Entity-centric knowledge discovery
A user interface of an entity-centric knowledge discovery system presents an entity profile including a mix of structured and unstructured data relating to an entity. As a user refines the entity profile based on information gathered from various sources, the changing entity profile can be used as a substantially continuously updating query to search, retrieve, and rank new and pertinent information specifically relevant to the profiled entity. The platform described herein provide an active loop for refining an entity description and searching for additional information in which human input and machine-based algorithms can cooperate to more quickly build a comprehensive description of an entity of interest.
1. A computer program product comprising computer executable code embodied in a non-transitory computer readable medium that, when executing on one or more computing devices, performs the steps of:
providing a human-in-the-loop interface for facilitating entity resolution and interactively building curated knowledge from unstructured and semi-structured data sources;
causing display of a document representation in a first window of the human-in-the-loop interface, wherein:
the document representation comprises an entity profile containing a description of an entity; and
the description of the entity comprises an attribute having a value;
causing display of one or more search results in a second window of the human-in-the-loop interface, wherein the one or more search results are:
identified with a machine-based algorithm;
from one or more of the unstructured and semi-structured data sources containing a mention with a predetermined likelihood of being associated with the entity profile in the document representation displayed in the first window; and
displayed based on the mention in the one or more of the unstructured and semi-structured data sources being semantically related to the entity profile in the document representation in the first window;
receiving a user selection associated with one of the one or more search results;
in response to the user selection associated with the one of the one or more search results, facilitating an entity resolution by creating a link between the one of the one or more search results in the second window and the document representation in the first window, thereby providing a change to the document representation; and
causing an updated search result based on the link between the one of the one or more search results and the document representation.
2. The computer program product of claim 1 , wherein causing the updated search result includes presenting the updated search result for display by a computer.
3. The computer program product of claim 1 , wherein causing the updated search result includes presenting the updated search result in response to a user action to save the change.
4. The computer program product of claim 1 , wherein causing the updated search result includes presenting the updated search result in response to a user action requesting the updated search result.
5. The computer program product of claim 1 , wherein causing the updated search result includes presenting an indication of an availability of the updated search result for display by a computer.
6. The computer program product of claim 1 , further comprising code that performs the step of determining a type for the entity, the type having a predetermined schema that organizes a plurality of attributes for the type.
7. The computer program product of claim 1 , further comprising code that performs the step of providing an indication of the updated search result for display by a computer.
8. The computer program product of claim 7 , wherein the indication of the updated search result includes at least one of a presentation of the updated search result in the second window, and a presentation of a notification to a user that the updated search result is available.
9. The computer program product of claim 1 , further comprising code that performs the step of receiving a user request to display the updated search result and responsively presenting the updated search result for display by a computer.
10. The computer program product of claim 1 , wherein each one of the search results is ranked according to a combination of a likelihood that the one of the search results is associated with the description of the entity in the document representation and a dissimilarity measurement quantifying a difference between the one of the search results and the description of the entity.
11. The computer program product of claim 10 , further comprising ranking the one or more search results with a ranking function that varies according to a state of the description of the entity in the document representation.
12. The computer program product of claim 1 , wherein the document representation includes one or more of a hyperlink, an explicit field, and plain text.
13. The computer program product of claim 1 , wherein the entity is selected from a group consisting of: a person, a corporation, an organization, and a chemical composition.
14. The computer program product of claim 1 , wherein the user selection adds the one of the one or more search results to an accumulated context for the entity in the document representation.
15. The computer program product of claim 1 , wherein the second window is responsive to a user selection of information from the first window to update a search result in the second window according to an inferred indication of user interest in the information.
16. A method for accumulating context for human-in-the-loop, computer assisted entity resolution, the method comprising:
providing a human-in-the-loop interface for facilitating entity resolution and interactively building curated knowledge from unstructured and semi-structured data sources;
causing display of a document representation in a first window of the human-in-the-loop interface, wherein:
the document representation comprises an entity profile containing a description of an entity; and
the description of the entity comprises an attribute having a value;
causing display of one or more search results in a second window of the human-in-the-loop interface, wherein the one or more search results are:
identified with a machine-based algorithm;
from one or more of the unstructured and semi-structured data sources containing a mention with a predetermined likelihood of being associated with the entity profile in the document representation displayed in the first window; and
displayed based on the mention in the one or more of the unstructured and semi-structured data sources being semantically related to the entity profile in the document representation in the first window;
receiving a user selection associated with one of the one or more search results;
in response to the user selection associated with the one of the one or more search results, facilitating an entity resolution by creating a link between the one of the one or more search results in the second window and the document representation in the first window, thereby providing a change to the document representation; and
causing an updated search result based on the link between the one of the one or more search results and the document representation.
17. The method of claim 16 , wherein each one of the search results is ranked according to a combination of a likelihood that the one of the search results is associated with the description of the entity in the document representation and a dissimilarity measurement quantifying a difference between the one of the search results and the description of the entity.
18. The method of claim 16 , further comprising ranking the one or more search results with a ranking function that varies according to a state of the description of the entity in the document representation.
19. A system comprising:
a memory storing a document representation comprising an entity profile containing a description of an entity, the description of the entity comprising an attribute having a value;
a search engine that locates possible mentions of the entity in a source of documents by identifying search results with a machine-based algorithm, the search results from one or more of unstructured and semi-structured data sources containing a mention with a predetermined likelihood of being associated with the entity profile in the document representation;
a server configured to provide a human-in-the-loop interface for facilitating entity resolution and interactively building curated knowledge from unstructured and semi-structured data in the source of documents by performing the steps of:
causing display of the document representation in a first window of the human-in-the-loop interface;
causing display of one or more of the search results in a second window of the human-in-the-loop interface based on the mention being semantically related to the entity profile in the document representation;
receiving a user selection associated with one of the search results;
in response to the user selection associated with the one of the search results, facilitating an entity resolution by creating a link between the one of the search results and the document representation, thereby providing a change to the document representation; and
causing an updated search result based on the change to the document representation.
20. The system of claim 19 wherein the search results from one or more of the unstructured and semi-structured data sources containing a mention with a predetermined likelihood of being associated with the description of the entity in the document representation.