SYSTEMS AND METHODS FOR LEARNING SEMANTIC PATTERNS FROM TEXTUAL DATA

• US 20160364476A1
• Filed: 06/11/2015
• Published: 12/15/2016
• Est. Priority Date: 06/11/2015
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Curin, Jan
• Granted Patent: US 9,959,341 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 40/205 Parsing
• CPC Class Codes: G06F 40/268 Morphological analysis
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

In some embodiments, a system is provided comprising at least one processor programmed to process an input text to identify a plurality of semantic patterns that match the input text, wherein, for at least one semantic pattern of the plurality of semantic patterns: the at least one semantic pattern comprises a plurality of semantic entities identified from the at least one input text, and the plurality of semantic entities occur in a common context within the at least one input text. The at least one processor may be further programmed to use statistical information derived from training data to associate a respective weight with each semantic pattern of the plurality of semantic patterns.

Claims

1. A system comprising at least one processor programmed to:

process an input text to identify a plurality of semantic patterns that match the input text, wherein, for at least one semantic pattern of the plurality of semantic patterns;

the at least one semantic pattern comprises a plurality of semantic entities identified from the at least one input text, and

the plurality of semantic entities occur in a common context within the at least one input text; and

use statistical information derived from training data to associate a respective weight with each semantic pattern of the plurality of semantic patterns.

2. The system of claim 1, wherein:

the statistical information comprises at least one measure of mutual information for the plurality of semantic entities of the at least one semantic pattern; and

the at least one measure of mutual information is derived from the training data.

3. The system of claim 2, wherein the at least one measure of mutual information comprises a measure of pointwise mutual information.

4. The system of claim 2, wherein the at least one measure of mutual information comprises a measure of weighted mutual information.

5. The system of claim 1, wherein:

the at least one semantic pattern comprises a valency frame;

the plurality of semantic entities comprise a plurality of valency frame components of the valency frame; and

the common context in which the plurality of valency frame components occur comprises a valency structure controlled by a controlling valency frame component.

6. The system of claim 1, wherein:

the at least one semantic pattern is in a lexicon of semantic patterns; and

the at least one processor is programmed to process the input text to identify, from the lexicon of semantic patterns, one or more semantic patterns that match the input text.

7. The system of claim 1, wherein the at least one processor is programmed to:

select one or more semantic patterns from the plurality of semantic patterns based on the respective weights; and

output, as semantic patterns matching the input text, the one or more semantic patterns and the respective weights.

8. A method comprising acts of:

processing an input text to identify a plurality of semantic patterns that match the input text, wherein, for at least one semantic pattern of the plurality of semantic patterns;

the at least one semantic pattern comprises a plurality of semantic entities identified from the at least one input text, and

the plurality of semantic entities occur in a common context within the at least one input text; and

using statistical information derived from training data to associate a respective weight with each semantic pattern of the plurality of semantic patterns.

9. The method of claim 8, wherein the statistical information comprises at least one measure of mutual information for the plurality of semantic entities of the at least one semantic pattern, and wherein the method further comprises an act of:

deriving the at least one measure of mutual information from the training data.

10. The method of claim 9, wherein the at least one measure of mutual information comprises a measure of pointwise mutual information.

11. The method of claim 9, wherein the at least one measure of mutual information comprises a measure of weighted mutual information.

12. The method of claim 8, wherein:

the at least one semantic pattern comprises a valency frame;

the plurality of semantic entities comprise a plurality of valency frame components of the valency frame; and

the common context in which the plurality of valency frame components occur comprises a valency structure controlled by a controlling valency frame component.

13. The method of claim 8, wherein the at least one semantic pattern is in a lexicon of semantic patterns, and wherein the method further comprises an act of:

processing the input text to identify, from the lexicon of semantic patterns, one or more semantic patterns that match the input text.

14. The method of claim 8, further comprising acts of:

selecting one or more semantic patterns from the plurality of semantic patterns based on the respective weights; and

outputting, as semantic patterns matching the input text, the one or more semantic patterns and the respective weights.

15. At least one computer-readable medium having encoded thereon instructions which, when executed by at least one processor, cause the at least one processor to perform a method comprising acts of:

processing an input text to identify a plurality of semantic patterns that match the input text, wherein, for at least one semantic pattern of the plurality of semantic patterns;

the at least one semantic pattern comprises a plurality of semantic entities identified from the at least one input text, and

the plurality of semantic entities occur in a common context within the at least one input text; and

using statistical information derived from training data to associate a respective weight with each semantic pattern of the plurality of semantic patterns.

16. The at least one computer-readable medium of claim 15, wherein the statistical information comprises at least one measure of mutual information for the plurality of semantic entities of the at least one semantic pattern, and wherein the method further comprises an act of:

deriving the at least one measure of mutual information from the training data.

17. The at least one computer-readable medium of claim 16, wherein the at least one measure of mutual information comprises a measure of pointwise mutual information or a measure of weighted mutual information.

18. The at least one computer-readable medium of claim 15, wherein:

the at least one semantic pattern comprises a valency frame;

the plurality of semantic entities comprise a plurality of valency frame components of the valency frame; and

the common context in which the plurality of valency frame components occur comprises a valency structure controlled by a controlling valency frame component.

19. The at least one computer-readable medium of claim 15, wherein the at least one semantic pattern is in a lexicon of semantic patterns, and wherein the method further comprises an act of:

processing the input text to identify, from the lexicon of semantic patterns, one or more semantic patterns that match the input text.

20. The at least one computer-readable medium of claim 15, wherein the method further comprises acts of:

selecting one or more semantic patterns from the plurality of semantic patterns based on the respective weights; and

outputting, as semantic patterns matching the input text, the one or more semantic patterns and the respective weights.