Goal-directed semantic search
AssigneeParsons Corporation
InventorMark Gerken
Life of this patent
Estimating the term…
Abstract
Embodiments of goal-directed semantic search allow a user to discover how one or more entities (organizations, people, events, places, etc.) are related without having to explicitly define multiple queries to discover those relationships. In those cases in which more than one structure could be used, separate hypotheses are generated, one for each such structure or relationship. Hypotheses are scored during the process.
Claims18
A system for querying a semantic knowledge base (KB), the system comprising:
- at least one processor configured to communicate with the semantic KB,
- data storage capability devoted to the system;
- wherein the at least one processor executed a query method for querying the semantic KB knowledge base, the method comprising,
- receiving a natural language query;
- performing an initial search of the semantic KB;
- based on results of the initial search, performing a split function which employs a goal-directed graph walk algorithm to create candidate solutions (hypotheses) which go beyond a structure defined by the natural language query, wherein the goal-directed graph walk algorithm creates a separate candidate solution for each possible entity relationships without explicitly defining queries to discover those relationships;
- using results from the split function, performing a filter function wherein the filter function takes candidate solutions and returns true when, and only when, a candidate solution within the candidate solutions yields an extracted solution through further application of the split function;
- using results from the filter function, performing an extract function;
- and
- generating a score for each search result and returning at least one search result.
A method for querying a semantic knowledge base (KB), wherein the semantic KB is in communication with one or more processors configured to execute instructions stored on a non-transitory storage medium, the method comprising,
- receiving a natural language query;
- performing an initial search of the semantic KB;
- based on results of the initial search, performing a split function which employs a goal-directed graph walk algorithm to create candidate solutions (hypotheses) which go beyond a structure defined by the natural language query, wherein the goal-directed graph walk algorithm creates a separate candidate solution for each possible entity relationships without explicitly defining queries to discover those relationships;
- using results from the split function, performing a filter function, wherein the filter function takes candidate solutions and returns true when, and only when, a candidate solution within the candidate solutions yields an extracted solution through further application of the split function;
- using results from the filter function, performing an extract function;
- and
- generating a score for each search result and returning at least one search result.
Dependencies are inferred from the claim text.