Structured user graph to support querying and predictions

A system comprising:

• one or more computers and one or more storage devices on which are stored instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
• receiving a query specific to a specific user, wherein the query specific to the user includes one or more terms that represent an entity, an entity type, or an entity attribute;
• identifying a user-specific knowledge graph generated from user-specific content representing past real-world physical activities of the specific user and date ranges representing when the real-world physical activities occurred, captured in one or more computer-implemented services and occurring outside the computer-implemented services, the user-specific knowledge graph comprising nodes that represent entities in the user-specific content, wherein the nodes include at least one timing node that represents a date range, and edges connecting pairs of nodes that each represent a respective relationship, identified from the user-specific content between the entities represented by the respective pairs of nodes,
• wherein an edge between:
• (i) a timing node and (ii) an activity node representing a real-world activity referenced in the user-specific content, indicates that the real-world physical activity occurred during the date range represented by the timing node;
• querying the user-specific knowledge graph to find nodes that represent entities that match the entity, entity type, or entity attribute of the one or more terms of the query;
• determining weights of edges connected to the found nodes in the graph, wherein the weight of each edge between a respective timing node and a respective activity node is based on how many times the real-world activity represented by the activity node is referenced in the user-specific content as occurring during the date range represented by the timing node;
• and
• providing an answer for the query specific to the user based on the determined weights of the edges between the found nodes in the graph.

A non-transitory computer readable medium storing instructions that, when executed by one or more computers, cause the one or more computers to perform operations comprising:

• receiving a query specific to a specific user, wherein the query specific to the user includes one or more terms that represent an entity, an entity type, or an entity attribute;
• identifying a user-specific knowledge graph generated from user-specific content representing past real-world physical activities of the specific user and date ranges representing when the real-world physical activities occurred, captured in one or more computer-implemented services and occurring outside the computer-implemented services, the user-specific knowledge graph comprising nodes that represent entities in the user-specific content, wherein the nodes include at least one timing node that represents a date range, and edges connecting pairs of nodes that each represent a respective relationship, identified from the user-specific content, between the entities represented by the respective pairs of nodes,
• wherein an edge between:
• (i) a timing node and (ii) an activity node representing a real-world physical activity referenced in the user-specific content, indicates that the real-world physical activity occurred during the date range represented by the timing node;
• querying the user-specific knowledge graph to find nodes that represent entities that match the entity, entity type, or entity attribute of the one or more terms of the query
• determining weights of edges connected to the found nodes in the graph, wherein the weight of each edge between a respective timing node and a respective activity node is based on how many times the real-world activity represented by the activity node is referenced in the user-specific content as occurring during the date range represented by the timing node;
• and
• providing an answer for the query specific to the user based on the determined weights of the edges between the found nodes in the graph.