Method and system for entity extraction and disambiguation

1. A method implemented on a machine having at least one processor, storage, and a communication platform connected to a network for determining an individual associated with an entity name, the method comprising:

generating by a person-centric index system, a person-centric space for a user and a person-centric space for each of other users related to the user, wherein each of the person-centric spaces is built by cross-linking information relevant to the corresponding user obtained from public, semi-private and private spaces, and wherein the information is cross-linked based on cross-linking keys extracted from the information;

determining, via machine learning and based on person-centric information associated with the user stored in the person-centric space for the user, one or more entity types associated with an entity name extracted from the person-centric space for the user, wherein the one or more entity types correspond to all the possible entity types that the entity name refers to, wherein each of the one or more entity types is associated with a confidence score indicating how likely the entity name refers to the entity type;

identifying one or more entity candidates with respect to each of the one or more entity types, wherein each of the one or more entity candidates is an instance of the entity type and is associated with a resolution score indicating how likely the entity name refers to the entity candidate, and wherein the resolution score is calculated by weighting, based on user input or preference, different types of contextual information and textual information that are extracted from the person-centric space for the user based on the entity type;

selecting, by an entity resolution engine, based on the confidence scores and the resolution scores, an entity candidate with respect to one of the one or more entity types to be an individual associated with the entity name, wherein the entity resolution engine is trained to perform the selecting based on the generated person-centric spaces associated with the user and the other users related to the user; and

generating a representation associated with the user based on the selected candidate, wherein the representation facilitates providing a response to a query issued by the user.

2. The method of claim 1 , wherein the information being cross-linked is obtained from at least one of the following: the user's emails, contacts, instant messages, browsing history, call records, and bookmarks.

3. The method of claim 1 , wherein the entity name is extracted from a card to be presented to the user.

4. The method of claim 1 , wherein the entity candidate is selected based on textual metadata associated with the entity name in the person-centric space for the user.

5. A system having at least one processor, storage, and a communication platform connected to a network for determining an individual associated with an entity name, the system comprising:

a person-centric space builder configured for generating by a person-centric index system, a person-centric space for a user and a person-centric space for each of other users related to the user, wherein each of the person-centric spaces is built by cross-linking information relevant to the corresponding user obtained from public, semi-private and private spaces, and wherein the information is cross-linked based on cross-linking keys extracted from the information;

an entity type determiner configured for determining, via machine learning and based on person-centric information associated with the user stored in the person-centric space for the user, one or more entity types associated with an entity name extracted from the person-centric space for the user, wherein the one or more entity types correspond to all the possible entity types that the entity name refers to, wherein each of the one or more entity types is associated with a confidence score indicating how likely the entity name refers to the entity type;

an entity candidate determiner configured for identifying one or more entity candidates with respect to each of the one or more entity types, wherein each of the one or more entity candidates is an instance of the entity type and is associated with a resolution score indicating how likely the entity name refers to the entity candidate, and wherein the resolution score is calculated by weighting, based on user input or preference, different types of contextual information and textual information that are extracted from the person-centric space for the user based on the entity type;

an entity individual selector configured for selecting, based on the confidence scores and the resolution scores, an entity candidate with respect to one of the one or more entity types to be an individual associated with the entity name, wherein the entity individual selector is trained to perform the selecting based on the generated person-centric spaces associated with the user and the other users related to the user; and

an entity relationship determiner configured to generate a representation associated with the user based on the selected candidate, wherein the representation facilitates providing a response to a query issued by the user.

6. The system of claim 5 , wherein the information being cross-linked is obtained from at least one of the following: the user's emails, contacts, instant messages, browsing history, call records, and bookmarks.

7. The system of claim 5 , wherein the entity name is extracted from a card to be presented to the user.

8. The system of claim 5 , wherein the entity candidate is selected based on textual metadata associated with the entity name in the person-centric space for the user.

9. A machine-readable, non-transitory and tangible medium having information recorded thereon for determining an individual associated with an entity name, the information, when read by the machine, causes the machine to perform the following:

generating by a person-centric index system, a person-centric space for a user and a person-centric space for each of other users related to the user, wherein each of the person-centric spaces is built by cross-linking information relevant to the corresponding user obtained from public, semi-private and private spaces, wherein the information is cross-linked based on cross-linking keys extracted from the information;

determining, via machine learning and based on person-centric information associated with the user stored in the person-centric space for the user, one or more entity types associated with an entity name extracted from the person-centric space for the user, wherein the one or more entity types correspond to all the possible entity types that the entity name refers to, wherein each of the one or more entity types is associated with a confidence score indicating how likely the entity name refers to the entity type;

identifying one or more entity candidates with respect to each of the one or more entity types, wherein each of the one or more entity candidates is an instance of the entity type and is associated with a resolution score indicating how likely the entity name refers to the entity candidate, and wherein the resolution score is calculated by weighting, based on user input or preference, different types of contextual information and textual information that are extracted from the person-centric space for the user based on the entity type;

selecting, by an entity resolution engine, based on the confidence scores and the resolution scores, an entity candidate with respect to one of the one or more entity types to be an individual associated with the entity name, wherein the entity resolution engine is trained to perform the selecting based on the generated person-centric spaces associated with the user and the other users related to the user; and

generating a representation associated with the user based on the selected candidate, wherein the representation facilitates providing a response to a query issued by the user.

10. The medium of claim 9 , wherein the information being cross-linked is obtained from at least one of the following: the user's emails, contacts, instant messages, browsing history, call records, and bookmarks.

11. The medium of claim 9 , wherein the entity name is extracted from a card to be presented to the user.

12. The medium of claim 9 , wherein the entity candidate is selected based on textual metadata associated with the entity name in the person-centric space for the user.

13. The method of claim 1 , wherein determining one or more entity types comprises:

determining a plurality of entity types associated with the entity name; and

selecting the one or more entity types from the plurality of entity types, based on the entity name and an entity type resolution model trained with some entities with known entity types.

14. The method of claim 13 , wherein the plurality of entity types includes people, place, business, country, and title.

15. The system of claim 5 , wherein determining one or more entity types comprises:

determining a plurality of entity types associated with the entity name; and

selecting the one or more entity types from the plurality of entity types, based on the entity name and an entity type resolution model trained with some entities with known entity types.

16. The system of claim 15 , wherein the plurality of entity types includes people, place, business, country, and title.

17. The medium of claim 9 , wherein determining one or more entity types comprises:

determining a plurality of entity types associated with the entity name; and

selecting the one or more entity types from the plurality of entity types, based on the entity name and an entity type resolution model trained with some entities with known entity types, wherein the plurality of entity types includes people, place, business, country, and title.

18. The method of claim 1 , wherein the selecting is further based on contextual information associated with the entity name, and wherein the selecting further includes determining a distance of the entity name from a phrase included in the contextual information.

19. The method of claim 1 , wherein the resolution engine is further trained to perform the selecting based on a location of a source of the information obtained from the public, semi-private and private spaces and a time of obtaining the information from the public, semi-private and private spaces.

20. The method of claim 1 , wherein the resolution engine is further trained to perform the selecting based on a type of format of the information stored in the person-centric space.

21. The method of claim 1 , wherein the entity resolution engine is selected based on the determined one or more entity types associated with the entity name.

22. The method of claim 1 , wherein the person-centric space is generated without user input.