DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The term “the financial service” in claim 16 lacks antecedent basis as there are no prior reference to a financial service made in the claims.
Claims 17-20 are dependent claims that depend on indefinite claim 16. They are therefore rejected by virtue of its dependency.
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.
Step 1
According to the first part of the analysis, claims 1-10 are directed to a system and claims 11-20 are directed to a method, which fall within one of the four statutory categories (i.e. process, machine, manufacture, or composition of matter).
Regarding Claim 1
Step2A Prong One
a plurality of entity nodes, each of the entity nodes … having an entity type, entity characteristics
(This element recites entity nodes with characteristics [conceptual data organization] which is to be understood as a recitation of a mental process.)
and one or more functional connections, each functional connection associating one a first entity node of the plurality of entity nodes to another a second entity node of the plurality of entity nodes,
(This element recites connections between entity nodes [conceptual data organization] which is to be understood as a recitation of a mental process.)
Step 2A Prong Two
each of the entity nodes including a hardware-based neuromorphic device
(This step for implementing the entity nodes using a hardware-based neuromorphic device does integrate the judicial exception into a practical application, "With the availability of the neuromorphic chips the implementation of knowledge graphs can be improved dramatically" (present application, paragraph [0059]. Thus, claim 1 is not rejected under 101.)
Regarding Claim 11
Step2A Prong One
a plurality of entity nodes, each entity node having an entity type, entity characteristics… and ii) one or more functional connections associating the plurality of entity nodes to others of the plurality of entity nodes,
(This element recites data analysis and organization at a high level of generality such that they could be performed in the human mind, which is to be understood as a mental process)
associating each query component with one or more entities and connections of the knowledge graph:
(This step for associating query components is considered a mental process)
and performing one or more computational operations by each entity node of the plurality of entity nodes.
(This step for performing computational operations is considered a mathematical concept)
Step2A Prong Two
and storage, processing and communication capability
(This step for including entity storage, processing, and communication capabilities at a high level of generality is considered mere instructions to apply an exception. See MPEP § 2106.05(f))
the method comprising: receiving the query at a controller operatively connected to at least one of the plurality of entity nodes;
(This step for receiving a query at a controller is considered insignificant extra-solution activity. See MPEP § 2106.05(g))
decomposing the received query into query components including functional operations;
(This step for decomposing the received query is recited at a high level of generality and does not integrate the exception)
distributing the query components to the one or more associated entities for processing;
(This step for distributing query components is considered insignificant extra-solution activity. See MPEP § 2106.05(g))
generating a response to the received query; and outputting the response through the controller,
(This step for generating and outputting a response is insignificant extra-solution activity. See MPEP § 2106.05(g))
wherein the processing of the received query includes message passing between the plurality of entity nodes by exchanging messages between all of the plurality of entity nodes to calculate connectivity between the plurality of entity nodes,
(This step for calculating connectivity to process a query at a high level of generality is considered insignificant extra-solution activity. See MPEP § 2106.05(g))
Step2B
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because, when considered individually and in combination, they do not add significantly more (also known as inventive concept) to the abstract idea. The claim recites abstract ideas while the additional elements are recited at a high level of generality are considered mere instructions to apply an exception and insignificant extra-solution activity.
Regarding Claim 12
Step2A Prong One
Claim 12 depends from claim 11 and therefore incorporates the abstract idea recited therein
Step2A Prong Two
The method of claim 11 further comprising pooling related entity nodes of the plurality of entity nodes to the associated query component such that the received query is parsed and processed by a subset of the related entity nodes in the knowledge graph.
(This step for including pooling related entities at a high level of generality is considered mere-instructions to apply an exception. See MPEP § 2106.05(f))
Step2B
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because, when considered individually and in combination, they do not add significantly more (also known as inventive concept) to the abstract idea. The claim recites abstract ideas while the additional elements are recited at a high level of generality are considered mere instructions to apply an exception.
Regarding Claim 13
Step2A Prong One
Claim 13 depends from claim 11 and therefore incorporates the abstract idea recited therein
Averaging, arithmetic (This element recites arithmetic operations which is to be understood as the recitation of a mathematical concept)
Pooling, aggregating, concatenating (This element recites manipulating data which is to be understood as the recitation of a mathematical concept)
Ontological operations (This element recites reasoning and inference related operations which is to be understood as the recitation of a mental process)
Risk, credit, compliance (This element recites fundamental economic practices or principles related operations which is to be understood as the recitation of a methods of organizing human activity)
Step2A Prong Two
message passing, routing, convoluting, updating, query operations, storage operations, logical operations, natural-language processing operations.
(This additional element merely lists categories of operations that entity nodes are capable of performing at a high level of generality, which amounts to mere instructions to apply the judicial exception, which does not integrate the exception into a practical application. See MPEP § 2106.05(f))
Step2B
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because, when considered individually and in combination, they do not add significantly more (also known as inventive concept) to the abstract idea. The claim recites the abstract ideas, while the additional element is recited at a high level of generality and is considered mere instructions to apply a judicial exception.
Regarding Claim 14
Step2A Prong One
i) a plurality of entity nodes, each entity node of the plurality of entity nodes of the computational knowledge graph having an entity type, entity characteristics and ii) one or more functional connections associating the plurality of entity nodes to others of the plurality of entity nodes, the method comprising:
(This element recites data analysis and organization at a high level of generality such that they could be performed in the human mind, which is to be understood as a mental process)
whereby each of the plurality of entity nodes accesses the one or more data sources and executes the accessed one or more learning functions; and processing data obtained from the accessed one or more data sources using the accessed one or more learning functions to determine an inference regarding the processed data.
(This element recites machine learning at a high level of generality which is to be understood as a mathematical concept)
Step2A Prong Two
providing each entity node of the plurality of entity nodes with a hardware-based neuromorphic device having a neuronal circuitry element configured to perform a learning function;
(This step for implementing the entity nodes using neuromorphic hardware does integrate the exception into a meaningful application; "With the availability of the neuromorphic chips the implementation of knowledge graphs can be improved dramatically" (present application, paragraph [0059]. Thus, claim 14 is not rejected under 101).
Regarding Claim 16
Step2A Prong One
i) a plurality of entity nodes, each entity node of the plurality of entity nodes having an entity type, entity characteristics… and ii) one or more functional connections associating the plurality of entity nodes to others of the plurality of entity nodes,
(This element recites data analysis and organization at a high level of generality such that they could be performed in the human mind, which is to be understood as a mental process)
configuring the entity nodes to represent financial or legal entities; configuring the one or more functional connections between the plurality of entity nodes that represent relationships or transactions between the plurality of entity nodes;
(This element recites using nodes and connections to represent financial or legal interactions which is to be considered methods of organizing human activity)
Step2A Prong Two
and storage, a processing capability, and a communication capability
(This step for including generic storage, processing, and communication capabilities is considered mere instructions to apply an exception. See MPEP § 2106.05(f))
providing a controller operatively connected to at least one entity node of the plurality of entity nodes to receive a query;
(Receiving a query via a controller is considered insignificant extra-solution activity. See MPEP § 2106.05(g))
embedding each processing capability of each entity node of the plurality of entity nodes to be customized to the financial service;
(This element recites tailoring general processing capabilities to a financial concept, which is merely indicating field of use in which to apply the judicial exception)
generating a response to the received query; and outputting the response through the controller.
(This step for generating and outputting a response is considered insignificant extra-solution activity. See MPEP § 2106.05(g))
Step2B
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because, when considered individually and in combination, they do not add significantly more (also known as inventive concept) to the abstract idea. The claim recites abstract ideas while the additional elements are recited at a high level of generality and are considered insignificant extra-solution activity or mere instructions to apply an exception.
Regarding Claim 17
Step2A Prong One
Claim 17 depends from claim 16 and therefore incorporates the abstract idea recited therein
wherein each of the plurality of entity nodes represents one or more of corporations, clients, products, services and events.
(This element recites using nodes represent specific financial or legal concepts which is to be considered methods of organizing human activity)
Step2A Prong Two
There are no additional elements that integrate the judicial exception into a practical application.
Step2B
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because, when considered individually and in combination, they do not add significantly more (also known as inventive concept) to the abstract idea. The claim recites abstract ideas without any technological improvement or inventive step.
Regarding Claim 18
Step2A Prong One
Claim 18 depends from claim 16 and therefore incorporates the abstract idea recited therein
wherein the connections represent one or more of service contracts, product-specific client relationships, third party services, and financial events.
(This element recites using connections to represent specific financial or legal transactions which is to be considered methods of organizing human activity)
Step2A Prong Two
There are no additional elements that integrate the judicial exception into a practical application.
Step2B
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because, when considered individually and in combination, they do not add significantly more (also known as inventive concept) to the abstract idea. The claim recites abstract ideas without any technological improvement or inventive step.
Regarding Claim 19
Step2A Prong One
Claim 19 depends from claim 16 and therefore incorporates the abstract idea recited therein
Averaging, arithmetic (This element recites arithmetic operations which is to be understood as the recitation of a mathematical concept)
Pooling, aggregating, concatenating (This element recites manipulating data which is to be understood as the recitation of a mathematical concept)
Ontological operations (This element recites reasoning and inference related operations which is to be understood as the recitation of a mental process)
Risk, credit, compliance (This element recites fundamental economic practices or principles related operations which is to be understood as the recitation of a methods of organizing human activity)
Step2A Prong Two
message passing, routing, convoluting, updating, query operations, storage operations, logical operations, natural-language processing operations.
(This additional element merely lists operations that entity nodes are capable of performing at a high level of generality, which amounts to mere instructions to apply the judicial exception, which does not integrate the exception into a practical application. See MPEP § 2106.05(f))
Step2B
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because, when considered individually and in combination, they do not add significantly more (also known as inventive concept) to the abstract idea. The claim recites the mental processes, mathematical processes, and methods of organizing human activity with the additional elements including mere instructions to apply a judicial exception recited at a high level of generality.
Regarding Claim 20
Step2A Prong One
Claim 20 depends from claim 19 and therefore incorporates the abstract idea recited therein
Step2A Prong Two
wherein the processing capability of each entity node of the plurality of entity nodes is customized for compliance related functions including assessing risks and determining compliance with regulations.
(This step for customizing the processing capabilities of the entity nodes at a high level of generality is considered mere instructions to apply an exception. See MPEP § 2106.05(f))
Step2B
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because, when considered individually and in combination, they do not add significantly more (also known as inventive concept) to the abstract idea. The claim recites abstract ideas while the additional element is recited at a high level of generality and is considered mere instructions to apply an exception.
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 is/are rejected under 35 U.S.C. 103 as being unpatentable by Soler Garrido et al., (hereinafter Garrido) (US20220229400, 2022-07-21) in view of Andrew Cassidy et al. (hereinafter Cassidy) (US 20190236444 A1, 2019-08-01) further in view of David Snelling (hereinafter Snelling) (US 20150193636 A1, 2015-07-09) further in view of Svitlana Vakulenko et al. (hereinafter Vakulenko) (“Message Passing for Complex Question Answering over Knowledge Graphs”, 2019-08-19).
Regarding claim 1, Garrido teaches;
A hardware-based computational knowledge graph system comprising: a plurality of entity nodes, … having an entity type,
([0190] For the following experiments, we use a recording from the industrial automation system with some default network and app activity, resulting in a knowledge graph KG with 3529 nodes, 11 node types)
NOTE: Teaches entity nodes of the knowledge graph having a type.
entity characteristics and storage,
([0132] Each entity in the graph is represented by one of the node embedding populations NEP, storing both its embeddings (real-valued entries) and accumulated gradient updates.)
NOTE: Teaches entity storage (storing embeddings) and characteristics (embeddings and gradient updates of the entity node embedding populations can be considered characteristics of the entity).
a processing capability, and a communication capability,
([0057] In an embodiment of the industrial device and method, learning updates for entity embeddings are computed using static feedback connections from each output neuron to neurons of the node embedding populations);
NOTE: Teaches a processing capability (neurons of the entity node embedding populations compute learning updates) and a communication capability of the entity nodes (entity node embeddings use feedback from output neurons, i.e., the entity node embeddings can at least communicate with output neurons).
and one or more functional connections, each functional connection associating one of a first entity node of the plurality of entity nodes to another a second entity node of the plurality of entity nodes,
([0058] In an embodiment of the industrial device and method, the learning component includes first neurons forming a first node embedding population, representing a first entity contained in the triple statements by first spike times of the first neurons during a recurring time interval. The learning component includes second neurons forming a second node embedding population, representing a second entity contained in the triple statements by second spike times of the second neurons during the recurring time interval. A relation between the first entity and the second entity is represented as the differences between the first spike times and the second spike times.)
NOTE: Teaches one or more functional connections, each functional connection associating one of a first entity node of the plurality of entity nodes to a second entity node of the plurality of entity nodes (differences of spike times between a first entity node and second entity node represent a relation / connection, with the difference in spike times indicating the strength of the connection, which can be considered a function of the connection).
Garrido fails to teach but Cassidy teaches
each of the
([0072] After consuming the input files, placement systems according to the present disclosure such as NSCP create an internal representation of the network in the form of a graph. Each vertex of the graph is a neurosynaptic core. The edges of the graph are the number of connections between each pair of cores (in terms of the static number of connections) or the number of spikes between those two cores (representing dynamic activity in the network).)
NOTE: Teaches a graph representing a neural network where each vertex/node of the graph includes a hardware-based neuromorphic device (neurosynaptic core).
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to implement the entity nodes taught by Garrido to each include a respective neurosynaptic core as taught by Cassidy, further explained below.
OBVIOUSNESS TO COMBINE CASSIDY WITH GARRIDO:
Cassidy is analogous art to the present disclosure as it pertains to a plurality of nodes each including neuromorphic hardware-based devices (neurosynaptic cores) while Garrido is analogous art to the present disclosure as it pertains to a plurality of entity nodes of a knowledge graph having at least processing, storage, and communication capabilities.
Garrido teaches neuromorphic hardware for storing a knowledge graph representation where entities are represented by a respective embedding population of neurons, allowing for at least processing, storage, and communication capabilities.
Cassidy supplies implementation details for a neuromorphic substrate implementing a graph representation of a neural network, where the plurality of nodes/vertices of the graph each correspond to a neurosynaptic core, each core containing axons, dendrites, synapses, and neurons.
Cassidy further states;
([0030] The present disclosure provides for efficiently mapping an abstract (logical) network of neurosynaptic cores (for example, a Corelet) to an array of physical neurosynaptic cores (for example, the TrueNorth chip). Efficiency corresponds to minimizing the overall active power of the system as well as minimizing the spike bandwidth between multiple chips.)
NOTE: Explaining that the disclosed mapping to physical cores is done in a way which minimizes the overall active power of the system as well as spike bandwidth of the hardware.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to implement the entity nodes of Garrido to include respective neurosynaptic cores as taught by Cassidy, because Garrido already represents each entity using a population of neurons in neuromorphic hardware, and Cassidy teaches a known neuromorphic substrate implementation in which neural network nodes correspond to neurosynaptic cores containing neurons, synapses, axons, and dendrites. The combination would have predictably provided Garrido’s entity nodes with localized neuromorphic circuitry in an arrangement that minimizes power usage and spike bandwidth of the neuromorphic hardware.
Garrido and Cassidy fail to teach but Snelling teaches;
wherein at least one of the plurality of entity nodes is operatively connected to a controller to receive a query; and outputs the response through the controller.
([0035-0038] In embodiments, the data items may be triples each consisting of a data value for each of three triple elements, the three triple elements being: a subject, identifying the subject graph node linked by the labelled link encoded by the data item; an object, identifying the object graph node linked by the labelled link encoded by the data item; and a predicate, being the label of the labelled link encoded by the data item.)
NOTE: Teaches data items comprising of nodes of the graph.
([0010] the stored data access controller comprising: a query module, configured to receive a query requesting, as a query result, a specified subset of the data items stored in the data storage apparatus, and configured to obtain data items belonging to the specified subset from the data storage apparatus as a preliminary query result; a suppression module, configured to obtain the preliminary query result from the query module, to generate a revised version of the preliminary query result by selectively suppressing information from the preliminary query result, and to output, as the requested query result in response to the received query, the revised version of the preliminary query result;)
NOTE: Teaches at least one of the plurality entity nodes being operatively connected to a controller (data access controller) because the data access controller is able to interact with entity nodes (data items, which comprise the entity nodes). Snelling also teaches outputting the response to the query through the controller (the suppression module of the controller outputs the query result).
OBVIOUSNESS TO COMBINE SNELLING WITH GARRIDO AND CASSIDY: Snelling is analogous art to the present disclosure as it pertains to processing a query using a graph with nodes representing an entity or instance and edges representing a relationship between nodes.
Additionally Snelling states:
([0063] The query module 12 may receive the query from an application being operated by a user, from another machine, or via any form of interface enabling a subset to be defined or otherwise specified.)
NOTE: Indicating that the query module of the aforementioned data access controller allows a user to provide a query to be processed by the system.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to implement the knowledge graph system taught by Garrido in view of Cassidy to include the controller taught by Snelling to allow users to present input to be processed by the system.
the plurality of entity nodes and the one or more functional connections, responsive to the received query, generates a response to the received query,
Garrido, Cassidy, and Snelling fail to teach but Vakulenko teaches;
([pg. 4] The second step of the answer inference phase involves message passing, i.e., propagation of the confidence scores from the entities Ei and predicates Pi, matched in the question interpretation phase, to adjacent entities in the extracted subgraph.)
NOTE: Teaches the plurality of entity nodes Ei and one or more functional connections Pi, responsive to the received query (entities Ei and predicates Pi are representative of terms in the query), generates a response (answer inference) to the received query, by performing message passing operations.
OBVIOUSNESS TO COMBINE VAKULENKO WITH GARRIDO, CASSIDY, AND SNELLING:
Vakulenko is analogous art to the present disclosure as it pertains to processing queries using a knowledge graph.
Vakulenko further states;
([pg. 3] Our approach overcomes the limitations of the previously proposed graph-based approach in terms of efficiency and scalability, which we demonstrate on a compelling benchmark… Our evaluation results demonstrate improvements in precision and recall, while reducing average execution time over the SPARQL-based WDAqua, which is also orders of magnitude faster than results reported for the previous graph-based approach Treo.)
NOTE: Indicating that their method of graph-based query answering shows improvements in precision, recall, and execution time when compared to other graph-based query answering methods.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to use the neuromorphic hardware based computational knowledge graph taught by Garrido in view of Cassidy, having the controller configured to receive and output queries as taught by Snelling, to implement the knowledge graph based query answering methods proposed by Vakulenko to allow for query answering in the system with competitive precision, recall, and execution time.
Regarding claim 2, Garrido teaches;
wherein the plurality of entity nodes further have learning and application-specific capabilities.
([0011] An example are methods that learn representations (so-called embeddings) for entities in the graph in order to perform an inference task such as performing knowledge graph completion by inferring/predicting unobserved relationships (link prediction) or finding multiple instances of the same entity (entity resolution).)
NOTE: Teaches the plurality of entity nodes further having learning (the entity node embeddings learn representations for entities) and application-specific capabilities (the entity node embeddings can perform various inference tasks, link prediction, entity resolution, etc.).
Regarding claim 3, Garrido teaches;
wherein the processing capability of each entity node of the plurality of entity nodes enables each entity node to perform one or more of the following operations: message passing, pooling, routing, averaging, aggregating, concatenating, convoluting, updating, query operations, storage operations, arithmetic operations, logical operations, ontological operations, risk-related operations, credit-related operations, compliance operations, and natural-language processing operations.
([0057] “In an embodiment of the industrial device and method, learning updates for entity embeddings are computed using static feedback connections from each output neuron to neurons of the node embedding populations.”)
NOTE: Teaches the processing capability of each entity node of the plurality of entity nodes enables each entity node to perform at least updating.
Regarding claim 4, Garrido teaches;
wherein the hardware-based neuromorphic device of each entity node of the plurality of entity nodes includes neuronal circuitry elements.
([0035] In some embodiments a plurality of neurosynaptic cores are tiled on a chip. In an exemplary embodiment, a 64 by 64 grid of cores is tiled, yielding 4,096 cores, for a total of 1,048,576 neurons and 268,435,456 synapses. In such embodiments, neurons, synapses, and short-distance connectivity are implemented by the core circuit.)
NOTE: Teaches the hardware-based neuromorphic device of each entity node of the plurality of entity nodes (the neurosynaptic core of each node/vertex) includes neuronal circuitry elements (the core circuit).
OBVIOUSNESS:
Using the same reasoning from claim 1, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to implement the entity nodes of Garrido to include respective neurosynaptic cores as taught by Cassidy, because Garrido already represents each entity using a population of neurons in neuromorphic hardware, and Cassidy teaches a known neuromorphic substrate implementation in which neural network nodes correspond to neurosynaptic cores containing neurons, synapses, axons, and dendrites. The combination would have predictably provided Garrido’s entity nodes with localized neuromorphic circuitry in an arrangement that minimizes power usage and spike bandwidth of the neuromorphic hardware.
Claim(s) 5-8, 16-17, 19-20 is/are rejected under 35 U.S.C. 103 as being unpatentable by Garrido (US20220229400, 2022-07-21) in view of Cassidy (US 20190236444 A1, 2019-08-01) further in view of Snelling (US 20150193636 A1, 2015-07-09) further in view of Vakulenko (“Message Passing for Complex Question Answering over Knowledge Graphs”, 2019-08-19) further in view of Joel Filliben et al. (hereinafter Filliben) (US20190377819A1, 2019-12-12).
Regarding claim 5, Garrido, Cassidy, Snelling, and Vakulenko fail to teach but Filliben teaches;
wherein each particular entity node of the plurality of the entity nodes is configured to perform a set of functions assigned to each particular entity node.
([0121] Referring to FIG. 6, the machine learning system calculates and assigns, in step 612 and step 614, one or more customized functions to nodes in the graph structure based on the entity type of the node. For example, five hot file functions that may be implemented on the graph structure include, but are not limited to: detect, label, spread, fade out, and restrict.)
NOTE: Teaches each particular entity node of the plurality of the entity nodes is configured to perform the set of functions assigned to each particular entity node (one or more customized ‘hotfile’ functions are assigned to entity nodes; detect, label, spread, etc.)
OBVIOUSNESS TO COMBINE FILLIBEN WITH GARRIDO:
Filliben is analogous art to the present disclosure as it pertains to a graph-based machine learning system for determining transaction risk and fraud.
Garrido further states;
([0037] According to other embodiments, the learning component and/or the control component are implemented with neuromorphic hardware. The neuromorphic hardware embodiments empower edge learning devices for online graph learning and analytics. Being inspired by the mammalian brain, neuromorphic processors promise energy efficiency, fast emulation times as well as continuous learning capabilities. In contrast, graph-based data processing is commonly found in settings foreign to neuromorphic computing, where huge amounts of symbolic data from different data silos are combined, stored on servers and used to train models on the cloud. The aim of the neuromorphic hardware embodiments is to bridge these two worlds for scenarios where graph-structured data has to be analyzed dynamically, without huge data stores or off-loading to the cloud—an environment where neuromorphic devices have the potential to thrive.)
NOTE: Garrido states that neuromorphic hardware improves energy efficiency, emulation times, and continuous learning capabilities in graph learning and analytics.
Filliben further states;
([0121] As previously described, a series of human-designed pattern recognition rules, which are sometimes referred to as “hot files,” may be used to detect fraud in, for example, banking transactions. At initialization, hot files may, in some examples, be provided and/or designed manually by a human. Although the term “hot file” or “hotfile” is used in this specification, it is not intended to mean just a rule-based system that simple reacts through manual human intervention. Rather, as illustrated in FIG. 6, the coupling of a machine learning with a graph structure, which stores nodes and edges representing transaction data, creates an automated machine learning system that is at least a technological improvement over prior, manual hot file methodologies.)
NOTE: The method of coupling machine learning with a graph structure allows for automation of hot file-based fraud detection in financial transactions, which is an improvement over existing manual hot file-based fraud detection methods.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to configure neuromorphic hardware-based knowledge graph taught in claim 1 to implement Filliben’s machine learning based financial transaction graph processing, including associated automated hot-file fraud risk detection functionalities, because Filliben teaches that coupling machine learning with a transaction graph automates and improves prior manual hot-file detection methods, and Garrido teaches that implementing graph learning and analytics using neuromorphic hardware can improve energy efficiency, emulation times, and continuous learning capabilities.
Regarding claim 6, Garrido, Cassidy, Snelling, and Vakulenko fail to teach but Filliben teaches;
wherein after each particular entity node of the plurality of the entity nodes is configured to perform the set of functions assigned to each particular entity node, the set of functions assigned to each particular entity node is customizable based on other characteristics of each particular entity node.
([0092] In step 624, the system may update a hot file level (e.g., hot, warm, or cold) for an entity/edge/node. A hot file level may be a characterization of one or more hot file functions and/or restrictions, a weighting of one or more hot file functions and/or restrictions, or some other manner in which the strength of a hot file may be characterized or quantified.)
NOTE: Teaches after each particular entity node of the plurality of the entity nodes is configured to perform the set of functions assigned to each particular entity node, the set of functions assigned to each particular entity node is customizable based on other characteristics of each particular entity node, because the set of hotfile functions is characterized based a hot file level of an entity node, and the hot file level can be updated, thereby customizing the hot file functions to represent the updated hot file level.
OBVIOUSNESS:
Using the same reasoning from claim 5, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to configure the neuromorphic hardware-based knowledge graph taught in claim 1 to implement Filliben’s machine learning based financial transaction graph processing, including associated automated hot-file fraud risk detection functionalities, because Filliben teaches that coupling machine learning with a transaction graph automates and improves prior manual hot-file detection methods, and Garrido teaches that implementing graph learning and analytics using neuromorphic hardware can improve energy efficiency, emulation times, and continuous learning capabilities.
Regarding claim 7, Garrido, Cassidy, Snelling, and Vakulenko fail to teach but Filliben teaches;
wherein the other characteristics include a geographic location of each particular entity node
([0092] In step 624, the system may update a hot file level (e.g., hot, warm, or cold) for an entity/edge/node. A hot file level may be a characterization of one or more hot file functions and/or restrictions, a weighting of one or more hot file functions and/or restrictions, or some other manner in which the strength of a hot file may be characterized or quantified…The reverse may apply as well: hot files may become more “hot” over time. For example, over time, a credit card used frequently may become increasingly exposed to risk, counseling for the strength of a hot file to increase over time.)
NOTE: The hot file level of an entity node, which characterizes the aforementioned hot file functions, can increase based on the entity’s exposure to risk.
([0108] For example, the personal computer may have a corresponding Internet Protocol (IP) address, Media Access Control (MAC) address, and other similar information ... For example, an IP address, rather than a personal computer, may be considered high risk.)
NOTE: The IP address (IP address indicates geographic location; country, state, etc.) characteristic of a given entity node (the personal computer entity node, for example) can expose the entity node to risk, thereby increasing the corresponding hot file level. As previously mentioned, the hotfile functions are characterized by the hot file level of an entity node. Thus, teaches the other characteristics including a geographic location (IP address) of each particular entity node.
OBVIOUSNESS:
Using the same reasoning from claim 5, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to configure the neuromorphic hardware-based knowledge graph taught in claim 1 to implement Filliben’s machine learning based financial transaction graph processing, including associated automated hot-file fraud risk detection functionalities, because Filliben teaches that coupling machine learning with a transaction graph automates and improves prior manual hot-file detection methods, and Garrido teaches that implementing graph learning and analytics using neuromorphic hardware can improve energy efficiency, emulation times, and continuous learning capabilities.
Regarding claim 8, Garrido, Cassidy, Snelling, and Vakulenko fail to teach but Filliben teaches;
wherein one or more entity nodes of the plurality of entity nodes are configured as financial entities and the set of functions are application-specific and include one or more of compliance, legal, risk and market-related functions.
PNG
media_image1.png
696
494
media_image1.png
Greyscale
NOTE: Teaches the one or more entity nodes of the plurality of entity nodes are configured as financial entities (merchant, financial institution, etc.)
([0085] Such hot file functions may be directly associated with an entity (e.g., an indication that a certain type of computing device is extremely risky), a relationship between entities (e.g., that two entities, when used together, are extremely risky), or the like. For example, the system may use the fraud risk scores determined in step 604 to determine a hot file function for each of the entities that was involved in a transaction corresponding to the data received in step 602.)
NOTE: Teaches the set of functions are application-specific (risk-related functions are specific to the financial-transaction/fraud-detection application) and include one or more of compliance, legal, risk and market-related functions (the aforementioned hot file functions include at least risk related functions).
OBVIOUSNESS:
Using the same reasoning from claim 5, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to configure the neuromorphic hardware-based knowledge graph taught in claim 1 to implement Filliben’s machine learning based financial transaction graph processing, including associated automated hot-file fraud risk detection functionalities, because Filliben teaches that coupling machine learning with a transaction graph automates and improves prior manual hot-file detection methods, and Garrido teaches that implementing graph learning and analytics using neuromorphic hardware can improve energy efficiency, emulation times, and continuous learning capabilities.
Regarding claim 16, Garrido teaches;
A method of implementing a
([0190] For the following experiments, we use a recording from the industrial automation system with some default network and app activity, resulting in a knowledge graph KG with 3529 nodes, 11 node types)
NOTE: Teaches the plurality of entity nodes of the knowledge graph having a type.
entity characteristics and storage,
([0132] Each entity in the graph is represented by one of the node embedding populations NEP, storing both its embeddings (real-valued entries) and accumulated gradient updates.)
NOTE: Teaches entity storage (storing embeddings) and characteristics (embeddings and gradient updates of the entity node embedding populations can be considered characteristics of the entity)
a processing capability, and a communication capability,
([0057] In an embodiment of the industrial device and method, learning updates for entity embeddings are computed using static feedback connections from each output neuron to neurons of the node embedding populations.);
NOTE: Teaches a processing capability (neurons of the entity node embedding populations compute learning updates) and a communication capability of the entity nodes (entity node embeddings use feedback from output neurons, i.e., the entity node embeddings communicate with output neurons).
and ii) one or more functional connections associating the plurality of entity nodes to others of the plurality of entity nodes,
PNG
media_image2.png
526
429
media_image2.png
Greyscale
NOTE: Teaches one or more functional connections associating the plurality of entity nodes to others of the plurality of entity nodes (functional connection #runsOn associates entity node #app with #edgePC, etc.)
Garrido and Cassidy fail to teach but Snelling teaches;
providing a controller operatively connected to at least one entity node of the plurality of entity nodes to receive a query; generating a response to the received query; and outputting the response through the controller.
([0035-0038] In embodiments, the data items may be triples each consisting of a data value for each of three triple elements, the three triple elements being: a subject, identifying the subject graph node linked by the labelled link encoded by the data item; an object, identifying the object graph node linked by the labelled link encoded by the data item; and a predicate, being the label of the labelled link encoded by the data item.)
NOTE: Teaches data items containing nodes of the graph.
([0010] the stored data access controller comprising: a query module, configured to receive a query requesting, as a query result, a specified subset of the data items stored in the data storage apparatus, and configured to obtain data items belonging to the specified subset from the data storage apparatus as a preliminary query result; a suppression module, configured to obtain the preliminary query result from the query module, to generate a revised version of the preliminary query result by selectively suppressing information from the preliminary query result, and to output, as the requested query result in response to the received query, the revised version of the preliminary query result;)
NOTE: Teaches at least one of the plurality entity nodes being operatively connected to a controller (data access controller) because the data access controller is able to interact with entity nodes (data items, which contain the entity nodes). Snelling also teaches receiving the query via the controller (the query module of the data access controller is configured to receive a query), generating a response to the received query (the suppression module of the data access controller generates a revised query result) and outputting the response to the query through the controller (the suppression module of the data access controller outputs the query result).
OBVIOUSNESS:
Using the same reasoning from claim 1, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to implement the knowledge graph system taught in claim 1 to include the controller taught by Snelling to allow users to present input to be processed by the system.
Garrido, Cassidy, Snelling, and Vakulenko fail to teach but Filliben teaches;
a financial services knowledge graph
the method comprising: configuring the plurality of entity nodes to represent financial or legal entities;
PNG
media_image3.png
399
490
media_image3.png
Greyscale
([Abstract] An ensemble machine learning system is coupled to a graph module storing a graph structure, wherein a collection of entities and the relationships between those entities forms nodes and connection arcs between the various nodes.)
NOTE: Teaches a financial knowledge graph with the plurality of entity nodes to represent financial or legal entities (Financial institution, ATM, etc.)
configuring the one or more functional connections between the plurality of entity nodes that represent relationships or transactions between the plurality of entity nodes;
([Abstract] An ensemble machine learning system is coupled to a graph module storing a graph structure, wherein a collection of entities and the relationships between those entities forms nodes and connection arcs between the various nodes.)
NOTE: Teaches configuring the one or more functional connections between the plurality of entity nodes that represent relationships or transactions between the plurality of entity nodes.
embedding each processing capability of each entity node of the plurality of entity nodes to be customized to the financial service;
PNG
media_image3.png
399
490
media_image3.png
Greyscale
([0121] the machine learning system calculates and assigns, in step 612 and step 614, one or more customized functions to nodes in the graph structure based on the entity type of the node. For example, five hot file functions that may be implemented on the graph structure include, but are not limited to: detect, label, spread, fade out, and restrict.)
NOTE: Teaches embedding each processing capability of each entity node of the plurality of entity nodes to be customized to the financial service (customized hot file functions are assigned to a node based on the entity type for a node [credit card, ATM, merchant], i.e., they are customized to the financial service).
OBVIOUSNESS:
Using the same reasoning from claim 1, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to configure the neuromorphic hardware-based knowledge graph taught in claim 1 to implement Filliben’s machine learning based financial transaction graph processing, including associated automated hot-file fraud risk detection functionalities, because Filliben teaches that coupling machine learning with a transaction graph automates and improves prior manual hot-file detection methods, and Garrido teaches that implementing graph learning and analytics using neuromorphic hardware can improve energy efficiency, emulation times, and continuous learning capabilities.
Regarding claim 17, Garrido, Cassidy, Snelling, and Vakulenko fail to teach but Filliben teaches;
wherein each of the plurality of entity nodes represents one or more of corporations, clients, products, services and events.
PNG
media_image3.png
399
490
media_image3.png
Greyscale
NOTE: Teaches the plurality of entity nodes represents one or more of corporations (financial institution), clients, products (personal computer), services (checking account) and events (fraudulent transaction).
OBVIOUSNESS:
Using the same reasoning from claim 5, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to configure the neuromorphic hardware-based knowledge graph taught in claim 1 to implement Filliben’s machine learning based financial transaction graph processing, including associated automated hot-file fraud risk detection functionalities, because Filliben teaches that coupling machine learning with a transaction graph automates and improves prior manual hot-file detection methods, and Garrido teaches that implementing graph learning and analytics using neuromorphic hardware can improve energy efficiency, emulation times, and continuous learning capabilities.
Regarding claim 19, Garrido teaches;
wherein the processing capability of each entity node of the plurality of entity nodes includes the ability to perform one or more of message passing, pooling, routing, averaging, aggregating, concatenating, convoluting, updating, query operations, storage operations, arithmetic operations, logical operations, ontological operations, risk-related operations, credit-related operations, compliance operations, and natural-language processing operations.
([0057] “In an embodiment of the industrial device and method, learning updates for entity embeddings are computed using static feedback connections from each output neuron to neurons of the node embedding populations.”
NOTE: Teaches the processing capability of each entity node of the plurality of entity nodes enables each entity node to perform at least updating (learning updates for entity embeddings).
Regarding claim 20, Garrido, Cassidy, Snelling, and Vakulenko fail to teach but Filliben teaches;
wherein the processing capability of each entity node of the plurality of entity nodes is customized for compliance related functions including assessing risks and determining compliance with regulations.
([0085] Such hot file functions may be directly associated with an entity (e.g., an indication that a certain type of computing device is extremely risky), a relationship between entities (e.g., that two entities, when used together, are extremely risky), or the like.)
NOTE: Teaches the processing capability of each entity node of the plurality of entity nodes (the aforementioned hot file functions assigned to entity nodes) is customized for compliance related functions including assessing risks (indicating risk of entities or relationships).
([0121] Referring to FIG. 6, the machine learning system calculates and assigns, in step 612 and step 614, one or more customized functions to nodes in the graph structure based on the entity type of the node. For example, five hot file functions that may be implemented on the graph structure include, but are not limited to: detect, label, spread, fade out, and restrict. As previously described, a series of human-designed pattern recognition rules, which are sometimes referred to as “hot files,” may be used to detect fraud in, for example, banking transactions. At initialization, hot files may, in some examples, be provided and/or designed manually by a human.)
NOTE: Teaches the processing capability of each entity node of the plurality of entity nodes (the aforementioned hot file functions assigned to entity nodes) is customized for compliance related functions including determining compliance with regulations (detecting fraud in transactions is considered determining compliance with regulations).
OBVIOUSNESS:
Using the same reasoning from claim 5, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to configure the neuromorphic hardware-based knowledge graph taught in claim 1 to implement Filliben’s machine learning based financial transaction graph processing, including associated automated hot-file fraud risk detection functionalities, because Filliben teaches that coupling machine learning with a transaction graph automates and improves prior manual hot-file detection methods, and Garrido teaches that implementing graph learning and analytics using neuromorphic hardware can improve energy efficiency, emulation times, and continuous learning capabilities.
Claim(s) 9-10 is/are rejected under 35 U.S.C. 103 as being unpatentable by Garrido (US20220229400, 2022-07-21) in view of Cassidy (US 20190236444 A1, 2019-08-01), further in view of Snelling (US 20150193636 A1, 2015-07-09), further in view of Vakulenko (“Message Passing for Complex Question Answering over Knowledge Graphs”, 2019-08-19), further in view of Ankit Gandhi et al. (hereinafter Gandhi) (US 11593622 B1, 2023-02-28).
Regarding claim 9, Garrido, Cassidy, Snelling, and Vakulenko fail to explicitly teach but Gandhi teaches;
wherein each entity node of the plurality of entity nodes is connected to another entity node of the plurality of entity nodes by more than one functional connection of the one or more functional connections.
([col. 17, ln. 53-65] Access to a data set whose records represent events/transactions pertaining to instances of the entity types and relationship types may also be obtained in at least some embodiments (element 904). Using the data set as well as the entity and relationship information, a graph comprising a plurality of nodes and edges may be generated (element 907). The nodes may represent instances of the entity types, while an edge between a pair of nodes may indicate a relationship between the instances represented by the nodes. A given node may have multiple edges linking it to another node in at least some embodiments, indicating that relationships of multiple types exist between entity instances represented by a pair of connected nodes.)
NOTE: Teaches each entity node of the plurality of entity nodes is connected to another entity node of the plurality of entity nodes by more than one functional connection of the one or more functional connections (a given entity node may have multiple edges linking it to another node, indicating that relationships of multiple types exist between entity nodes).
OBVIOUSNESS TO COMBINE GANDHI WITH GARRIDO, CASSIDY, SNELLING, VAKULENKO, AND GANDHI:
Gandhi is analogous art to the present disclosure as it pertains to generating inference using a machine learning models used in conjunction with a knowledge graph (a graph containing entity nodes with connections representing relationships).
Additionally, Gandhi states;
PNG
media_image4.png
328
426
media_image4.png
Greyscale
([col. 11, ln. 60-65] Taking all the different combinations similar to those illustrated in FIG. 4 into account, a very large number of logical relationship types may be captured for a relatively small number of entity types in various embodiments, potentially enabling a rich set of inferences to be drawn regarding the interactions between entity instances.)
NOTE: Gandi states that the logical relationship types presented in their disclosure allow for a rich set of inferences to be drawn regarding interactions between entity nodes.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to allow for multiple nodes of the computational knowledge graph to be connected by more than one functional connection to enable a rich set of inferences to be drawn regarding the interactions between entity instances.
Regarding claim 10, Garrido teaches;
wherein the one or more functional connections comprises one of relationship connections and computational connections.
[pg. 1]
PNG
media_image5.png
444
393
media_image5.png
Greyscale
NOTE: Teaches the one or more functional connections comprises one of relationship connections and computational connections (comprises at least relationship connections; reads, measures, etc.).
Claim(s) 11-13 is/are rejected under 35 U.S.C. 103 as being unpatentable by Vakulenko (“Message Passing for Complex Question Answering over Knowledge Graphs”, 2019-08-19), in view of Garrido (US20220229400, 2022-07-21), further in view of Snelling (US 20150193636 A1, 2015-07-09), further in view of Sricharan Kumar et al. (hereinafter Kumar) (US 20210326531 A1, 2021-10-21).
Regarding claim 11, Vakulenko teaches;
A method of processing a query in a computational knowledge graph that includes i) a plurality of entity nodes, each entity node of the plurality of entity nodes of the computational knowledge graph having an entity type, entity characteristics and
[pg. 3]
PNG
media_image6.png
229
693
media_image6.png
Greyscale
[pg. 4]
PNG
media_image7.png
107
612
media_image7.png
Greyscale
NOTE: Teaches method of processing a query in a computational knowledge graph (answer inference based on knowledge graph utilizing computations) that includes a plurality of entity nodes (set of entities E), each entity node of the plurality of entity nodes of the computational knowledge graph having an entity type (class), entity characteristics (a class can be considered a characteristic) and storage, processing and communication capability (message passing between entities) and one or more functional connections associating the plurality of entity nodes to others of the plurality of entity nodes (predicates p associating entities and allow for message passing between entities).
decomposing the received query into query components
([pg. 3] To produce a question model we follow two steps: (1) parse, which extracts references (entity, predicate and class mentions) from the natural language question and identifies the question type;)
NOTE: Teaches decomposing the received query into query components (extracting references from the natural language question, which are considered components of the query / natural language question)
associating each query component with one or more entity nodes of the plurality of entity nodes and the one or more functional connections of the knowledge graph;
([pg. 3] To produce a question model we follow two steps: (1) parse, which extracts references (entity, predicate and class mentions) from the natural language question and identifies the question type; and (2) match, which assigns each of the extracted references to a ranked list of candidate entities, predicates and classes in the KG.)
NOTE: Teaches associating each query component with one or more entity nodes of the plurality of entity nodes and the one or more functional connections of the knowledge graph (assigning each of the extracted references to a ranked list of entities and predicates / functional connections).
distributing the query components to the one or more associated plurality of entity nodes for processing; generating a response to the received query;
[pg. 4]
PNG
media_image8.png
161
615
media_image8.png
Greyscale
PNG
media_image9.png
189
609
media_image9.png
Greyscale
PNG
media_image10.png
372
612
media_image10.png
Greyscale
NOTE: Teaches distributing the query components to the one or more associated plurality of entity nodes for processing (the references / query components are matched with their associated entities, which is then used in processing for answer inference) and generating a response to the received query (compute a final answer to the received query).
wherein the processing of the received query includes message passing between the plurality of entity nodes by exchanging messages between all of the plurality of entity nodes to calculate connectivity between the plurality of entity nodes, and performing one or more computational operations by each entity node of the plurality of entity nodes.
([pg. 4] The second step of the answer inference phase involves message passing)
PNG
media_image11.png
472
610
media_image11.png
Greyscale
PNG
media_image12.png
121
636
media_image12.png
Greyscale
[pg. 3]
PNG
media_image6.png
229
693
media_image6.png
Greyscale
NOTE: Teaches the processing of the received query includes message passing (second step of answer inference involves message passing) between the plurality of entity nodes by exchanging messages between all of the plurality of entity nodes (a subset of entities [i.e. a subgraph] can include all of the plurality of entities E) to calculate connectivity between the plurality of entity nodes (message passing over the KG subgraph by propagating entity-reference confidence scores through adjacency matrices corresponding to the KG properties/relations, thereby determining query-relevant connectivity between entity nodes), and performing one or more computational operations by each entity node of the plurality of entity nodes (message passing operation by each entity node of the plurality of entity nodes in the subgraph).
Vakulenko fails to teach but Garrido teaches;
i) a plurality of entity nodes, each entity node of the plurality of entity nodes of the computational knowledge graph having storage
([0132] Each entity in the graph is represented by one of the node embedding populations NEP, storing both its embeddings (real-valued entries) and accumulated gradient updates.)
NOTE: Teaches a plurality of entity nodes, each entity node of the plurality of entity nodes of the computational knowledge graph having storage (each entity node in the knowledge graph is represented by a node embedding population, storing its embeddings and accumulated gradient updates).
OBVIOUSNESS TO COMBINE GARRIDO WITH VAKULENKO:
Garrido is analogous art to the present disclosure as it pertains to knowledge graph having entity nodes implemented using neuromorphic hardware. Vakulenko is analogous art to the present disclosure as it pertains to query decomposition and processing using a knowledge graph.
Garrido further states;
([0037] According to other embodiments, the learning component and/or the control component are implemented with neuromorphic hardware. The neuromorphic hardware embodiments empower edge learning devices for online graph learning and analytics. Being inspired by the mammalian brain, neuromorphic processors promise energy efficiency, fast emulation times as well as continuous learning capabilities. In contrast, graph-based data processing is commonly found in settings foreign to neuromorphic computing, where huge amounts of symbolic data from different data silos are combined, stored on servers and used to train models on the cloud. The aim of the neuromorphic hardware embodiments is to bridge these two worlds for scenarios where graph-structured data has to be analyzed dynamically, without huge data stores or off-loading to the cloud—an environment where neuromorphic devices have the potential to thrive.)
NOTE: Garrido states that neuromorphic hardware improves energy efficiency, emulation times, and continuous learning capabilities in graph learning and analytics.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to implement the knowledge graph-based query answering system of Vakulenko using the neuromorphic hardware-based knowledge graph of Garrido, including entities having storage capabilities operatively connected to a controller for receiving queries and outputting responses, to improve energy efficiency and emulation times of the graph-based system.
Vakulenko and Garrido fail to teach but Snelling teaches;
the method comprising: receiving the query at a controller operatively connected to at least one of the plurality of entity nodes; and outputting the response through the controller,
([0035-0038] In embodiments, the data items may be triples each consisting of a data value for each of three triple elements, the three triple elements being: a subject, identifying the subject graph node linked by the labelled link encoded by the data item; an object, identifying the object graph node linked by the labelled link encoded by the data item; and a predicate, being the label of the labelled link encoded by the data item.)
NOTE: Teaches data items comprising of nodes of the graph.
([0010] the stored data access controller comprising: a query module, configured to receive a query requesting, as a query result, a specified subset of the data items stored in the data storage apparatus, and configured to obtain data items belonging to the specified subset from the data storage apparatus as a preliminary query result; a suppression module, configured to obtain the preliminary query result from the query module, to generate a revised version of the preliminary query result by selectively suppressing information from the preliminary query result, and to output, as the requested query result in response to the received query, the revised version of the preliminary query result;)
NOTE: Teaches at least one of the plurality entity nodes being operatively connected to a controller (data access controller) because the data access controller is able to interact with entity nodes (data items, which comprise the entity nodes). Snelling also teaches receiving the query via the controller (the query module of the data access controller is configured to receive a query), and outputting the response to the query through the controller (the suppression module of the controller outputs the query result).
OBVIOUSNESS TO COMBINE SNELLING WITH VAKULENKO AND GARRIDO: Snelling is analogous art to the present disclosure as it pertains to processing a query using a graph with nodes representing an entity / instance and edges representing relationships between nodes.
Vakulenko already provides a method of processing received query using a knowledge graph to generate a query result. Snelling provides a means of receiving a query via a controller, and outputting the results of processing said query.
Additionally Snelling states:
([0063] The query module 12 may receive the query from an application being operated by a user, from another machine, or via any form of interface enabling a subset to be defined or otherwise specified.)
NOTE: Indicating that the query module of the aforementioned data access controller allows a user to provide a query to be processed by the system.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to implement the neuromorphic hardware-based knowledge graph query answering system taught by Vakulenko in view of Garrido to include the controller taught by Snelling to allow users to present queries to be processed by the system.
Vakulenko, Garrido, and Snelling fail to teach but Kumar teaches;
decomposing the received query into query components including functional operations;
([Abstract] Certain aspects of the present disclosure provide techniques for processing natural language utterances in a knowledge graph. An example method generally includes receiving a long-tail query comprising a natural language utterance from a user of an application. Operands and operators are extracted from the natural language utterance using a natural language model.)
NOTE: Teaches decomposing the received query into query components including functional operations (extracting operators from a query).
OBVIOUSNESS TO COMBINE KUMAR WITH VAKULENKO, SNELLING, AND GARRIDO:
Kumar is analogous art to the present disclosure as it pertains to processing queries using query decomposition and a knowledge graph.
Additionally, Kumar states;
([0041] By extracting operators and operands from a natural language utterance, query processor 110 can generate an answer to rarely encountered questions for which intent-based query resolver is unable to generate an answer. For any given calculation to be performed on data associated with nodes in the knowledge graph, the calculation can be performed based on the extracted operators and operands rather than relying on an operation defined in a knowledge graph. Thus, chatbots or other support agents may be made more scalable, as various operations not defined in the knowledge graph may be answered by identifying an operation to perform on specified data in the knowledge graph.)
NOTE: Instead of requiring the chatbot/knowledge graph to already have a predefined operation or intent for every possible user question, the system can break the natural-language question (query) into operators and operands and then perform the needed calculation dynamically. Thus, decomposing the query into functional operations allows chatbots or other support agents to be more scalable.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, for the query components of the knowledge graph-based query answering system taught by Vakulenko to include functional operators to improve scalability of the query answering system.
Regarding claim 12, Vakulenko teaches;
The method of claim 11 further comprising pooling related entity nodes of the plurality of entity nodes to the associated query component such that the received query is parsed and processed by a subset of the related entity nodes in the knowledge graph.
([pg. 3] To produce a question model q we follow two steps: (1) parse, which extracts references (entity, predicate and class mentions) from the natural language question and identifies the question type;)
NOTE: Teaches the received query being parsed.
[pg. 4]
PNG
media_image13.png
219
607
media_image13.png
Greyscale
PNG
media_image7.png
107
612
media_image7.png
Greyscale
NOTE: The received query is parsed (as previously taught), and then is processed by a subset of related entity nodes in the knowledge graph, Ei, which can thus be considered the pooling related entities.
Regarding claim 13, Vakulenko teaches;
The method of claim 11, wherein the one or more computation operations include one or more of message passing, pooling, routing, averaging, aggregating, concatenating, convoluting, updating, query operations, storage operations, arithmetic operations, logical operations, ontological operations, risk-related operations, credit-related operations, compliance operations, and natural-language processing operations.
[pg. 4]
PNG
media_image7.png
107
612
media_image7.png
Greyscale
NOTE: Teaches the one or more computation operations include at least message passing.
Claim(s) 14-15 is/are rejected under 35 U.S.C. 103 as being unpatentable by Garrido (US20220229400, 2022-07-21) in view of Cassidy (US 20190236444 A1, 2019-08-01), further in view of Luhui Hu et al. (hereinafter Hu) (US 20250209383 A1, 2025-06-26).
Regarding claim 14, Garrido teaches;
Method of performing learning in a computational knowledge graph that includes i) a plurality of entity nodes, each entity node of the plurality of entity nodes of the computational knowledge graph having an entity type,
([0190] For the following experiments, we use a recording from the industrial automation system with some default network and app activity, resulting in a knowledge graph KG with 3529 nodes, 11 node types)
NOTE: Teaches a plurality of entity nodes, each entity node of the plurality of entity nodes of the computational knowledge graph having an entity type.
entity characteristics and storage,
NOTE: Teaches entity storage (storing embeddings) and characteristics (embeddings and gradient updates of the entity node embedding populations can be considered characteristics of the entity).
processing and communication capability
([0057] In an embodiment of the industrial device and method, learning updates for entity embeddings are computed using static feedback connections from each output neuron to neurons of the node embedding populations);
NOTE: Teaches a processing capability (neurons of the entity node embedding populations compute learning updates) and a communication capability of the entity nodes (entity node embeddings use feedback from output neurons, i.e., the entity node embeddings communicate with output neurons).
and ii) one or more functional connections associating the plurality of entity nodes to others of the plurality of entity nodes, the method comprising:
([0058] In an embodiment of the industrial device and method, the learning component includes first neurons forming a first node embedding population, representing a first entity contained in the triple statements by first spike times of the first neurons during a recurring time interval. The learning component includes second neurons forming a second node embedding population, representing a second entity contained in the triple statements by second spike times of the second neurons during the recurring time interval. A relation between the first entity and the second entity is represented as the differences between the first spike times and the second spike times.)
NOTE: Teaches one or more functional connections associating the plurality of entity nodes to others of the plurality of entity nodes (differences of spike times between a first entity node and second entity node represent a relation / connection, with the difference in spike times indicating the strength of the connection, which can be considered a function of the connection).
a neuronal circuitry element configured to perform a learning function;
([0050] In an embodiment of the industrial device and method, the learning component and/or the control component are implemented as neuromorphic hardware, in particular as an application specific integrated circuit, a field-programmable gate array, a wafer-scale integration, a hardware with mixed-mode VLSI neurons, or a neuromorphic processor, in particular a neural processing unit or a mixed-signal neuromorphic processor.)
NOTE: The learning component is implemented as neuromorphic hardware having neuronal circuitry elements (the neuromorphic hardware can be an application specific integrated circuit).
([0118] One of the selected modes of operation of the learning component LC is a learning mode, where the triple statements T are provided to the learning component LC, which in response iteratively updates its internal state with learning updates LU according to a specific cost function as described below.)
NOTE: The teaches the neuronal circuitry element of the neuromorphic hardware (the learning component, implemented as the aforementioned neuronal circuitry element of the neuromorphic hardware) configured to perform a learning function (learning mode).
and processing data obtained from the accessed one or more data sources using the accessed one or more learning functions to determine an inference regarding the processed data.
([0118] A further mode of operation [of the Learning Component] is inference mode, where the learning component LC makes predictions about the likelihood of unobserved triple statements. Inference mode can either be a free-running mode, whereby random triple statements are generated by the learning component LC based on the accumulated knowledge, or a targeted inference mode, where the control component CC specifically sets the learning component LC in such a way that the likelihood of specific triple statements is evaluated.)
NOTE: Teaches processing data obtained from the accessed one or more data sources (triple statements) using the accessed one or more learning functions (inference modes) to determine an inference regarding the processed data (make predictions).
Garrido fails to teach but Cassidy teaches;
providing each
([0035] In some embodiments a plurality of neurosynaptic cores are tiled on a chip. In an exemplary embodiment, a 64 by 64 grid of cores is tiled, yielding 4,096 cores, for a total of 1,048,576 neurons and 268,435,456 synapses. In such embodiments, neurons, synapses, and short-distance connectivity are implemented by the core circuit.)
NOTE: The hardware-based neuromorphic devices (neurosynaptic cores) include neuronal circuitry elements (core circuits).
([0072] After consuming the input files, placement systems according to the present disclosure such as NSCP create an internal representation of the network in the form of a graph. Each vertex of the graph is a neurosynaptic core. The edges of the graph are the number of connections between each pair of cores (in terms of the static number of connections) or the number of spikes between those two cores (representing dynamic activity in the network).)
NOTE: Teaches a graph representing a neural network where each vertex/node of the graph includes a hardware-based neuromorphic device (neurosynaptic core) including neuronal circuitry elements (the aforementioned core circuits).
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to implement the entity nodes taught by Garrido to each include the neurosynaptic cores taught by Cassidy, further explained below.
OBVIOUSNESS TO COMBINE CASSIDY WITH GARRIDO:
Cassidy is analogous art to the present disclosure as it pertains to a plurality of nodes each including neuromorphic hardware-based devices (neurosynaptic cores) while Garrido is analogous art to the present disclosure as it pertains to a plurality of entity nodes of a knowledge graph having at least processing, storage, communication, and learning capabilities.
Garrido teaches neuromorphic hardware for storing a knowledge graph representation where entities are represented by a respective embedding population of neurons, allowing for at least processing, storage, communication, and learning capabilities.
Cassidy supplies implementation details for a neuromorphic substrate implementing a graph representation of a neural network, where the plurality of nodes/vertices of the graph each correspond to a neurosynaptic core, each core containing axons, dendrites, synapses, and neurons.
Cassidy further states;
([0030] The present disclosure provides for efficiently mapping an abstract (logical) network of neurosynaptic cores (for example, a Corelet) to an array of physical neurosynaptic cores (for example, the TrueNorth chip). Efficiency corresponds to minimizing the overall active power of the system as well as minimizing the spike bandwidth between multiple chips.)
NOTE: Explaining that the disclosed mapping to physical cores is done in a way which minimizes the overall active power of the system as well as spike bandwidth of the hardware.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to implement the entity nodes of Garrido to include respective neurosynaptic cores as taught by Cassidy, because Garrido already represents each entity using a population of neurons in neuromorphic hardware, and Cassidy teaches a known neuromorphic substrate implementation in which neural network nodes correspond to neurosynaptic cores containing neurons, synapses, axons, and dendrites. The combination would have predictably provided Garrido’s entity nodes with localized neuromorphic circuitry in an arrangement that minimizes power usage and spike bandwidth of the neuromorphic hardware.
Garrido and Cassidy fail to teach but Hu teaches;
accessing one or more data sources and one or more learning functions of the plurality of entity nodes at an entity node level,
([Abstract] In one embodiment, one or more computing systems may generate a knowledge graph representing relationships between a global model and a number of local models for updating the global model. Each local model may have access to a local dataset for machine-learning training.)
NOTE: Teaches accessing one or more data sources (local dataset) and one or more learning functions (local machine learning model) of the plurality of entity nodes at an entity node level (each node of the knowledge graph has a local model).
whereby each of the plurality of entity nodes accesses the one or more data sources and executes the accessed one or more learning functions;
([0005] Each local node may continuous refine/train its local model.)
([Abstract] Each local model may have access to a local dataset for machine-learning training.)
NOTE: Teaches each of the plurality of entity nodes accesses the one or more data sources (local dataset) and executes the accessed one or more learning functions (each node trains its local model).
OBVIOUSNESS TO COMBINE HU WITH GARRIDO AND CASSIDY:
Hu is analogous art to the present disclosure as it pertains to machine learning processing using a knowledge graph.
Specifically, Hu teaches a knowledge graph to coordinate federated learning, where entity nodes represent federated nodes.
Garrido teaches a neuromorphic hardware-based knowledge graph in which entity nodes and relations are represented in spiking/neuronal hardware, and the learning component performs learning/inference over graph triples. Garrido further teaches that such neuromorphic graph-learning systems are useful for graph analytics and can provide energy-efficient and flexible processing/learning units for online evaluation of graph data.
Additionally, Hu states;
([0005] Furthermore, the knowledge graph system may use graph learning or machine-learning to determine one or more inferences to coordinate and optimize the federated learning process. For example, two local sites represented by two graph nodes in the knowledge graph may have a particular type of data that is similar and can be combined to generate better training samples. The knowledge graph system may use the knowledge graph to determine that these two nodes cannot share that particular type of data with the server, but these two nodes can share this data with each other. The knowledge graph system may send instructions to these two local sites to cause the data to be shared between these two local sites. As a result, the ML (machine learning) models may be trained more effectively due to the combined data.)
NOTE: Hu improves local learning of the entity nodes by combining complementary local data when allowed.
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Garrido’s neuromorphic hardware-based KG to include Hu’s KG based federated learning framework to improve local learning of the entity nodes.
Regarding claim 15, Garrido teaches;
using the accessed one or more learning functions to determine the inference
([0118] A further mode of operation [of the Learning Component] is inference mode, where the learning component LC makes predictions about the likelihood of unobserved triple statements. Inference mode can either be a free-running mode, whereby random triple statements are generated by the learning component LC based on the accumulated knowledge, or a targeted inference mode, where the control component CC specifically sets the learning component LC in such a way that the likelihood of specific triple statements is evaluated.)
NOTE: Teaches using the accessed one or more learning functions (inference modes) to determine the inference (predictions).
Garrido and Cassidy fail to teach but Hu teaches;
wherein two or more entity nodes of the plurality of entity nodes process the processed data collaboratively using the accessed one or more learning functions
([0005] For example, two local sites represented by two graph nodes in the knowledge graph may have a particular type of data that is similar and can be combined to generate better training samples. The knowledge graph system may use the knowledge graph to determine that these two nodes cannot share that particular type of data with the server, but these two nodes can share this data with each other. The knowledge graph system may send instructions to these two local sites to cause the data to be shared between these two local sites. As a result, the ML (machine learning) models may be trained more effectively due to the combined data.)
NOTE: Teaches two or more entity nodes of the plurality of entity nodes process the processed data collaboratively (data being shared between local sites) using the accessed one or more learning functions (the local models of the entity nodes are trained on the combined data).
OBVIOUSNESS:
Using the same reasoning from claim 14, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Garrido’s neuromorphic hardware-based KG to include Hu’s KG based federated learning framework to improve local learning of the entity nodes, thereby improving the inference of the system.
Claim(s) 18 is/are rejected under 35 U.S.C. 103 as being unpatentable by Garrido (US20220229400, 2022-07-21) in view of Cassidy (US 20190236444 A1, 2019-08-01) further in view of Snelling (US 20150193636 A1, 2015-07-09) further in view of Vakulenko (“Message Passing for Complex Question Answering over Knowledge Graphs”, 2019-08-19) further in view of Filliben (US20190377819A1, 2019-12-12) further in view of Gandhi (US 11593622 B1, 2023-02-28).
Regarding claim 18, Garrido, Cassidy, Snelling, Vakulenko, and Filliben fail to teach but Gandhi teaches;
wherein the one or more functional connections represent one or more of service contracts, product-specific client relationships, third party services, and financial events.
PNG
media_image14.png
322
445
media_image14.png
Greyscale
NOTE: Teaches the one or more functional connections represent one or more of service contracts, product-specific client relationships, third party services, and financial events (has-sold-to, has-granted-premier-status-to, etc.).
OBVIOUSNESS:
Using the same reasoning from claim 9, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the functional connections of the knowledge graph of Garrido to be represented by the logical functional connections taught by Gandhi to enable a rich set of inferences to be drawn regarding the interactions between entity instances in a financial setting.
Response to Arguments
Applicant's arguments filed 04/03/2026 regarding 35 U.S.C. 112 have been fully considered and they are persuasive. Amended claims 1-20 filed 04/03/2026 overcome all of the 35 U.S.C. 112 rejections except the following issue recited in claim 16;
The term “the financial service” in claim 16 lacks antecedent basis as there are no prior reference to a financial service made in the claims. Claims 17-20 are dependent claims that depend on indefinite claim 16. They are therefore rejected by virtue of its dependency.
Applicant's arguments filed 04/03/2026 regarding step 1 of the Alice/Mayo analysis under 35 U.S.C. 101 have been fully considered and are persuasive. Amended claims 1-10 filed 04/03/2026 are now directed to a physical system, which is considered eligible subject matter (“A hardware-based computational knowledge graph system comprising: a plurality of entity nodes, each of the entity nodes including a hardware-based neuromorphic device ...”), and therefore passes step 1 of the Alice/Mayo analysis.
Applicant's arguments filed 04/03/2026 regarding the abstract idea rejections from the previous office action made under 35 U.S.C. 101 have been fully considered but are not persuasive;
Beginning on page 3, applicant states “It is respectfully submitted that the claimed invention of claims 1-20 is not directed to an abstract idea, or to a law of nature or a natural phenomenon… the claimed invention of claims 1-20 is directed to an improvement to computer functionality or computer-related technology, and such an improvement is shown in the teachings in the specification about how the claimed invention improves a computer, a computer functionality, or other technology. Alternatively, such an improvement to computer functionality or computer-related technology is shown in the teaching in the specification of a particular solution to a problem or a particular way to achieve a desired outcome defined by the claimed invention.”
Applicant further cites reasoning passages from the specification outlining the alleged improvements;
"computational knowledge graph has numerous advantages including response time and performance improvements, resilience, and increased parallelism" (present application, paragraph [0044]);
This excerpt sets forth an improvement in a conclusory manner, which does not indicate an improvement in technology (see MPEP 2106.05(a)).
"For other knowledge graph operations, tasks can also be assigned to local entity nodes for processing, or alternatively (or additionally), can be distributed between centralized processing and local processing and anything in between. This flexibility improves the overall efficiency and performance of the knowledge graph system" (present application, paragraph [0048]);
This excerpt states that having the ability to alternate between centralized and local processing improves efficiency and performance of the KG, but this feature is not included in the claims, and thus does not indicate the claimed invention providing an improvement (see MPEP 2106.05(a)).
"the integrated knowledge graph system is particularly useful in the context of financial services, in which transactions need to both refer to volumes of distinct data sources and be processed quickly" (present application, paragraph [0061]);
This excerpt sets forth an improvement in a conclusory manner, which does not indicate an improvement in technology (see MPEP 2106.05(a)).
and "The improved speed and efficiency of the disclosed system-level implementations enable updating of large knowledge graphs in real-time through parallel access to the knowledge graph. Similarly, the improvements allow machine learning to be applied to large scale knowledge graphs in ways previously unfeasible and also allows parallel learning operations" (present application, paragraph [0077]).
This excerpt sets forth an improvement in a conclusory manner, which does not indicate an improvement in technology (see MPEP 2106.05(a)).
Such improvements to computer functionality or computer-related technology solve the technical problem of using knowledge graphs for structured queries (present application, paragraph [0003])
Paragraph [0003] presents general use cases for knowledge graphs, which does not indicate that the claimed invention provides an improvement.
"[improve] knowledge graph processing performance" (present application, paragraph [0006])
This excerpt recites the problem without addressing an improvement from the claimed invention.
"the size and complexity of financial services knowledge graphs grow (involving tens to hundreds of millions of entities, numerous relationship types, complex schemas, etc.)" (present application, paragraph [0077])
Paragraph [0077] addresses improvements regarding parallel querying in a conclusory manner. Additionally, parallel querying is not addressed in the claims, and thus does not indicate that the claimed invention provides an improvement.
However, one of the applicant’s arguments with respect to the abstract idea rejection under 35 U.S.C 101, see pg. 4 ("With the availability of the neuromorphic chips the implementation of knowledge graphs can be improved dramatically" (present application, paragraph [0059]), is persuasive. Paragraph [0059] does provide a technical explanation as to why the neuromorphic hardware implementation of the computational knowledge graph (which is recited in claims 1-10, 14-15) does provide an improvement to the technology. Thus, the abstract idea rejection of amended claims 1-10, 14-15, which include the implementation of entity nodes using neuromorphic hardware, has been withdrawn.
Applicant's arguments filed 04/03/2026 regarding the 35. U.S.C 102 rejection of claims 14-15 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Applicant's arguments filed 04/03/2026 regarding the 35. U.S.C 103 rejection of claims 1-10, 16-20 have been fully considered but they are not persuasive;
Applicant argues that amended independent claim 1 (and dependent claims 2-10) is patentable over Garrido, because Garrido does not disclose or suggest at least one of the plurality of entity nodes is operatively connected to a controller to receive a query ... wherein the plurality of entity nodes and the one or more functional connections, responsive to the received query, generates a response to the received query, and outputs the response through the controller, as in amended independent claim 1.
However, Snelling teaches a controller operatively connected to at least one entity node of a plurality of entity nodes to receive a query, and output a response to the query through the controller. Snelling teaches a data access controller that interacts with entity nodes (wherein the interactions with the entity nodes can be considered an operative connection), receives a query through its query module, and outputs a response to the query through the its suppression module. Additionally, Vakulenko teaches a plurality of entity nodes (entities Ei) and one or more functional connections (predicates Pi), responsive to the received query, generating a response to the received query by performing a series of message passing operations. Snelling and Vakulenko are used in the new 103 rejection of claim 1, which is necessitated by the amendments to claim 1.
Applicant argues that amended independent claim 16 (and dependent claims 17-20) is patentable over Garrido, because Garrido does not disclose or suggest a controller operatively connected to at least one entity node of the plurality of entity nodes to receive a query ... generating a response to the received query; and outputting the response through the controller, as in amended independent claim 16.
However, Snelling teaches a controller operatively connected to at least one entity node of a plurality of entity nodes to receive a query, generate a response to the query, and output a response through the controller. Snelling teaches a data access controller that interacts with entity nodes (wherein the interactions with the entity nodes can be considered an operative connection), receives a query through its query module, and generates and outputs the response to the query through the its suppression module. Snelling is used in the new 103 rejection of claim 16, which is necessitated by the amendments of claim 16.
Applicant’s arguments filed on 04/03/2026 with respect to regarding the 35. U.S.C 103 rejection of claim(s) 11-13 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
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 Matthew Alan Cady whose telephone number is (571) 272-7229. The examiner can normally be reached Monday - Friday, 7:30 am - 5:00 pm ET.
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, Cesar Paula can be reached on (571)272-4128. 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.
/MATTHEW ALAN CADY/ Examiner, Art Unit 2145
/CESAR B PAULA/ Supervisory Patent Examiner, Art Unit 2145
Prosecution Insights