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 . This action is made non-final.
This action is in response to the amendments filed April 9, 2026.
Claims 1, 12, and 17 have been amended.
Response to Amendment
The amendments filed April 9, 2026 has been entered. Claims 1, 3-6, 8-12, 14-17, and 19-20 are pending in the case and have been examined. Claims 1, 3-6, 8-12, 14-17, and 19-20 are rejected.
Response to Arguments
Regarding the 101 Arguments
Applicant’s arguments, see pages 9-11, filed April 9, 2026, with respect to Step 2A Prong Two have been fully considered and are persuasive. The rejection of March 10, 2026 has been withdrawn.
Regarding the 103 Arguments
Applicant’s arguments with respect to claim(s) 1, 3-6, 8-12, 14-17, and 19-20 have been considered but are moot because the amendments necessitated the new ground of rejection. Applicant’s arguments are directed to the rejection set forth in the prior Office Action, whereas the present rejection has been modified to account for the amened claim language and further relies on the new ground of rejection for the amended language.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claim(s) 1, 3-6, 8, 12, 14-17, 19, and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Tang et al. (US 11496493 B2, referred to as Tang), in view of Ferreira Moreno et al. (“US 20220198211 A1”, referred to as Ferreira Moreno), in view of Yang et al. (“Conditional structure generation through graph variational generative adversarial nets”, referred to as Yang), in view of Tang et al. (US 11496493 B2, referred to as Tang), in view of Sun et al. (US 20230351215 A1, referred to as Sun), in view of Cai et al. (“Time in a box: advancing knowledge graph completion with temporal scopes.”, referred to as Cai).
Regarding claim 1, Tang teaches, a computer-implemented method comprising:
generating a test graph corresponding to an entity (Col. 5, lines 28 cont. Col. 6, lines 1-21: Describes a dynamic graph analysis engine that processes communications and profile data associated with multiple devices to determine dynamic graphs, in the graphs, each node represents a device/server/source entity and edges represent communications between nodes. An input preprocessor that performs identifier matching to extract valid data for generating the dynamic graphs, each graph being represented by noes and edges.), wherein the test graph has a hybrid structure comprising a static graph that includes a reference vertex associated with an entity node and a dynamic graph connected to the entity node by a reference vertex, and wherein the static graph comprises a first plurality of peripheral nodes connected to the entity node, and the dynamic graph has a second plurality of peripheral nodes connected to the reference vertex(Col. 6, lines 22-41: Describes separate static and dynamic graph components, corresponding to a hybrid graph construction.; Col. 8, lines 27-45: Describes a focal node, which is a reference/anchor vertex associated with the entity node. The “nodes to whom ego is directly connected to (for example, “alters”) plus the ties, if any, among the alters” are a plurality of peripheral nodes connected to that entity/reference node. The static one-hop/multi-hop features are computes from this topology, i.e., the static graph side. Dynamic features are computed per time snapshot around the same focal/anchor node, corresponding to the dynamic graph being attached via the same reference node and includes another set of neighbors (peripheral nodes) observable in the time-varying snapshots.)
Although Tang teaches generating a test graph corresponding to an entity, wherein the test graph has a hybrid structure comprising a static graph that includes a reference vertex associated with an entity node and a dynamic graph connected to the entity node by a reference vertex, and wherein the static graph comprises a first plurality of peripheral nodes connected to the entity node, and the dynamic graph has a second plurality of peripheral nodes connected to the reference vertex. It does not teach, decoupling various properties of the first plurality of peripheral nodes and the second plurality of peripheral nodes from the entity node and the reference vertex.
Ferreira Moreno teaches, decoupling various properties of the first plurality of peripheral nodes and the second plurality of peripheral nodes from the entity node and the reference vertex ([0037-0040], and [0048-0050]: Describes a knowledge structurer that constructs a knowledge graph including nodes, anchors, and edges, wherein a content node represents content, concept nodes represent concepts extracted from the content, and edges connect the nodes. For each extracted label, a concept node is created to represent the label, spatial/temporal/spatiotemporal anchors are created, and a link r relationship is created between the appropriate anchor and the respective concept node fragment. A concept node can be represented as a <fragment, concept> pair, with the anchor representing spatial and/or temporal fragments of the content node. which represents properties/concepts as separate graph nodes linked by anchors/edges, rather than embedding those properties in the entity/content node itself, corresponding to decoupling various properties of the peripheral nodes form the entity node/reference vertex.)
It would have been obvious to one of ordinary skill in the art at the time of the claimed invention to incorporate Ferreira Morenos separately linked concept nodes with Tang’s graph structure. Doing so would have improved the graphs reasoning and retrieval over entity related information by allowing properties to be separately represented, linked, and traversed rather than embedded within a single entity node.
Although Tang in view of Ferreira Moreno teaches, decoupling various properties of the first plurality of peripheral nodes and the second plurality of peripheral nodes from the entity node and the reference vertex. They do not teach, applying the test graph to a conditional generative model, wherein the conditional generative model has a graph neural network structure.
Yang teaches, applying the test graph to a conditional generative model, wherein the conditional generative model has a graph neural network structure (Pages 1-5 Sections 1, 2.1-2.2 and Fig. 2: Describes a conditional structure generation, where the graph structure is generated according to semantic conditions indicated by contextual attributes or labels. A set of graphs G={G1, G2, …, Gn}, where each graph Gi={Vi, Ei} corresponds to a graph structure described by nodes Vi and edges Ei, and further constructs a condition vector Ci for each graph Gi describing graph context. It trains a model on graph/condition pairs T={Gi, Ci} to generate graph structures for seen and unseen conditions. The CONDGEN model is a graph variational generative adversarial network (CGN) that leverages graph convolutional neural networks including a GCN based graph encoder and GCN based discriminator/generator framework.) It would have been obvious to one of ordinary skill in the art at the time of the claimed invention to have combined the graph structure of Tang, in view of Ferreira Moreno, with the conditional graph model of Yang. Doing so would have allowed the system to generate or predict graph relationships and graph structure based on contextual graph information.
Yang further teaches, accessing one or more speculative transient attributes (Pages 2-3 Sections 1, 2.1 and Fig. 1: Describes that graphs may be associated with contextual attributes or labels and that a condition vector is constructed for each graph to describe the graph context. It generates graph structures for seen conditions and for unseen conditions, including situations where network data for certain conditions is scarce or missing. );
applying the one or more speculative transient attributes and the test graph to the conditional generative model (Pages 3-6 Sections 2.1-2.3: Describes using a graph Gi together with a condition vector Ci (graph context), and training on graph/condition pairs T={Gi, Ci} to generate structures for seen and unseen conditions.)
Although Tang, in view of Ferreira Moreno, in view of Yang teaches, applying the test graph to a conditional generative model, wherein the conditional generative model has a graph neural network structure. They do not teach predicting a first chronological state for the test graph based on the applying
Sun teaches, predicting a first chronological state for the test graph based on the applying ([0052-0057], [0070], and [0189-0197]: Describes a dynamic graph which may be represented as a sequence of graph snapshots from different time steps, and that each graph snapshot can include graph data associated with a timestamp. It uses temporal convolution across graph snapshots to capture temporal; evolutionary patterns and generate final node representations representing changes in graph structure over time. It performs graph context prediction using the final node representations, including predicting whether two nodes will be linked or connected by an edge in a future graph snapshot and/or time associated with such a snapshot.
It would have been obvious to one of ordinary skill in the art at the time of the claimed invention to have combined the graph structure of Tang in view of Ferreira Moreno, in view of Yang with the temporal graph prediction of Sun. Doing so would have allowed the system to account for time varying graph relationships and predict chronological graph states, improving prediction of future graph structure based on temporal evolution.
Although Tang in view of Ferreira Moreno, in view of Yang in view of Sun teaches predicting a first chronological state for the test graph based on the applying. They do not teach, wherein the one or more speculative transient attributes are represented as conjectural nodes unconnected to any chronological vertex prior to inference.
Cai teaches wherein the one or more speculative transient attributes are represented as conjectural nodes unconnected to any chronological vertex prior to inference (Pages 121-122 Introduction, and Pages 122-123 Section 3.1Describes that temporal knowledge base statements may lack known temporal scope and that prior systems excluding such statements lose useful information. The TIME2BOX handles statements with no validity information together with temporal statements and performs time prediction for queries of the form (s, r, o, ?I). Which maintains information without a known temporal association prior to inference and predicting the missing temporal scope during inference.; Ferreira Moreno teaches the concept of nodes and temporal/spatiotemporal anchors in a knowledge graph, Cai’s statement or attribute without known temporal scope corresponds to information maintained without connection to a chronological/temporal anchor prior to inference, and Cai’s time prediction corresponds to predicting the missing temporal scope during inference.)
It would have been obvious to one of ordinary skill in the art at the time of the claimed invention to have combined the graph structure of Tang in view of Ferreira Moreno, in view of Yang in view of Sun, with the temporal knowledge base completion. Doing so would have allowed the system to retain and use temporally incomplete graph information while predicting the missing chronological/temporal association.
Cai further teaches predicting a second chronological state for the test graph (Page 123 Section 3.2 and Page 126 Section 6: Describes temporal knowledge base completion, wherein statements in a temporal knowledge base may have known or missing temporal scope. The validity time period of a statement is often missing and that TIME2BOX is configured to handle statements with and without known temporal scope. A time prediction task using queries of the form (s, r, o, ?I), where the model predicts a missing time interval for a statement. Predicting a chronological state, corresponding to a predicted temporal scope or time interval associated with a graph statement/fact.); and
Tang in view of Ferreira Moreno, in view of Yang in view of Sun, in view of Cai teaches, training the conditional generative model on a set of training graphs (Yang Page 3, Section 2.1 and Page 6 Section 2.3: Describes training a model on a set of graphs with conditions (T = {Gi, Ci}) where each Gi={Vi, Ei} is a graph structure with nodes and edges, and Ci is a condition vector describing graph context.) to learn temporal evolutions of chronological states of the set (Sun [0038-0044], and [0195-0197] Describes dynamic graphs represented as graph snapshots from different time steps, and specifically teaches learning node representations that capture structural features and temporal evolution patterns. Real world graphs evolve over time and that modeling temporal evolutionary patterns is important for accurately predicting node properties and future links.), wherein each chronological state is represented by one or more chronological vertices (Ferreira Moreno [0027-0039], and FIG. 2: Describes creating spatial, temporal, and/or spatiotemporal anchors in the knowledge graph and creating nodes representing concepts and anchors. In FIG. 2,it shows “Create Nodes Representing Concepts, Create Spatial, Temporal, and/or Spatiotemporal Anchors.”), and wherein the first plurality and the second plurality of peripheral nodes are represented by transient attributes distinct from the one or more chronological vertices (Ferreira Moreno [0036-0039], and [0047-0049]: Separates concept nodes form temporal/spatiotemporal anchors. Concept nodes represent extracted labels/concepts, while anchors represent spatial/temporal information. ), and wherein the peripheral nodes and the one or more chronological vertices are connected via edges (Ferreira Moreno [0037-0040], and [0049]: Describes that the knowledge graph has nodes, anchors, and edges/links, ad that a link or relationship is created between the appropriate anchor and the respective concept node fragment. Corresponding to peripheral/concept nodes connected to chronological/temporal anchors via edges/links.), and wherein the predicting further comprises:
predicting at least one chronological vertex (Cai, Page 121 Introduction and Page 123 Section 3.2: Describes a temporal knowledge base completion framework in which statements may have missing temporal scope, and further teaches a time prediction task using queries of the form (s, r, o, ?I), where the model predicts a missing time interval for a statement. It represents temporal information in a knowledge graph using temporal or spatiotemporal anchors associated with graph nodes.): and
associating a subset of the peripheral nodes with the at least one chronological vertex (Ferreira Moreno [0037-0044]: Describes a knowledge graph including nodes, anchors, and edges, wherein concept nodes represent extracted concepts and temporal or spatiotemporal anchors represent time associated fragments. It creates link or relationship between the appropriate temporal/spatiotemporal anchor and the respective concept node fragment.)
Regarding claim 2 (Cancelled)
Regarding claim 3, Tang in view of Ferreira Moreno, in view of Yang in view of Sun, in view of Cai teaches the computer-implemented method of claim 1 wherein predicting a second chronological state further comprises: predicting a second chronological vertex associated with at least one of the one or more speculative transient attributes(Cai, Page 123 Section 3.2: Describes time prediction queries of the form (s, r, o, ?I), where the model predicts a missing time interval for a statement.; Ferreira Moreno [0038-0039], and [0049]: Describes temporal/spatiotemporal anchors associated with concept node fragments and creating a link or relationship between the appropriate anchor and the respective concept node fragment.).
Regarding claim 4, Tang in view of Ferreira Moreno, in view of Yang in view of Sun, in view of Cai teaches the computer-implemented method of claim 1.
Yang further teaches accessing a modified version of the one or more speculative transient attributes (Page 3, Section 2.1: Describes constructing a condition vector Ci for each graph Gi to describe graph context, training a model on graph/condition pairs T={Gi, Ci}, and generating graphs for seen and unseen conditions, where a different or modified condition vector corresponds to a modified version of the contextual/speculative attributes.); and
determining the second chronological state of the test graph based on the modified version of the one or more speculative transient attributes (Yang Page 3, Section 2.1: Describes generating graph structures based on the graph condition/context.; Cai Page 123 Section 3.2: Describes time prediction queries of the form (s, r, o, ?I), where the model predicts a missing time interval for a statement.).
Regarding claim 5, Tang in view of Ferreira Moreno, in view of Yang in view of Sun, in view of Cai teaches the computer-implemented method of claim 4.
Ferreira Moreno teaches analyzing the second chronological state; and
determining an outcome for the second chronological state, based on the analyzing ([0042-0044]: Describes extracting one or more subgraphs from the knowledge graph, calculating and deciding whether connected ,fragment, concept. pairs match spatial, temporal, or spatiotemporal requirements, processing the subgraphs to create event chains, calculating and deciding which chains meet temporal, spatial, or spatiotemporal restrictions, selecting a sequence that meets the query, and composing new content representing the result of the search query.).
Regarding claim 6, Tang in view of Ferreira Moreno, in view of Yang in view of Sun, in view of Cai teaches the computer-implemented method of claim 1.
Tang further teaches, adding one or more actual chronological vertices associated with transient actual properties (Col. 6, lines 1-21, and Col. 7, lines 12-67, Cont. Col. 8, lines 1-19: Describes per-time instances of the node’s states, chronological vertices with transient properties.); and
ordering the added vertices chronologically via oriented chronological edges, with respect to a most recent one of the chronological vertices of the test graph (Col. 6, lines 65-67, Cont. Col. 7, lines 1-11 Describes directed edges (oriented in-going and out-going edges.; Col. 7, lines 12-67, Cont. Col. 8, lines 1-19: Describes computing next/current month and year ratios, requiring chronological ordering anchored at the most recent snapshot, corresponding to comparing t and t-1 in an ordered series.).
Regarding claim 7 (Cancelled)
Regarding claim 8, Tang in view of Ferreira Moreno, in view of Yang in view of Sun, in view of Cai teaches the computer-implemented method of claim 6.
Cai teaches wherein training further comprises: imposing a loss function to a predetermined minimal smoothness of time variations for the predicted states (Page 125 Sections 4.3-4.5: Describes a time smoothness regularizer, where neighboring timestamps are expected to have similar representations in vector space, differences between embeddings of consecutive timestamps are penalized using an L2 term, and the regularizer is added to the loss function during training with a weight scalar β specifying the degree of penalization.).
Regarding claims 12, 14, and 15, which recites substantially the same limitations as claims 1, 3, and 4. Claims 12, 14, and 15 further recite a computer system (Tang Col. 2, lines 55 cont. Col.32: Describes components incorporated into a computer system.) to perform the method steps of claims 1, 3, and 4, respectively, and are therefore rejected on the same premise.
Regarding claim 16, which recites substantially the same limitations as claim 6. Claim 16 further recites a computer system (Tang Col. 2, lines 55 cont. Col.32: Describes components incorporated into a computer system.) to perform the method steps of claim 6, respectively, and are therefore rejected on the same premise.
Regarding claims 17, 19, and 20, which recites substantially the same limitations as claims 1, 3, and 4. Claims 17, 19, and 20, further recite computer program product (Tang Col. 2, lines 55 cont. Col.32: Describes a computer system to store information that is run by a computer program to execute instructions.) to perform the method steps of claim 1, 3, and 4, respectively, and are therefore rejected on the same premise.
Claim(s) 9, 10, and 11 is/are rejected under 35 U.S.C. 103 as being unpatentable over Tang et al. (US 11496493 B2, referred to as Tang), in view of Ferreira Moreno et al. (“US 20220198211 A1”, referred to as Ferreira Moreno), in view of Yang et al. (“Conditional structure generation through graph variational generative adversarial nets”, referred to as Yang), in view of Tang et al. (US 11496493 B2, referred to as Tang), in view of Sun et al. (US 20230351215 A1, referred to as Sun), in view of Cai et al. (“Time in a box: advancing knowledge graph completion with temporal scopes.”, referred to as Cai), in view of Knuff et al. (US 20220188654 A1, referred to as Knuff).
Regarding claim 9, Tang in view of Ferreira Moreno, in view of Yang in view of Sun, in view of Cai teaches the computer-implemented method of claim 1.
Sun further teaches each graph of the training graphs spans a full chronological sequence decomposing into two contiguous chronological sequences, including a first sequence followed by a second sequence ([0213]: Describes that the training procedure uses snapshots of the dynamic graph as input, corresponding to a first chronological subsequence.; [0235]: Describes an evaluation loop “trains on the first t snapshots” of every dynamic graph and treats the remaining snapshots as the next contiguous block, thereby splitting every full timeline into a second chronological subsequence).
Although Tang in view of Ferreira Moreno, in view of Yang in view of Sun, in view of Cai teaches each graph of the training graphs spans a full chronological sequence decomposing into two contiguous chronological sequences, including a first sequence followed by a second sequence, it does not teach the conditional generative model is implemented as a variational autoencoder by the graph neural network, the latter including two input channels consisting of a first input channel and a second input channel.
Knuff teaches the conditional generative model is implemented as a variational autoencoder by the graph neural network, the latter including two input channels consisting of a first input channel and a second input channel (FIG. 20, [0160-0161]: “The exemplary model described herein is a variational autoencoder” these describe a VAE whose encoder/decoder blocks are graph neural network components.; [0143]: Describes that the VAE takes a node-feature matrix and an edge-feature tensor as two distinct matrices, corresponding to two separate input channels.).
It would have been obvious to one of ordinary skill in the art at the time of the claimed invention to modify Tang in view of Ferreira Moreno, in view of Yang in view of Sun, in view of Cai’s input channels with Knuff’s graph based variational autoencoder architecture. Doing so would produce clearer temporal conditioning with routine data-pipeline wiring.
Regarding claim 10, Tang in view of Ferreira Moreno, in view of Yang in view of Sun, in view of Cai teaches, in view of Knuff teaches the computer-implemented method of claim 9.
Sun teaches, wherein training the model on the set of training graphs further comprises: comprises: extracting a first data from the chronological vertices and the associated transient attributes corresponding to the first sequence (SUN [0088]: Describes each snapshot as time varying weights/attributes and states that the first database of intermediate vectors is pulled from those vertices across the first sequence)
Although sun teaches each graph of the training graphs spans a full chronological sequence decomposing into two contiguous chronological sequences, including a first sequence followed by a second sequence, it does not teach the conditional generative model is implemented as a variational autoencoder by the graph neural network, the latter including two input channels consisting of a first input channel and a second input channel.
Knuff further teaches extracting a second data from the sole transient attributes associated with the chronological vertices corresponding to the second sequence ([0142]: Describes the stored attributes are in a node feature matrix N even when link structure is omitted, corresponding to passing only transient attributes without edges.); and inputting the first data into a first input channel and the second data into a second input channel extracted, respectively, for the variational autoencoder to learn to reconstruct a representation of said each graph ([0143]: Describes that two matrices N and A being supplied to the VAE.).
It would have been obvious to one of ordinary skill in the art at the time of the claimed invention to modify Sun’s vertex and attribute data with attribute only data of Knuff. Doing so would allow for selection of readily available data for pre-processing decision making.
Regarding claim 11, Tang in view of Ferreira Moreno, in view of Yang in view of Sun, in view of Cai teaches, in view of Knuff teaches the computer-implemented method of claim 10.
Knuff teaches wherein: the variational autoencoder includes an encoder and a decoder ([0160]: Describes how the3 VAE incorporates a MPNN encoder and is also used to decode.); the encoder is designed to encode input data in a latent space representation in an inner layer block, while the decoder is designed to decode data from the inner layer block; and the first input channel connects to the encoder, while the second channel connects to the inner layer block ([0143]: Describes that the node feature matrix N is input to the encoder during message passing.; [0163] Describes latent samples being “ passed through a sequence of dense layers 2303a-n and subsequently processed via two different matrices” including the edge feature tensor, placing the second matrix/channel at the latent/decoder of the VAE.).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. See attached PTO-892 for additional art including:
US 20180039894 A1: temporal prediction
US 20180032506 A1: dynamic graph concept
US 20210374499 A1: loss function and GNN processing
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DONALD T RODEN whose telephone number is (571)272-6441. The examiner can normally be reached Mon-Thur 8:00-5:00 EST.
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, Omar Fernandez Rivas can be reached at (571) 272-2589. 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.
/D.T.R./Examiner, Art Unit 2128
/OMAR F FERNANDEZ RIVAS/Supervisory Patent Examiner, Art Unit 2128
Prosecution Insights