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 Objections
Claim 1 is objected to because of the following informalities: The proposed claim is written in outline form, using boxes and circles to list the steps of the claimed method. This is non-standard form of claim writing. The claim should be written as a single sentence and not an outline.
Appropriate correction is required.
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 following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 1 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 1 recites the limitation "retrieving a topic linked to the first node in the learning path; " in line 4. There is insufficient antecedent basis for this limitation in the claim.
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.
Claim 1 is rejected under 35 U.S.C. 101 because it is directed to an abstract idea.
Regarding claim 1, limitations include the actions of: collecting and organizing information related to a first user in an organization, identifying a path in a graph, guiding a conversation, extracting an idea, comparing information, judging similarity, and contradiction, and storing that information. Those steps constitute an abstract idea directed to a mental process that can be executed by a human mentally or using pen and paper, which is a judicial exception to patent eligibility. A human can create a learning strategy based on the job position of an employee, organize a series of learning task including other coworkers, providing them a list of topics and insights for the conversation, later transcribing, summarizing and annotating the insights of the conversation, to then check how similar and contradictory the conversation was and then update the knowledge associated with the task as needed. The claim recites additional elements such as a generic knowledge graph, but it doesn’t integrate the mental process into a practical application, and the sole purpose of the system is not significantly more than performing the mental steps listed.
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 factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claim 1 is rejected under 35 U.S.C. 103 as being unpatentable over Shakeri; Cirrus et al. (US 20160189034 A1), hereinafter SHAKERI, in view of Desai; Nirav S. et al. (US 11868736 B1), hereinafter DESAI, in view of Glennon; Barry Dean et al. (US 9659052 B1), hereinafter GLENNON, in view of CHEN; Wei-Peng et al. (US 20240160852 A1), hereinafter CHEN, in further view of Raghavendra; Arun Kumar et al. (US 20180365336 A1) hereinafter RAGHAVENDRA.
Regarding claim 1, SHAKERI teaches:
A method comprising: accessing a position of a first user in an organization;
SHAKERI [0042] "In various embodiments, a variety of operations may be performed against a learning graph. For example, the system may determine a user's profile to obtain some context about what type of learner the person is (who am I and what do I want or need to learn), including what content the individual has consumed (learned) in the past and how it was consumed (online, classroom, etc.), who is the individual connected to and how are they connected. Accordingly, the system may derive what are the individual is interested in and provide the user with relevant learning content recommendations, for example. As described in more detail below, user profile information may be stored in the learning graph as well (e.g., as attributes of a person node and as attributes or node types of other related nodes). Profile information relevant to determining relevant learning content may include a user's role in an organization, years on the job (or experience level), skills, performance data, goals (e.g., areas they need to develop or improve), topics they post or message about, connections to other individuals in the organization, and a variety of other information which may be extracted from external modules (e.g., an HR database, talent management database, or social collaboration tool) and stored in the learning graphs as illustrated in certain example implementations described herein."
based on the position, identifying a learning path in a knowledge graph for the first user, the learning path comprising a set of connected nodes within the knowledge graph;
SHAKERI [0038] “Features and advantages of the present disclosure include a learning graph that comprises nodes and edges used to store data and other information to enable particular embodiments of a dynamic learning management system. Example embodiments of a learning graph may include nodes corresponding to individuals in an organization, learning content, learning collections (e.g., curriculum), learning goals, and a variety of other useful information. Edges between the nodes comprise association types defining particular relationships between particular nodes. For example, a person node representing an individual may be connected to a content node representing an item of learning content (e.g., a research paper or a video). The edge between the nodes may have an association type “hasTaken”, indicating that the person has consumed (or completed) the learning content. As another example, the edge may have an association type “hasLearningItem” together with an attribute “status,” where a “completed” status may indicate the content has been consumed or an “in progress” status may indicate the content has been started but not finished.”
SHAKERI [0041] "A learning graph may be used to create learning paths (e.g., curriculum) dynamically and automatically for individual learning as opposed to contemporary prescription techniques. Individuals in an organization will be able to discover, using the computer automated techniques described herein, including the learning graph, what they need to learn for their job functions and/or career goals, and the learning content will be provided to them automatically without burdensome internet searching, for example. Accordingly, certain embodiments may be user driven or “bottom up” approaches to connecting content with users, for example."
SHAKERI [0045] “In another embodiment, customized content and learning collections (such as a curriculum) are generated for particular users from the learning graph. As described in more detail below, a variety of algorithms (aka learning strategies) may be applied against information embodied in a learning graph to make recommendations about content to be consumed by particular individuals. Particular learning paths may be based on data in the learning graph data model and impacted by what other similar individuals in an organization have done to obtain knowledge, for example.”
retrieving a topic linked to the first node in the learning path;
SHAKERI [0048] “As one example application of a learning graph, derivation of a learning path may proceed as follows. The system may start by determining a model of a particular learner. Learning goals for a person node may be extracted from the learning graph. Topic nodes representing topics of interest in an organization (described in more detail below) may be extracted from the learning graph. Learning preferences (how does the particular individual like to learn) may be extracted by examining consumed content nodes (learning history) connected to the user in the learning graph. Similar learners may be captured by examining other person nodes that meet predefined characteristics (e.g., similar attributes or connections in the learning graph). In certain example embodiments, one or more predefined content identification strategy algorithms (or “strategies”) may be applied to the learning graph data described above, and other data, to generate customized content and/or curriculum recommendations for a particular individual based on that individuals unique collection of connections and attributes in the learning graph, for example. As time passes and the connections between nodes in the graph change and the association types and attributes around the graph change, the recommendations will also change, creating an adaptive environment for deriving learning content for each user.”
SHAKERI [0043] “FIG. 2A illustrates an example learning graph according to a particular embodiment. As shown in FIG. 2A, the learning graph 200 includes nodes (e.g., person node 208) and edges (e.g., edge 202). In this example, there are multiple different types of nodes, including multiple person nodes (labeled “P”) representing individuals (e.g., in an organization) and multiple content nodes (labeled “C”) representing learning content. In this example, nodes further include learning collection nodes and learning goal nodes. Learning collection nodes (labeled “LC”) represent a collection of learning content as a unit (e.g., a curriculum) and have a plurality of edges to a plurality of content nodes. Learning goal nodes have edges to person nodes and may include edges to learning collection nodes, for example, to represent a goal of an individual to consume a compilation of particular learning content, for example. For instance, in this example, person node 209 has an edge to learning collection node 220, and learning collection node 220 has edges to content nodes 221-224. Accordingly, content associated with content nodes 221-224 is part of a particular learning connection 220, which is associated with an individual represented by person node 209 (e.g., has been, or is being, consumed by the individual). Similarly, as another example, individuals represented by person nodes 207-209 may all have the same learning goal 210 (e.g., where all three individuals are new college hires in the same department). Learning goal 210 in turn has an edge to a learning collection node 211, which in turn has edges to content nodes 212-215.”
SHAKERI does not teach but DESAI teaches:
prompting the first user and a second user to discuss the topic during a guided conversation;
DESAI (Col 7 line 57 to Col 8 line 11) “In some embodiments, the ideation module 216 programmatically associates the data structures associated with a given conversation to produce a “topic cloud” that is representative of a graph data structure. Documenting conversations in this manner provides several notable benefits. First, participants can “reach into” the topic cloud and then identify a concept to be further elaborated on. Accordingly, concepts can be expanded or extended over time, as additional conversations occur. Second, the ideation module 216 can act as a guided agent by posing questions that are intended to provoke further discussion or disclosure. When acting as a guided agent, the ideation module 216 can integrate aspects of conversational artificial intelligence, spatial computing, and context-relevant metadata as further discussed below, so as to gain insights into the conversation being guided. And third, the ideation module 216 can permit data structures that are representative of documented conversations to be readily stored and exported. As an example, the data structure that is populated with information derived from a conversation may be exportable as an XML file whose contents can be visualized in various data visualization tools.”
DESAI (Col 2 line 64 to Col 3 line 16) “Documenting conversations in this manner provides several notable benefits. First, a participant may be able to “reach into” a topic cloud and then identify a concept that she would like to further elaborate on. Accordingly, concepts may be expanded or extended over time, as additional conversations occur. Second, the collaboration platform can guide a conversation by prompting participants to answer related questions, so as to obtain information related to different aspects of a given concept. As an example, if the conversation is determined to relate to a new company, the collaboration platform may ask the participants whether there are any known competitors. Responses to this query can be further cataloged in the same “knowledge object.” And third, the collaboration platform may permit data structures that are representative of documented conversations to be readily stored and exported. As an example, the data structure that is populated with information derived from a conversation may be exportable as an Extensible Markup Language (“XML”) file whose contents can be visualized in various data visualization tools.”
generating a transcript of the guided conversation;
DESAI (Col 2 lines 50 to 63) “Introduced here is a computer program that is representative of a software-implemented collaboration platform (or simply “collaboration platform”) that is designed to facilitate conversations in virtual environments, document those conversations, and analyze those conversations, all in real time. As further discussed below, the collaboration platform can include or integrate tools for turning ideas—expressed through voice—into templatized, metadata-rich data structures called “knowledge objects.” Discourse throughout a conversation can be converted into a transcription (or simply “transcript”), parsed to identify topical shifts, and then segmented based on the topical shifts. Together, the various topics discussed over the course of a conversation may be called a “topic listing,” “topic cloud,” or “mind map.””
DESAI (Col 18 lines 21 to 40) “FIGS. 8A-D illustrate a virtual environment at different stages of an ideation session in which three topics are to be discussed. As shown in FIG. 8A, the virtual environment can initially include avatars for the participants, as well as an agenda that lists the topics to be discussed. As the speakers speak over the course of the conversation, the collaboration platform can obtain a transcript as discussed above. The transcript, or snippets thereof, may be posted for review as shown in FIG. 8B. Further, summaries of the topics can be posted for review. In FIG. 8B, a summary has been added beneath the first topic. This summary can be extracted, inferred, or otherwise derived from the transcript (and more specifically, the segment of the transcript that corresponds to the first topic). FIG. 8C illustrates how a summary can be added beneath the second topic to account for contribution by a second speaker, while FIG. 8D illustrates how a summary can be added beneath the third topic to account for contribution by a third speaker. As discussed above, these summaries can be output by a semantic analyzer (e.g., semantic analyzer 410 of FIG. 4).”
based on the transcript, extracting a captured insight discussed by the first user and the second user during the guided conversation;
DESAI (Col 18 line 61 to Col 19 line 5) “FIG. 9 illustrates an example of a transcript, as can be provided to the collaboration platform by a transcription service. As can be seen in FIG. 9, the transcription service can distinguish between different participants, allowing for improved knowledge of which participant contributed which concepts. Various analytics can also be output by the transcription service. As an example, topics can be identified based on a semantic analysis of the transcript. These topics may be representative of insights into the conversation, and therefore can be posted in the virtual environment for review by the participants, such that the participants can readily determine which topics have been discussed.”
DESAI (Col 7 line 57 to Col 8 line 11) “In some embodiments, the ideation module 216 programmatically associates the data structures associated with a given conversation to produce a “topic cloud” that is representative of a graph data structure. Documenting conversations in this manner provides several notable benefits. First, participants can “reach into” the topic cloud and then identify a concept to be further elaborated on. Accordingly, concepts can be expanded or extended over time, as additional conversations occur. Second, the ideation module 216 can act as a guided agent by posing questions that are intended to provoke further discussion or disclosure. When acting as a guided agent, the ideation module 216 can integrate aspects of conversational artificial intelligence, spatial computing, and context-relevant metadata as further discussed below, so as to gain insights into the conversation being guided. And third, the ideation module 216 can permit data structures that are representative of documented conversations to be readily stored and exported. As an example, the data structure that is populated with information derived from a conversation may be exportable as an XML file whose contents can be visualized in various data visualization tools.”
accessing a stored insight linked to the first node;
DESAI (Col 17 line 56 to Col 18 line 9) “For each of the multiple topics, the collaboration platform can populate information derived from the corresponding transcript segment into a first data structure (step 705), populate metadata associated with the corresponding transcript segment into a second data structure (step 706), and link the first and second data structures together, so as to produce a graph data structure for that topic (step 707). The metadata can specify, for example, a date of the conversation, a time of the conversation, identifiers of the multiple participants who engaged in the conversation, locations of the multiple participants who engaged in the conversation, an agenda of the conversation, a document referenced in the conversation, or any combination thereof. As mentioned above, the metadata may only include identifiers for those participants that actually participated in a conversation segment that is associated with the corresponding transcript segment. The collaboration platform can store the multiple graph data structures—each of which is representative of the knowledge object produced for the corresponding topic—produced for the multiple topics in a graph database (step 708).”
DESAI (Col 11 line 60 to Col 12 line 22) “During a conversation, the participants 406 will speak and perform actions in the virtual environment 402. These actions can be documented and recorded by the collaboration platform on a continuous “scrolling tape” 424 that is representative of a “current knowledge object.” The term “current knowledge object” can be used to refer to a knowledge object that is currently in an active editing state. The term “past knowledge object” can be used to refer to a knowledge object that was previously in an active editing state and is currently in a locked state with data stored therein, while the term “future knowledge object” can be used to refer to a knowledge object that has yet to be in an active editing state and is currently largely or entirely devoid of data. As further discussed below, the collaboration platform can record content of the conversation and contextual metadata related to the conversation into the current knowledge object. For example, the current knowledge object can include: A list of participants who are engaged in the conversation or who have contributed to the corresponding topic; Outputs from processing of the voice data 404, such as the transcript 414, topic sets 416, insights such as questions and action items 418, and summary 420; and Contextual metadata such as the geographical location of each participant, the local date and time of each participant, the weather conditions experienced by each participant, and the audio environment (e.g., background music, noise levels, etc.) experienced by each participant.”
DESAI (Col 2 lines 50 to 63) “Introduced here is a computer program that is representative of a software-implemented collaboration platform (or simply “collaboration platform”) that is designed to facilitate conversations in virtual environments, document those conversations, and analyze those conversations, all in real time. As further discussed below, the collaboration platform can include or integrate tools for turning ideas—expressed through voice—into templatized, metadata-rich data structures called “knowledge objects.” Discourse throughout a conversation can be converted into a transcription (or simply “transcript”), parsed to identify topical shifts, and then segmented based on the topical shifts. Together, the various topics discussed over the course of a conversation may be called a “topic listing,” “topic cloud,” or “mind map.”
storing the transcript of the guided conversation and a summary of the captured insight
DESAI (Col 2 line 64 to Col 3 line 16) “Documenting conversations in this manner provides several notable benefits. First, a participant may be able to “reach into” a topic cloud and then identify a concept that she would like to further elaborate on. Accordingly, concepts may be expanded or extended over time, as additional conversations occur. Second, the collaboration platform can guide a conversation by prompting participants to answer related questions, so as to obtain information related to different aspects of a given concept. As an example, if the conversation is determined to relate to a new company, the collaboration platform may ask the participants whether there are any known competitors. Responses to this query can be further cataloged in the same “knowledge object.” And third, the collaboration platform may permit data structures that are representative of documented conversations to be readily stored and exported. As an example, the data structure that is populated with information derived from a conversation may be exportable as an Extensible Markup Language (“XML”) file whose contents can be visualized in various data visualization tools.”
DESAI (Col 18 lines 21 to 40) “FIGS. 8A-D illustrate a virtual environment at different stages of an ideation session in which three topics are to be discussed. As shown in FIG. 8A, the virtual environment can initially include avatars for the participants, as well as an agenda that lists the topics to be discussed. As the speakers speak over the course of the conversation, the collaboration platform can obtain a transcript as discussed above. The transcript, or snippets thereof, may be posted for review as shown in FIG. 8B. Further, summaries of the topics can be posted for review. In FIG. 8B, a summary has been added beneath the first topic. This summary can be extracted, inferred, or otherwise derived from the transcript (and more specifically, the segment of the transcript that corresponds to the first topic). FIG. 8C illustrates how a summary can be added beneath the second topic to account for contribution by a second speaker, while FIG. 8D illustrates how a summary can be added beneath the third topic to account for contribution by a third speaker. As discussed above, these summaries can be output by a semantic analyzer (e.g., semantic analyzer 410 of FIG. 4).”
It would have been obvious to someone of ordinary skill in the art before the effective filling date of the claimed invention to include in the teachings of SHAKERI the capability to proportionate knowledge objects or topics related to a learning goal for a guided conversation, as well as obtain transcripts, summaries and insights of the conversation to update the knowledge updates. The benefit and motivation of such modification is discussed by DESAI in the following portion: DESAI (Col 3 lines 17 to 42) “Documenting conversations in virtual environments in the manner set forth below also provides technical benefits over traditional approaches to documentation or recordation. As an example, documenting the various topics that are discussed over the course of a conversation in the form of “knowledge objects” allows interconnections to be programmatically formed, for example, between related topics. These interconnections not only allow relationships to be more consistently established, but also allow relationships to be surfaced, for example, in response to receiving a query for topics related to a given topic. Simply put, documenting the conversation via interconnected “knowledge objects” leads to more consistency, as the conversation can be represented in the form of a graph data structure as discussed below, and improved searchability (e.g., fewer computational resources consumed, less time taken). Through the approach set forth below, spatial information can also be added to conversational intelligence in a way that adds clarify, for example, in who participants are addressing, the level of engagement of those participants, etc. As an example, information regarding the gaze of participants can be used to improve pronoun understanding. As another information, information regarding the gaze of participants can be used to improve engagement understanding, and therefore which participants may be interested in further conversation on a given topic.”
SHAKERI in view of DESAI does not teach, but GLENNON teaches:
calculating a similarity score between the stored insight and the captured insight;
GLENNON (Col 8 lines 1 to 3) “FIG. 5 is a flow diagram illustrating an exemplary method for comparing relation-side pairs between new objects and objects existing in the knowledge base. ...”
GLENNON (Col 13 lines 48 to 52) “When these scores are strong, the method may exit early and generate an output. For example, if there is a high conflict score (indicating inconclusive results) or if weight is high with a high positivity or negativity and a low conflict (indicating strongly meaningful results).”
calculating a contradiction score between the stored insight and the captured insight;
GLENNON (Col 13 lines 35 to 47) “In an aspect, a similarity score and a conflict score may be calculated based on the positive score, negative score, and weight, illustrated as block 902. For example, the similarity and conflict scores may be calculated as follows:
Similarity score=0.5 if weight<minimum threshold (i.e. not enough information)
Otherwise similarity score=(1+positivity−negativity)/weight
Conflict score=(max(positivity,negativity)−abs(positiviy−negativity))/weight, or (min/weight)”
in response to the similarity score exceeding a similarity threshold and in response to the
GLENNON (Col 13 lines 48 to 52) “When these scores are strong, the method may exit early and generate an output. For example, if there is a high conflict score (indicating inconclusive results) or if weight is high with a high positivity or negativity and a low conflict (indicating strongly meaningful results).”
GLENNON (Col 13 lines 53 to 67) “In an aspect, the similarity score is a comparison vector based on similarity and conflict scores, and is compared to a threshold score, illustrated as block 904, for example, a lower threshold limit, illustrated as block 906, and an upper threshold limit, illustrated as block 908. When the similarity score is lower than the lower threshold limit, illustrated as block 910, the objects may be output as being distinct, illustrated as block 912. When the similarity score is higher than the upper threshold limit, illustrated as block 914, the objects may be output as matching, illustrated as block 916. Further, when the similarity score is between the upper and lower threshold limits, illustrated as block 918, the objects may be output as indicating a manual review, illustrated as block 920. The thresholds may be adjusted as desired when merging objects.”
confirming correspondence between the captured insight and the stored insight; and
GLENNON (Col 13 lines 53 to 67) “In an aspect, the similarity score is a comparison vector based on similarity and conflict scores, and is compared to a threshold score, illustrated as block 904, for example, a lower threshold limit, illustrated as block 906, and an upper threshold limit, illustrated as block 908. When the similarity score is lower than the lower threshold limit, illustrated as block 910, the objects may be output as being distinct, illustrated as block 912. When the similarity score is higher than the upper threshold limit, illustrated as block 914, the objects may be output as matching, illustrated as block 916. Further, when the similarity score is between the upper and lower threshold limits, illustrated as block 918, the objects may be output as indicating a manual review, illustrated as block 920. The thresholds may be adjusted as desired when merging objects.”
It would have been obvious to someone of ordinary skill in the art before the effective filling date of the claimed invention to include in the teachings of SHAKERI in view of DESAI the capability to perform a similarity analysis compared to a second analysis to compare the insights of the one guided conversation to the insights stored in the knowledge object. Regarding the second analysis, the one presented in GLENNON is a conflict analysis, the contradiction analysis would be explained by CHEN further ahead. The benefit and motivation of such modification is discussed by GLENNON in the following portion: GLENNON (Col 1 line 55 to Col 2 line 5) “When a new data source is imported into a knowledge database(s), such as a knowledge database that aggregates data from numerous different sources, the new data source may include data objects that represent objects that are already stored in the knowledge database(s). The new data objects may be entirely repetitive of information already in the knowledge base or may contain additional information that may be merged with the existing data. The knowledge database(s) may include information about the data objects in a structured form. In this aspect, the present disclosure describes systems and methods that can recognize and/or resolve when data objects are the same or similar, using structured information about each object. In an aspect, the present disclosure describes systems and methods that can determine what and how similar a new source data object is to data objects that are already present in the knowledge database(s) and aggregate the information in the objects if appropriate.”
SHAKERI in view of DESAI in view of GLENNON does not teach, but CHEN teaches:
calculating a contradiction score between the stored insight and the captured insight;
CHEN [0077] “FIG. 6 is a diagram that illustrates an exemplary scenario for computing the final NLI score for each sentence in the set of texts, in accordance with at least one embodiment described in the present disclosure. FIG. 6 is described in conjunction with elements from FIG. 1, FIG. 2, FIG. 3, FIG. 4, and FIG. 5. With reference to FIG. 6, there is shown an exemplary scenario 600. The exemplary scenario 600 may include an input sentence 602, a positive statement (K r) 604, a negative statement (K.sub.n) 606, a score for contradiction relation label based on positive statement (contradiction_p) 608, a score for neutral relation label based on positive statement (neutral_p) 610, a score for entailment relation label based on positive statement (entailment_p) 612, a score for contradiction relation label based on negative statement (contradiction_n) 614, a score for neutral relation label based on negative statement (neutral_n) 616, a score for entailment relation label based on negative statement (entailment_n) 618, a negative label 620, a neutral label 622, and a positive label 624. The contradiction_p 608, the neutral_p 610, the entailment_p 612, the contradiction_n 614, the neutral_n 616, the entailment_n 618 may be the intermediate NLI scores. The negative label 620, the neutral label 622, and the positive label 624 may be the final NLI score.”
CHEN EQUATION (1): “Final_score(j==NEGATIVE)=w.sub.p.Math.avg_i(avg_k.sub.p(NLI_score(i,j,k.sub.p,CON))+w.sub.n.Math.avg_i(avg_k.sub.n(NLI_score(i,j,k.sub.n,ENT)) (1) “
CHEN EQUATION (3): “Final_score(j==POSITIVE)=w.sub.p.Math.avg_i(avg_k.sub.p(NLI_score(i,j,k.sub.p,ENT))+w.sub.n.Math.avg_i(avg_k.sub.n(NLI_score(i,j,k.sub.n,CON)) (3)”
CHEN [0079] “Here, “w.sub.p” may be weights for a positive statement, “w.sub.n” may be weights for a negative statement, “CON” may represent contradiction, “NEU” may represent neutral, and “ENT” may represent entailment. With reference to FIG. 6, the processor 204 may determine whether the input sentence 602 (i.e., “j”) is a positive statement or a negative statement. Each input sentence 602 (i.e., “j”), i.e., a premise sentence, followed by the positive statement 604 (i.e., “K.sub.p”), i.e., a hypothesis sentence, might determine the intermediate NLI scores as the contradiction_p 608, the neutral_p 610, or the entailment_p 612. …”
contradiction score falling below a contradiction threshold
CHEN [0094] “At block 904, the determined NLI prediction score, of each sentence of the set of sentences, over each of the set of predefined NLI classes may be compared with a first predefined threshold. In an embodiment, the processor 204 may be configured to compare the determined NLI prediction score, of each sentence of the set of sentences, over each of the set of predefined NLI classes, with the first predefined threshold. The first predefined threshold may be a minimum value that the class of the set of predefined NLI classes may have in order to determine the overall NLI score as the class.”
CHEN [0072] “… The inference relation may be one of, but not limited to, an entailment relation, a contradiction relation, or a neutral relation. The intermediate NLI score associated with the selected first sentence may be determined based on the determined inference relation. The intermediate NLI score may be a tuple of probabilities of three labels, such as, an entailment relation label, a contradiction relation label, or a neutral relation label. For example, if the determined inference relation corresponds to the contradiction relation, then the intermediate NLI score of contradiction relation label might be highest among three possible labels. If the determined inference relation is the neutral relation, then the intermediate NLI score of neutral relation label might be highest among three possible labels. Further, if the determined inference relation is the entailment relation, then the intermediate NLI score of entailment relation label might be highest among three possible labels.”
It would have been obvious to someone of ordinary skill in the art before the effective filling date of the claimed invention to include in the teachings of SHAKERI in view of DESAI in view of GLENNON the capability to include in the comparison of the insights a contradiction analysis. The contradiction score being in CHEN part of the analysis to determine the multi-dimensional natural language inference score in order to determine if a relationship between 2 elements is an entailment relationship, a neutral relationship or a contradiction relationship (which is the element that would be incorporated to the teachings of SHAKERI in view of DESAI in view of GLENNON). The benefit and motivation of such modification is discussed by CHEN in the following portion: CHEN [0038] “The pre-trained NLI model 114, also known as recognizing textual entailment (RTE) model, may be used to infer a relationship between two short and ordered texts. The two short and ordered texts may be a premise and a hypothesis. The inference relationship may be one of entailment, contradiction, or neutral. In an example, if the premise is “A boat sank in the Pacific Ocean.” and the hypothesis is “A boat sank in the Atlantic Ocean.”, then the inference relation between the premise and the hypothesis may be contradiction as the premise may contradict the hypothesis. In an example, if the premise is “Ruth Bader Ginsburg was appointed to the US Supreme Court.” and the hypothesis is “I had a sandwich for lunch today.”, then the inference relation between the premise and the hypothesis may be neutral, as the premise and the hypothesis may not be related to each other. In an example, if the premise is “Two women are embracing while holding to go packages.” And the hypothesis is “Two woman are holding packages.”, then the inference relation between the premise and the hypothesis may be entailment, as the premise and the hypothesis may conform with each other.”
SHAKERI in view of DESAI in view of GLENNON in view of CHEN does not teach, but RAGHAVENDRA teaches:
as a new version of the stored insight at the first node.
RAGHAVENDRA [0061] “In an embodiment, the temporal graph stores one or more versions of the same graph element. When a graph element is modified or removed, the previous version of the graph element remains. A subsequent version of the graph element is created that reflects the modification or deletion. The subsequent version is associated with time information indicating when the subsequent version was created. In other words, each version of a graph element corresponds to the state of the graph element at a particular point in time.”
RAGHAVENDRA [0065] “FIG. 2 illustrates a visual representation of an example temporal graph 200. Temporal graph 200 comprises vertices v1, v2, v3, v4, v5, and v6, and edges e1, e2, e3, e4, and e5. Temporal graph 200 comprises one or more versions of each graph element. Each version of a graph element is associated with one of time T1, time T2, or time T3.”
It would have been obvious to someone of ordinary skill in the art before the effective filling date of the claimed invention to include in the teachings of SHAKERI in view of DESAI in view of GLENNON in view of CHEN the capability to create and store versions of the graph, not only the most updated version of the graph but also previous versions. The benefit and motivation of such modification is discussed by RAGHAVENDRA in the following portion: RAGHAVENDRA [0174] “Temporal graph structure and graph element properties change over time. Historical data retrieval comprises retrieving graph element history from between any two points in time. Retrieving historical data for graph elements includes retrieving graph elements from a period of time that satisfy particular criteria or conditions. For example, graph elements that were created, modified, deleted, and/or existed during a particular time range may be retrieved. As another example, graph elements with particular property values during a particular time range may be retrieved. Additionally, for a given vertex, all of the inbound and/or outbound edges that were created, deleted, modified, and/or existed within a specified time period may be retrieved.”
Summary of references used as prior art
SHAKERI is used as the main reference, it provides a base for a learning graph related to a company knowledge for employees, where it connects nodes to knowledge topics. DESAI provides a guided conversation between 2 individuals and store transcripts, summaries and insights and condenses them into knowledge objects. GLENNON provides a multi parameter framework to compare the stored knowledge to the knowledge from the conversation, this framework includes the similarity score and a second parameter. CHEN is used to modify the invention to replace the second parameter with a contradiction score. RAGHAVENDRA provides graph versioning, to store and modify information stored in the nodes of a graph.
Pertinent art not relied upon in this rejection
Adinarayan; Geetha et al. (US 11227018 B2): Auto generating reasoning query on a knowledge graph. Alternative reference to represent relationships in a knowledge graph for semantic analysis and reasoning.
KIM; Hyeongsik et al. (US 20220351716 A1): SYSTEM AND METHOD FOR A PERSONALIZED DIALOGUE SYSTEM USING KNOWLEDGE-BASED ENTITY SERVICES. Alternative reference to represent relationships in a knowledge graph for semantic analysis and reasoning.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to HECTOR J. CRESPO FEBLES whose telephone number is (571)272-4512 and email is hcrespofebles@uspto.gov. The examiner can normally be reached Mon - Fri 7:30 - 5:00.
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, Daniel Washburn can be reached at (571) 272-5551. 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.
/HECTOR J. CRESPO FEBLES/Examiner, Art Unit 2657
/DANIEL C WASHBURN/Supervisory Patent Examiner, Art Unit 2657
