, the answer data to generate a neuro-symbolic metamodel that comprises a semantic reasoner and a plurality of layers ranging from a sub-symbolic layer to a symbolic layer, wherein using the answer data to generate the neuro-symbolic metamodel comprises populating a knowledge graph that represents cross-layer mappings as relations between nodes of the sub-symbolic layer corresponding to learned features and nodes of the symbolic layer corresponding to concepts, wherein the semantic reasoner operates on the symbolic layer using the relations from Latapie into the method as disclosed by Xiong. The motivation for doing this is to improve reasoning engines.
Regarding claim 2, the combination of Xiong and Latapie teach the method as in claim 1, wherein the student agent uses natural language processing to identify the topic of interest (Latapie ¶166 the DFRE may apply semantic analysis to the feature labels, such as by using natural language processing (NLP) analysis of the feature labels and/or graph analysis and ontology matching of Yang models with other ontologies like the ConceptNet ontology and/or other ontologies,).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein the student agent uses natural language processing to identify the topic of interest from Latapie into the method as disclosed by the combination of Xiong. The motivation for doing this is to improve reasoning engines.
Regarding claim 3, the combination of Xiong and Latapie teach the method as in claim 1, wherein the answer data comprises images (Latapie ¶58 For example, output 314 may comprise a video feed/stream augmented with inferences or conclusions made by the DFRE, such as the locations of unstocked or under-stocked shelves, etc.).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein the answer data comprises images from Latapie into the method as disclosed by Xiong. The motivation for doing this is to improve reasoning engines.
Regarding claim 4, the combination of Xiong and Latapie teach the method as in claim 1, wherein the relations in the knowledge graph include symmetric and anti-symmetric relations and link the sub-symbolic layer to the symbolic layer (Latapie ¶75 DFRE agent 404 may feed and interact with the AIKR reasoner so as to populate and leverage a DFRE knowledge graph with knowledge; ¶77 For example, DFRE agent 404 may perform semantic graph decomposition on DFRE KB 416 (e.g., a knowledge graph), so as to compute a graph from the knowledge graph of KB 416 that addresses a particular problem; ¶118 Feeding and interacting with the AIKR reasoner via bidirectional translation layer to the DFRE knowledge graph; ¶143 The overall structure of this knowledge is also based on anti-symmetric and symmetric relations).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein the relations in the knowledge graph include symmetric and anti-symmetric relations and link the sub-symbolic layer to the symbolic layer89 from Latapie into the method as disclosed by the combination of Xiong. The motivation for doing this is to improve reasoning engines.
Regarding claim 8, the combination of Xiong and Latapie teach the method as in claim 1, wherein the topic of interest comprises a particular type of action associated with a particular type of object (Xiong Fig. 5 showing questions associated with a particular type of object).
Regarding claim 10, the combination of Xiong and Latapie teach the method as in claim 1, wherein the neuro-symbolic metamodel is used to analyze video data captured from at least one of: a port, a train station, a bus station, an airport, or a stadium (Xiong Fig. 4 and Fig. 5 shows examples of images which include a stadium).
Regarding claim(s) 11-14 and 18 (drawn to an apparatus):
The rejection/proposed combination of Xiong and Latapie, explained in the rejection of method claim(s) 1-4 and 8, anticipates/renders obvious the steps of the apparatus of claim(s) 11, 13, 15 and 18-19 because these steps occur in the operation of the proposed combination as discussed above. Thus, the arguments similar to that presented above for claim(s) 1-4 and 8 is/are equally applicable to claim(s) 11-14 and 18. See further Latapie ¶40-43 & ¶230.
Regarding claim(s) 20 (drawn to a CRM):
The rejection/proposed combination of Xiong and Latapie, explained in the rejection of method claim(s) 1, anticipates/renders obvious the steps of the computer readable medium of claim(s) 20 because these steps occur in the operation of the proposed combination as discussed above. Thus, the arguments similar to that presented above for claim(s) 1 is/are equally applicable to claim(s) 20. See further Latapie ¶40-43 & ¶230.
Claim(s) 5, 9, 15 and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Xiong and Latapie as applied to claim 1 and 11 above, and further in view of Latapie et al (US 20210042532, hereinafter Latapie2).
Regarding claim 5, the combination of Xiong and Latapie teach the method as in claim 1, but fails to teach where Latapie2 teaches wherein using the answer data to generate the neuro-symbolic metamodel (Latapie2 ¶55 assessing video feeds/streams using a hybrid neuro-symbolic system) comprises: performing semantic segmentation and object detection on the answer data (Latapie2 ¶82-84 Such object detection and tracking can be achieved, in some embodiments, by training a convolutional neural network (CNN)-based classifier to identify the predefined object classes, such as people, cars, etc. and tracking using, for example ;semantic or instance segmenter).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein using the answer data to generate the neuro-symbolic metamodel comprises: performing semantic segmentation and object detection on the answer data from Latapie2 into the system as disclosed by Xiong and Latapie. The motivation for doing this is to improve reasoning engines.
Regarding claim 9, the combination of Xiong and Latapie teach the method as in claim 1, but fails to teach where Latapie2 teaches wherein using the answer data to generate the neuro-symbolic metamodel (Latapie2 ¶55 assessing video feeds/streams using a hybrid neuro-symbolic system) comprises: training a neural network at the sub-symbolic layer of the neuro-symbolic metamodel using the answer data (Latapie2 ¶126 For example, training a neural network on gallon jugs of milk will enable it to identify only gallon jugs of milk. However, by linking the sub-symbolic processing to a symbolic layer, the system can ‘learn’ the concept of a jug and identify other jugs of different shapes and sizes).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein using the answer data to generate the neuro-symbolic metamodel comprises: training a neural network at the sub-symbolic layer of the neuro-symbolic metamodel using the answer data from Latapie2 into the system as disclosed by Xiong and Latapie. The motivation for doing this is to improve reasoning engines.
Regarding claim(s) 15 and 19 (drawn to an apparatus):
The rejection/proposed combination of Xiong, Latapie, and Latapie2 explained in the rejection of method claim(s) 5 and 9 anticipates/renders obvious the steps of the apparatus of claim(s) 15 and 19 because these steps occur in the operation of the proposed combination as discussed above. Thus, the arguments similar to that presented above for claim(s) 5 and 9 is/are equally applicable to claim(s) 15 and 19. See further Latapie ¶40-43 & ¶230.
Claim(s) 6-7 and 16-17 is/are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Xiong and Latapie as applied to claim 1 and 11 above, and further in view of Chatterjee et al (US 20210256966).
Regarding claim 6, the combination of Xiong and Latapie teach the method as in claim 1, but fails to teach where Chatterjee teaches wherein the teacher agent bases the answer data on results from a search engine (¶21 the intelligent systems 107 may include, but not limited to, a chat bot, question answering systems, virtual assistance, crowdsourcing platforms and search engines.).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein the teacher agent bases the answer data on results from a search engine from Chatterjee into the system as disclosed by the combination of Xiong and Latapie. The motivation for doing this is to improve natural language based intelligent systems.
Regarding claim 7, the combination of Xiong and Latapie teach the method as in claim 1, but fails to teach where Chaterjee teaches wherein the teacher agent bases the answer data on crowdsourced or human-provided information (Chaterjee ¶21 the intelligent systems 107 may include, but not limited to, a chat bot, question answering systems, virtual assistance, crowdsourcing platforms and search engines.).
Therefore, it would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to have implemented the teaching of wherein the teacher agent bases the answer data on crowdsourced or human-provided information from Chatterjee into the system as disclosed by the combination of Xiong and Latapie. The motivation for doing this is to improve natural language based intelligent systems.
Regarding claim(s) 16-17 (drawn to an apparatus):
The rejection/proposed combination of Xiong, Latapie, and Chatterjee, explained in the rejection of method claim(s) 6-7, anticipates/renders obvious the steps of the apparatus of claim(s) 16-17 because these steps occur in the operation of the proposed combination as discussed above. Thus, the arguments similar to that presented above for claim(s) 6-7 is/are equally applicable to claim(s) 16-17. See further Latapie ¶36-38 & ¶128.
Response to Arguments
Applicant's arguments filed 12/11/2025 have been fully considered but they are not persuasive.
Regarding claim 1, the applicant argues that Latapie does not teach “using, by the student agent, the answer data to generate a neuro-symbolic metamodel that comprises a semantic reasoner and a plurality of layers ranging from a sub-symbolic layer to a symbolic layer, wherein using the answer data to generate the neuro-symbolic metamodel comprises populating a knowledge graph that represents cross-layer mappings as relations between nodes of the sub-symbolic layer corresponding to learned features and nodes of the symbolic layer corresponding to concepts, wherein the semantic reasoner operates on the symbolic layer using the relations”.
Regarding the above argument, the examiner respectfully disagrees. Latapie expressly discloses a layered DFRE architecture including sub-symbolic layer (e.g. machine learning models for analyzing time series and detecting structural breaks) and a symbolic layer including concepts and prior knowledge (see Fig. 3 & ¶50-52). Latapie further discloses that observations derived from linear and/or non-linear models are used to create a causal model based on both observations and prior knowledge, and that this causal model “may be represented as a DFRE knowledge graph” (¶70 & ¶75-77). The knowledge graph therefore structurally integrates the model-derived outputs (sub-symbolic features such as detected structural breaks) with symbolic constructs, with relations encoding the casual and contextual relationship among them.
Latapie additionally discloses that the symbolic layer includes reasoning engine configured to operate over the knowledge representation (¶58) and that the DFRE framework provides contextual parameters, appropriates subsets of time series, and initial casual hypotheses within which reasoning is performed (¶60). Because the knowledge graph represents relations derived from machine-learned observations and symbolic prior knowledge, the reasoning engine necessarily operates on relations thar reflect cross-layer integration. Applicant’s arguments that Latapie does not disclose “an explicit graph schema” in which learned features are labeled as graph nodes linked to concept nodes improperly imposes a level of specificity not required by the claims. The claims require relations representing cross-layer mappings, not any particular naming convention or graphical annotation. Latapie’s teaching of representing a casual model, built from model-derived observations and prior knowledge, as a knowledge graph stratifies this limitation.
With respect to the use of answer data, Xiong discloses a student agent that issues questions to a teacher (question-generating) agent and receives answer data in response, thereby refining knowledge representations through iterative interaction (Xiong Sections 3.2-3.3). Latapie teaches incorporating machine-derived observations into a knowledge graph within a layered neurosymbolic architecture. It would have been obvious to one with ordinary skill in the art to use the answer data generated by Xiong’s student-teacher interaction as observational input to the knowledge graph population mechanism of Latapie, as Latapie expressly contemplates constructing its knowledge graph from observations produced by analytic models (¶60).
Accordingly, Latapie teaches a neurosymbolic metamodel comprising sub-symbolic and symbolic layers, population of knowledge graph from model-derived observations and prior knowledge, and operation of a semantic reasoner over relations encoded in that graph, while Xiong teaches student-teacher question-and-answer framework that generates answer data. The combination therefore teaches or renders obvious the argued limitations of claim 1.
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.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KEVIN KY whose telephone number is (571)272-7648. The examiner can normally be reached Monday-Friday 9-5PM.
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, Vincent Rudolph can be reached at 571-272-8243. 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.
/KEVIN KY/Primary Examiner, Art Unit 2671