# US 11,907,643 B2 · Dynamic persona-based document navigation > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,907,643 B2 - Document ID: US11907643B2 - Status: Granted - Kind: B2 - Application number: 17/732,920 - Filed: 2022-04-29 - Published: 2023-11-02 - Granted: 2024-02-20 - Earliest claimed: 2022-04-29 - Assignees: Adobe Inc. - Applicants: Adobe Inc. - Inventors: Sumit Shekhar; Tanvi V. Karandikar; Nethraa Sivakumar; Shelly Jain; Himanshu Maheshwari; Vinay Aggarwal; Navita Goyal - Web page: https://patentgrep.com/patent/US11907643B2 ## CPC - G06F 40/134: Hyperlinking - G06F 40/137: Hierarchical processing, e.g. outlines - G06F 40/20: Natural language analysis (semantic analysis of natural language G06F40/30) - G06F 40/258: Heading extraction; Automatic titling; Numbering - G06F 40/279: Recognition of textual entities - G06F 40/295: Named entity recognition - G06N 3/0455: Auto-encoder networks; Encoder-decoder networks - G06N 3/0895: Weakly supervised learning, e.g. semi-supervised or self-supervised learning - G06N 20/00: Machine learning - G06V 30/413: Classification of content, e.g. text, photographs or tables - G06V 30/416: Extracting the logical structure, e.g. chapters, sections or page numbers; Identifying elements of the document, e.g. authors - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING ## Abstract Embodiments of the technology described herein are directed to a persona-specific navigation interface for a document. Initially, a user may select a persona associated with a document through a document navigation interface. A machine-learning model may identify an interest within a portion of the document. The interest may be mapped to the persona. A navigation interface that includes a navigable link to the portion of the document is generated and output for display. A user interaction with the navigable link is received. In response to the interaction, the portion of the document corresponding to the navigable link is output for display. ## Claims ### Claim 1 A method comprising: receiving, by a processing device, a selection of a persona associated with a document through a document navigation interface; identifying, using a machine learning model, an interest within a portion of the document; mapping the interest to the persona; generating a navigation interface that includes a navigable link to the portion of the document, wherein the navigation interface is output for display, wherein the navigable link is phrased as a question that is automatically generated by a natural language processing model having an encoder-decoder architecture, wherein the natural language processing model is trained using a correctness reward that is positive when a generated question contains an entity found in the linked portion of the document; receiving an interaction with the navigable link; and in response to the interaction, outputting the portion of the document for display. ### Claim 2 (depends on 1) The method of claim 1, wherein the natural language processing model is trained using an answerability award calculated based on whether a generated question is answered in the linked portion of the document, wherein the answerability award is calculated using a classifier. ### Claim 3 (depends on 1) The method of claim 1, wherein the natural language processing model is trained using a coverage reward calculated that is positive when a generated question contains a first set of keywords similar to a second set of keywords found in the linked portion of the document. ### Claim 4 (depends on 1) The method of claim 1, wherein the document navigation interface includes multiple navigable links to different portions of the document and the navigable links are arranged in an order that is different from a native order of occurrence in the document. ### Claim 5 (depends on 1) The method of claim 1, wherein the method further comprises identifying the interest in the portion of the document using a self-supervised contrastive learning framework that adjusts a keyword representation using a TF-IDF (“Term Frequency-Inverse Document Frequency”) score for a keyword associated with the keyword representation. ### Claim 6 (depends on 1) The method of claim 1, wherein the mapping of the interest to the persona is completed using a persona-interest matrix generated by a domain expert for a document type. ### Claim 7 (depends on 1) The method of claim 1, wherein the method includes: associating a table in the document to the portion of the document by finding a similarity between entities in the portion of the document and entities in the table; and associating the table with the interest identified in the portion of the document. ### Claim 8 (depends on 1) The method of claim 1, wherein the navigation interface includes multiple navigable links to different portions of the document and the navigable links are arranged in an order that is different from a native order of occurrence in the document. ### Claim 9 A non-transitory computer-readable medium storing computer-executable instructions that, when executed by a processing device, cause the processing device to perform operations comprising: identifying, using a machine learning model, an interest within a portion of a document; associating an object in the document to the portion of the document by finding a similarity between a first set of entities in the portion of the document and a second set of entities in the object; associating the object with the interest identified in the portion of the document; generating a navigation interface that includes navigable links to portions of the document that are associated with the interest, wherein the navigable links include a first link to the portion of the document and a second link to the object, wherein the navigable link is phrased as a question that is automatically generated by a natural language processing model having an encoder-decoder architecture, wherein the natural language processing model is trained using a coverage reward calculated that is positive when a generated question contains a first set of keywords similar to a second set of keywords found in the linked portion of the document; outputting the navigation interface for display: receiving an interaction with the second link; and in response to the interaction, outputting the object for display. ### Claim 10 (depends on 9) The non-transitory computer-readable medium of claim 9, wherein the method further comprises applying named entity recognition and key phrase extraction to portions of the document and the object. ### Claim 11 (depends on 10) The non-transitory computer-readable medium of claim 10, wherein the method further comprises: adding each unique entity and unique key phrase extracted from the object into a first single string for the object; and adding each unique entity and unique key phrase extracted for a portion into to form a second single string for the portion; generating a similarity score for the first single string and the second single string. ### Claim 12 (depends on 11) The non-transitory computer-readable medium of claim 11, wherein the method further comprises associating the object with the portion because the similarity score is above a threshold similarity. ### Claim 13 (depends on 9) The non-transitory computer-readable medium of claim 9, wherein the method further comprises identifying the interest in the portion of the document using a self-supervised contrastive learning framework that adjusts a keyword representation using a TF-IDF (“Term Frequency-Inverse Document Frequency”) score for a keyword associated with the keyword representation. ### Claim 14 (depends on 9) The non-transitory computer-readable medium of claim 9, wherein the object is a table. ### Claim 15 A system comprising: a memory component; and a processing device coupled to the memory component, the processing device to perform operations comprising: generating a document interface that displays a document content; generating a navigation interface that includes a document navigation aid, wherein the navigation interface includes a navigable link phrased as a question answered by a portion of the document linked to the navigable link, wherein the document interface is output for display adjacent to the navigation interface, wherein the question is automatically generated by a natural language processing model having an encoder-decoder architecture, wherein the natural language processing model is trained using an answerability award calculated based on whether a generated question is answered in the linked portion of the document, wherein the answerability award is calculated using a classifier; receiving an interaction with the navigable link; and in response to the interaction, outputting the portion of the document linked of the navigable link for display on the document interface. ### Claim 16 (depends on 15) The system of claim 15, wherein the method further comprises receiving a selection of a persona associated with the document content through a document interface; identifying, using a machine learning model, an interest within the portion of the document; mapping the interest to the persona. ### Claim 17 (depends on 16) The system of claim 16, wherein the method further comprises identifying the interest in the portion of the document using a self-supervised contrastive learning framework that adjusts a keyword representation using a TF-IDF (“Term Frequency-Inverse Document Frequency”) score for a keyword associated with the keyword representation. ### Claim 18 (depends on 16) The system of claim 16, wherein the mapping of the interest to the persona is completed using a persona-interest matrix generated by a domain expert for a document type. ### Claim 19 (depends on 17) The system of claim 17, wherein the natural language processing model is trained using a correctness reward that is positive when a generated question contains an entity found in the linked portion of the document. ### Claim 20 (depends on 17) The system of claim 17, wherein the natural language processing model is trained using a coverage reward calculated that is positive when a generated question contains a first set of keywords similar to a second set of keywords found in the linked portion of the document.