Efficiently Representing Word Sense Probabilities

• US 20090094019A1
• Filed: 08/29/2008
• Published: 04/09/2009
• Est. Priority Date: 08/31/2007
• Status: Active Grant

• Current Assignee: Zhigu Holdings Limited
• Original Assignee: Powerset (Microsoft Corporation)
• Inventors: Converse, Timothy, Thione, Giovanni Lorenzo, Walters, Chad, Snow, Rion, Waterman, Scott A.
• Granted Patent: US 8,280,721 B2
• Time in Patent Office: Days
• US Class Current: 704/9
• CPC Class Codes: G06F 40/268 Morphological analysis

Abstract

Word sense probabilities are compressed for storage in a semantic index. Each word sense for a word is mapped to one of a number of “buckets” by assigning a bucket score to the word sense. A scoring function is utilized to assign the bucket scores that maximizes the entropy of the assigned bucket scores. Once the bucket scores have been assigned to the word senses, the bucket scores are stored in the semantic index. The bucket scores stored in the semantic index may be utilized to prune one or more of the word senses prior to construction of the semantic index. The bucket scores may also be utilized to prune and rank the word senses at the time a query is performed using the semantic index.

Claims

1. A method for efficiently representing word sense probabilities, the method comprising:

identifying one or more word senses associated with a word;

obtaining a word sense probability associated with each of the word senses;

mapping each word sense to a bucket by assigning a bucket score to each word sense, the bucket score based upon the word sense probability; and

storing the bucket score for each word sense in a semantic index.

2. The method of claim 1, wherein mapping each word sense to a bucket comprises generating a monotonic mapping between the word sense probabilities and the bucket scores.

3. The method of claim 1, further comprising utilizing the bucket scores stored in the semantic index to prune one or more of the word senses prior to construction of the semantic index.

4. The method of claim 1, further comprising utilizing the bucket scores stored in the semantic index to rank the word senses when a query is performed.

5. The method of claim 1, further comprising utilizing the bucket scores stored in the semantic index to prune one or more of the word senses when a query is performed.

6. The method of claim 1, wherein approximately equal percentages of the word sense probabilities are associated with each bucket.

7. The method of claim 1, wherein the bucket score comprises an N bit binary number.

8. A computer storage medium having computer executable instructions stored thereon which, when executed by a computer, cause the computer to:

store a semantic index;

identify one or more word senses associated with a word;

obtain a word sense probability for each of the word senses;

map each word sense to a bucket score; and

to

store the bucket score for each word sense probability in the semantic index.

9. The computer storage medium of claim 8, wherein each word sense probability identifies a probability that an associated word refers to a corresponding word sense.

10. The computer storage medium of claim 9, wherein each bucket score comprises an N bit binary number.

11. The computer storage medium of claim 10, wherein each word sense is mapped to a bucket score by assigning a bucket score to the word sense based upon the word sense probability.

12. The computer storage medium of claim 11, wherein a monotonic mapping exists between the word sense probabilities and the bucket scores.

13. The computer storage medium of claim 12, wherein approximately equal percentages of the word sense probabilities are associated with each bucket score.

14. The computer storage medium of claim 13, comprising further computer executable instructions which, when executed by the computer, will cause the computer to utilize the bucket scores stored in the semantic index to prune one or more of the word senses prior to construction of the semantic index.

15. The computer storage medium of claim 14, comprising further computer executable instructions which, when executed by the computer, will cause the computer to utilize the bucket scores stored in the semantic index to rank the word senses when a query is performed.

16. The computer storage medium of claim 15, comprising further computer executable instructions which, when executed by the computer, will cause the computer to utilize the bucket scores stored in the semantic index to prune one or more of the word senses at the time a query is performed.

17. A computing system for efficiently representing word sense probabilities, the computing system comprising:

a central processing unit;

a memory; and

a mass storage device coupled to the central processing unit storing a semantic index and program code that is executable by the central processing unit and which, when executed by the central processing unit, will cause the computing system to identify one or more word senses associated with a word, to obtain a word sense probability for each of the word senses, to create a monotonic mapping between each word sense and a bucket by assigning a bucket score to each word sense, and to store the bucket score for each word sense probability in the semantic index, and wherein approximately equal percentages of the word sense probabilities are associated with each bucket score.

18. The computing system of claim 17, comprising further program code stored on the mass storage device which, when executed by the central processing unit, will cause the computing system to utilize the bucket scores stored in the semantic index to prune one or more of the word senses prior to construction of the semantic index.

19. The computing system of claim 18, comprising further program code stored on the mass storage device which, when executed by the central processing unit, will cause the computing system to utilize the bucket scores stored in the semantic index to rank the word senses when a query is performed.

20. The computing system of claim 19, comprising further program code stored on the mass storage device which, when executed by the central processing unit, will cause the computing system to utilize the bucket scores stored in the semantic index to prune one or more of the word senses at the time a query is performed.