# US 12,118,310 B2 · Extracting explainable corpora embeddings > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,118,310 B2 - Document ID: US12118310B2 - Status: Granted - Kind: B2 - Application number: 17/554,931 - Filed: 2021-12-17 - Published: 2023-06-22 - Granted: 2024-10-15 - Earliest claimed: 2021-12-17 - Assignees: Intuit Inc. - Applicants: Intuit Inc. - Inventors: Maria Kissa; Glenn Carter Scott - Web page: https://patentgrep.com/patent/US12118310B2 ## CPC - G06F 16/9535: Search customisation based on user profiles and personalisation - G06F 40/284: Lexical analysis, e.g. tokenisation or collocates - G06F 16/335: Filtering based on additional data, e.g. user or group profiles (filtering in web context G06F16/9535, G06F16/9536) - G06F 16/367: {Ontology} - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract A method extracts explainable corpora embeddings. The method includes constructing a graph with nodes representing terms from a text sequence and edges that include pointwise values generated between pairs of terms. The method further includes generating a rank vector from the graph. Elements of the rank vector correspond to the edges of the graph. A rank value, of the rank values, corresponds to a term from the text sequence. The method further includes selecting the term by comparing the rank vector to a previous rank vector generated for a previous text sequence. ## Claims ### Claim 1 A method comprising: constructing a graph with a plurality of nodes representing a plurality of terms from a text sequence and a plurality of edges comprising a plurality of pointwise values generated between pairs of terms from the plurality of terms, wherein the text sequence is extracted from a session; generating a rank vector from the graph, wherein a plurality of elements of the rank vector correspond to the plurality of edges of the graph, and wherein a rank value, of a plurality of rank values, corresponds to a term, of the plurality of terms, from the text sequence; selecting the term by comparing the rank vector for the session to a previous rank vector generated from a previous text sequence for a previous session; selecting a recommended text sequence using the term; and presenting a link to the recommended text sequence. ### Claim 2 (depends on 1) The method of claim 1, further comprising: presenting the recommended text sequence. ### Claim 3 (depends on 1) The method of claim 1, further comprising: selecting the recommended text sequence by: comparing the term to a plurality of rank vectors for a plurality of text sequences that include the recommended text sequence. ### Claim 4 (depends on 1) The method of claim 1, further comprising: initiating the session from which the text sequence is selected. ### Claim 5 (depends on 1) The method of claim 1, wherein the text sequence comprises text from one or more of a document and a user action description. ### Claim 6 (depends on 1) The method of claim 1, further comprising: selecting a corpus window of the session as the text sequence. ### Claim 7 (depends on 1) The method of claim 1, further comprising: tokenizing words from the text sequence to generate the plurality of terms. ### Claim 8 (depends on 1) The method of claim 1, further comprising: selecting a pair of terms, of the plurality of terms, using a term distance; generating a pointwise value, of the plurality of pointwise values, for the pair of terms using a pointwise mutual information algorithm; and filtering out, from the plurality of edges, one or more edges with pointwise values that do not satisfy a threshold. ### Claim 9 (depends on 1) The method of claim 1, further comprising: generating the rank value of the term using a plurality of ranks of terms within a term distance to the term and a number of terms within the term distance to the term. ### Claim 10 (depends on 1) The method of claim 1, further comprising: comparing the rank vector to the previous rank vector by generating a delta vector from the rank vector and the previous rank vector. ### Claim 11 (depends on 1) The method of claim 1, further comprising: selecting the term by selecting a delta element from a delta vector using a criterion, wherein the delta element corresponds to the term. ### Claim 12 A system comprising: at least one processor; a graph controller executing on the at least one processor to generate a graph; a rank controller executing on the at least one processor to generate a rank vector; a selection controller executing on the at least one processor configured to select a term; and a server application executing on the at least one processor, performing: constructing the graph with a plurality of nodes representing a plurality of terms from a text sequence and a plurality of edges comprising a plurality of pointwise values generated between pairs of terms from the plurality of terms, wherein the text sequence is extracted from a session, generating the rank vector from the graph, wherein a plurality of elements of the rank vector correspond to the plurality of edges of the graph, and wherein a rank value, of a plurality of rank values, corresponds to the term, of the plurality of terms, from the text sequence, selecting the term by comparing the rank vector for the session to a previous rank vector generated from a previous text sequence for a previous session, selecting a recommended text sequence using the term, and presenting a link to the recommended text sequence. ### Claim 13 (depends on 12) The system of claim 12, wherein the server application is further configured for: presenting the recommended text sequence. ### Claim 14 (depends on 12) The system of claim 12, wherein the server application is further configured for: selecting the recommended text sequence by: comparing the term to a plurality of rank vectors for a plurality of text sequences that include the recommended text sequence. ### Claim 15 (depends on 12) The system of claim 12, wherein the server application is further configured for: initiating the session from which the text sequence is selected. ### Claim 16 (depends on 12) The system of claim 12, wherein the text sequence comprises text from one or more of a document and a user action description. ### Claim 17 (depends on 12) The system of claim 12, wherein the server application is further configured for: selecting a corpus window of the session as the text sequence. ### Claim 18 (depends on 12) The system of claim 12, wherein the server application is further configured for: tokenizing words from the text sequence to generate the plurality of terms. ### Claim 19 (depends on 12) The system of claim 12, wherein the server application is further configured for: selecting a pair of terms, of the plurality of terms, using a term distance; generating a pointwise value, of the plurality of pointwise values, for the pair of terms using a pointwise mutual information algorithm; and filtering out, from the plurality of edges, one or more edges with pointwise values that do not satisfy a threshold. ### Claim 20 A method comprising: interacting, by a client application, with a server application to generate a corpus of text sequences recorded by a server, wherein, a graph is constructed with a plurality of nodes representing a plurality of terms from a text sequence and a plurality of edges comprising a plurality of pointwise values generated between pairs of terms from the plurality of terms, wherein the text sequence is extracted from a session, wherein a rank vector is generated from the graph, wherein a plurality of elements of the rank vector correspond to the plurality of edges of the graph, and wherein a rank value, of a plurality of rank values, corresponds to a term, of the plurality of terms, from the text sequence, and wherein the term is selected by comparing the rank vector for the session to a previous rank vector generated from a previous text sequence for a previous session, and wherein a recommended text sequence is selected using the term; and displaying a link to the recommended text sequence.