ENTITY RECOGNITION USING PROBABILITIES FOR OUT-OF-COLLECTION DATA

• US 20120323839A1
• Filed: 06/16/2011
• Published: 12/20/2012
• Est. Priority Date: 06/16/2011
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Kiciman, Emre, Wang, Kuansan, Aji, Abulimiti
• Granted Patent: US 9,104,979 B2
• Time in Patent Office: Days
• US Class Current: 706/52
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

A classifier that disambiguates among entities based on a dictionary, such as corpus of documents about those entities, is built by incorporating probabilities that an entity exists that is not in the dictionary. Given a document it is associated by the classifier with an entity. By incorporating out of collection probabilities into the classifier, a higher level of confidence in the match between an entity and a document is achieved.

Claims

1. A computer-implemented process for building a classifier for associating an entity with a document, comprising:

accessing a dictionary that maps entities to related terms;

partitioning the dictionary by entity names to provide a set of partitions, each partition relating to an entity name;

estimating a probability that an entity having the entity name for a partition is not represented in the dictionary; and

creating a classifier for the partition including the estimated probability.

2. The computer implemented process of claim 1, wherein the classifier is interpretable.

3. The computer implemented process of claim 2, wherein the classifier is a decision tree.

4. The computer implemented process of claim 3, wherein the decision tree includes at least one node having a probability incorporating the estimated probability that an entity having the entity name for a partition is not represented in the dictionary.

5. The computer implemented process of claim 4, wherein the probability incorporating the estimated probability is computed by |G|*P(‘

name’

)*P(f_1)* . . . *P(f_k), where |G| is the estimated size of the larger population, both including and in addition to the dictionary, P(‘

name’

) is the probability of an entity having the same name, and P(f_i) is the probability of an entity passing the feature test f_i.

6. The computer implemented process of claim 5, wherein P(f_i) is the dependent probability of an entity passing the feature test f_i given one or more features on which the feature test f_i depends.

7. The computer implemented process of claim 2, wherein the probability incorporating the estimated probability is computed by |G|*P(‘

name’

)*P(f_1)* . . . *P(f_k), where |G| is the estimated size of the larger population, both including and in addition to the dictionary, P(‘

name’

) is the probability of an entity having the same name, and P(f_i) is the probability of an entity passing the feature test f_i.

8. The computer implemented process of claim 7, wherein P(f_i) is the dependent probability of an entity passing the feature test f_i given one or more features on which the feature test f_i depends.

9. A computing machine comprising:

a partitioner having an input for receiving a dictionary of documents related to entities and an output providing a set of partitions, each partition relating to an entity name;

a statistics module having an input for receiving the dictionary and an output providing statistics regarding terms used in the dictionary;

a classification builder having an input for receiving data about the partitions and an input for receiving the statistics, and providing data describing a classifier as an output, wherein the classifier incorporates an estimated probability that an entity having the entity name for a partition is not represented in the dictionary.

10. The computing machine of claim 9, wherein the classifier is interpretable.

11. The computing machine of claim 10, wherein the classifier is a decision tree.

12. The computing machine of claim 11, wherein the decision tree includes at least one node having a probability incorporating the estimated probability that an entity having the entity name for a partition is not represented in the dictionary.

13. The computing machine of claim 12, wherein the probability incorporating the estimated probability is computed by |G|*P(‘

name’

)*P(f_1)* . . . *P(f_k), where |g| is the estimated size of the larger population, both including and in addition to the dictionary, P(‘

name’

) is the probability of an entity having the same name, and P(f_i) is the probability of an entity passing the feature test f_i.

14. The computing machine of claim 13, wherein P(f_i) is the dependent probability of an entity passing the feature test f_i given one or more features on which the feature test f_i depends.

15. The computing machine of claim 10, wherein the probability incorporating the estimated probability is computed by |G|*P(‘

name’

)*P(f_1)* . . . *P(f_k), where |G| is the estimated size of the larger population, both including and in addition to the dictionary, P(‘

name’

) is the probability of an entity having the same name, and P(f_i) is the probability of an entity passing the feature test f_i.

16. The computing machine of claim 15, wherein P(f_i) is the dependent probability of an entity passing the feature test f_i given one or more features on which the feature test f_i depends.

17. A computer-implemented process for associating an entity with a document, comprising:

accessing a classifier associated with an entity name in the document, wherein the classifier incorporates an estimated probability that an entity having the entity name for a partition is not represented in a dictionary; and

applying the classifier to the document to obtain probabilities that the document is associated with specific entities having the entity name.

18. The computer implemented process of claim 17, further comprising selecting an entity from among the specific entities according to the probabilities.

19. The computer implemented process of claim 18, wherein the classifier is a decision tree including at least one node having a probability incorporating the estimated probability that an entity having the entity name for a partition is not represented in the dictionary.

20. The computer implemented process of claim 17, further comprising identifying the document in response to a user query on a database of documents separate from the dictionary.