Abstract: Examples of the present disclosure describe systems and methods for providing textual summaries based on personalized prior knowledge. In examples, a user request for a summary of document or an entity is received. The document or documents associated with the entity are separated into segments and semantic embeddings are created for each segment. The semantic embeddings, a context of the user request, and relevant information available in the requesting user's previous knowledge base are provided as input to a personal knowledge system. Based on the input, the personal knowledge system outputs an indication of the segments that should be summarized. The indicated segments are provided to a summarization system. The summarization system generates a document summary or an entity summary and provides the summary to the user.

Abstract: The disclosure herein describes a system for measuring probability of influence in digital communications to determine if communication content originated in a person's own prior knowledge or new information more recently obtained from interaction with communications of others. An estimated probability a new communication by a first user comes from the same distribution as prior communications of the first user are generated using multidimensional statistics on embeddings representing the communications. A second estimated probability that the new communication comes from the same distribution as communication(s) of a second user that were accessible to the first user are generated. If the second probability is greater than the first probability, the new communication is more likely influenced by exposure of the first user to the second user's communications rather than the first user's own historical knowledge.

Abstract: Technology is disclosed for programmatically identifying topics between a user and a contact of the user using a knowledge graph corresponding to the collaboration of the user and at least one of the user's contacts through various applications or platforms. The knowledge graph of the user includes (1) nodes corresponding to contacts, data objects, and topics extracted from the data objects and (2) edges corresponding to interactions between the contacts and data objects and (b) relationships between the data objects and topics. The knowledge graph is preprocessed in order to prune corresponding contacts, data objects, and/or topics. Following preprocessing of the knowledge graph, a subgraph of the knowledge is generated for each contact in order to identify, prune, and rank the topics for each contact. A number of the highest-ranked topics for each contact can be formatted for presentation to the user to provide an improved user computing experience.

Abstract: The disclosed technology is generally directed to a personalized feed. In one example of the technology, selected key-value pairs from a profile associated with a user are provided. Based on a prompt that includes natural-language text instructions, the selected key-value pairs, and ranked content, a large language model is used to generate: pill prompts associated with the ranked content, such that the pill prompts are information requests that are unique and personalized to have particular relevance to the user based on selected key-value pairs, and a response to each pill prompt such that the response includes content corresponding to the requested information. A content feed is displayed to the user, including displaying selectable pills to the user as part of the displayed content feed such that each selectable pill includes a corresponding pill prompt. The response to the pill prompt that corresponds to the selection is displayed to the user.

Abstract: The disclosed technology is generally directed to a personalized feed. In one example of the technology, selected key-value pairs from a profile associated with a user are provided. Based on a prompt that includes natural-language text instructions, the selected key-value pairs, and ranked content, a large language model is used to generate: pill prompts associated with the ranked content, such that the pill prompts are information requests that are unique and personalized to have particular relevance to the user based on selected key-value pairs, and a response to each pill prompt such that the response includes content corresponding to the requested information. A content feed is displayed to the user, including displaying selectable pills to the user as part of the displayed content feed such that each selectable pill includes a corresponding pill prompt. The response to the pill prompt that corresponds to the selection is displayed to the user.

Abstract: The disclosure herein describes a system for measuring probability of influence in digital communications to determine if communication content originated in a person's own prior knowledge or new information more recently obtained from interaction with communications of others. An estimated probability a new communication by a first user comes from the same distribution as prior communications of the first user are generated using multidimensional statistics on embeddings representing the communications. A second estimated probability that the new communication comes from the same distribution as communication(s) of a second user that were accessible to the first user are generated. If the second probability is greater than the first probability, the new communication is more likely influenced by exposure of the first user to the second user's communications rather than the first user's own historical knowledge.

Abstract: The disclosure herein describes a system for measuring probability of influence in digital communications to determine if communication content originated in a person's own prior knowledge or new information more recently obtained from interaction with communications of others. An estimated probability a new communication by a first user comes from the same distribution as prior communications of the first user are generated using multidimensional statistics on embeddings representing the communications. A second estimated probability that the new communication comes from the same distribution as communication(s) of a second user that were accessible to the first user are generated. If the second probability is greater than the first probability, the new communication is more likely influenced by exposure of the first user to the second user's communications rather than the first user's own historical knowledge.