Systems and methods for entity resolution
Systems and methods are described for identifying a target entity that a user intends to refer to based on input that may include only a partial name of the entity. A system may receive a user's entity search query, which may include a partial entity name, then may determine initial candidate matching entities and associated name match scores for each of the initial candidate matching entities relative to the input string, such as using string matching or fuzzy matching techniques. The system may apply a probability match model to generate an overall match score for individual ones of the initial candidate matching entities, then may generate a response to the entity search query that includes information regarding one or more of the initial candidate matching entities with an overall match score exceeding a threshold.
1 . A computer-implemented method comprising:
receiving an entity search query via user interaction with a user interface, the entity search query including an input string that includes a partial name of an entity for which a user would like to retrieve information, wherein the partial name does not uniquely identify the entity;
determining initial candidate matching entities by at least performing initial string matching relative to a plurality of stored entity records for a plurality of entities;
generating a name match score for each of the initial candidate matching entities relative to the input string, wherein generating the name match score for each of the initial candidate matching entities utilizes at least one of: a string matching approach or a fuzzy matching approach;
generating an overall match score for each of at least a subset of the initial candidate matching entities, wherein the overall match score for each individual candidate matching entity of the subset of the initial candidate matching entities is generated based at least in part on (a) the name match score for the individual candidate matching entity and (b) one or more attributes previously stored in association with the individual candidate matching entity; and
generating a response to the entity search query, the response including information, for display in the user interface, regarding one or more of the initial candidate matching entities having an overall match score exceeding a threshold.
2 . The computer-implemented method of claim 1 further comprising causing display, in the user interface in response to the entity search query, of an initial candidate matching entity having a highest overall match score, wherein the initial candidate matching entity having the highest overall match score is determined by the computing system without receiving any information from the user regarding the initial candidate matching entity other than the partial name.
3 . The computer-implemented method of claim 1 , wherein the one or more attributes comprise one or more of: an industry, credit attributes, firmographic details, past search queries, or entity linkage information.
4 . The computer-implemented method of claim 1 , wherein generating the name match score comprises a fuzzy matching approach that calculates an edit distance between an individual initial candidate matching entity and the input string.
5 . The computer-implemented method of claim 1 , wherein the user represents an insurer that is considering insuring the entity for which the user would like to retrieve information, the computer-implemented method further comprising applying a filter to the initial candidate matching entities based on a likelihood of each of the initial candidate matching entities to be seeking insurance.
6 . The computer-implemented method of claim 1 , wherein the response includes, for display in the user interface, a Business Identification Number (BIN) for at least one of the initial candidate matching entities having an overall match score exceeding the threshold.
7 . The computer-implemented method of claim 1 further comprising, prior to generating the response, reducing an entity search space for the entity search query based on corporate linkage information that associates entities having a parent and child relationship.
8 . A computing system comprising:
a non-transitory data store configured to store executable instructions; and
one or more computer processors configured to execute the executable instructions, wherein execution of the executable instructions causes the one or more computer processors to:
receive an entity search query via user interaction with a user interface, the entity search query including an input string that includes a partial name of an entity for which a user would like to retrieve information, wherein the partial name does not uniquely identify the entity;
determine initial candidate matching entities by at least performing initial string matching relative to a plurality of stored entity records for a plurality of entities;
generate a name match score for each of the initial candidate matching entities relative to the input string, wherein generating the name match score for each of the initial candidate matching entities utilizes at least one of: string matching or fuzzy matching;
generate an overall match score for each of at least a subset of the initial candidate matching entities, wherein the overall match score for each individual candidate matching entity of the subset of the initial candidate matching entities is generated based at least in part on (a) the name match score for the individual candidate matching entity and (b) one or more attributes previously stored in association with the individual candidate matching entity; and
generate a response to the entity search query, the response including information, for display in the user interface, regarding one or more of the initial candidate matching entities having an overall match score exceeding a threshold.
9 . The computing system of claim 8 , wherein execution of the executable instructions further causes the one or more computer processors to cause display, in the user interface in response to the entity search query, of an initial candidate matching entity having a highest overall match score, wherein the initial candidate matching entity having the highest overall match score is determined by the computing system without receiving any information from the user regarding the initial candidate matching entity other than the partial name.
10 . The computing system of claim 8 , wherein the entity search query does not include any geographical location information associated with the entity for which the user would like to retrieve information.
11 . The computing system of claim 8 , wherein execution of the executable instructions further causes the one or more computer processors to apply a client-specific model that scores the initial candidate matching entities based at least in part on attributes of entities that a client associated with the user has interacted with previously.
12 . The computing system of claim 8 , wherein execution of the executable instructions further causes the one or more computer processors to apply natural language processing (NLP) to create vector representations of a plurality of entity names such that a distance between two vector representations corresponds to an extent of similarity between respective entity names represented by the two vector representations.
13 . The computing system of claim 8 , wherein the overall match score is based at least in part on graph-based record linkage data whereby nodes representing entities having a corporate relationship are connected in a graph structure.
14 . The computing system of claim 13 , wherein execution of the executable instructions further causes the one or more computer processors to apply a rollup operation to a top-level entity node in the graph structure with respect to a connected node.
15 . The computing system of claim 8 , wherein the overall match score for an individual candidate matching entity is based at least in part on one or more query patterns associated with search queries previously submitted by the user.
16 . The computing system of claim 8 , wherein the overall match score for an individual candidate matching entity is based at least in part on how unique an individual entity name is relative to entity names of the initial candidate matching entities.
17 . The computing system of claim 8 , wherein the response includes, for display in the user interface, a Business Identification Number (BIN) for at least one of the initial candidate matching entities having an overall match score exceeding the threshold.
18 . The computing system of claim 8 , wherein generating the name match score comprises a fuzzy matching approach that calculates an edit distance between an individual initial candidate matching entity and the input string.
19 . The computing system of claim 8 , wherein the user represents an insurer that is considering insuring the entity for which the user would like to retrieve information, wherein execution of the executable instructions further causes the one or more computer processors to apply a filter to the initial candidate matching entities based on a likelihood of each of the initial candidate matching entities to be seeking insurance.