System and method for a spoken language interface to a large database of changing records

• US 20030149566A1
• Filed: 12/31/2002
• Published: 08/07/2003
• Est. Priority Date: 01/02/2002
• Status: Abandoned Application

• Current Assignee: Telelogue (Microsoft Corporation)
• Original Assignee: Telelogue (Microsoft Corporation)
• Inventors: Burke, Robert, Levin, Esther, Zeljkovic, Ilija, Lyudovyk, Yevgeniy, Boyce, Susan, Helfrich, Brian
• Application Number:
  US10/331,343
• Publication Number:
  US 20030149566A1
• Time in Patent Office: Days
• US Class Current: 704/256
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models

Abstract

Embodiments of the present invention provide a spoken language interface to an information database. A grammars database based on the entries contained in the information database may be generated. The entries in the grammars database may be a compact representation of the entries in the information database. An index database based on the entries contained in the information database may be generated. The grammars database and the index database may be updated periodically based on updated entries contained in the information database. A recognized result of a user'"'"'s communication based on the updated grammars database may be generated. The updated index database may be searched for a list of matching entries that match the recognized result. The list of matching entries may be output.

Claims

1. A method for providing a spoken language interface to an information database, comprising:

generating a grammars database based on the entries contained in the information database, wherein entries in the grammars database are a compact representation of the entries in the information database;

generating an index database based on the entries contained in the information database;

periodically updating the grammars database based on updated entries contained in the information database;

periodically updating the index database based on the updated entries contained in the information database;

generating a recognized result of a user'"'"'s communication based on the updated grammars database;

searching the updated index database for a list of matching entries that match the recognized result; and

outputting the list of matching entries.

2. The method of claim 1, wherein generating the grammars database comprises:

generating entries in the grammars database based on the entries in the information database using estimated N-gram statistics.

3. The method of claim 2, wherein entries in the grammars database include bi-gram grammars.

4. The method of claim 1, wherein periodically updating the grammars database comprises:

generating entries in the grammars database based on the entries in the information database using estimated N-gram statistics.

5. The method of claim 1, wherein the entries in the grammars database do not directly correspond to entries in the listings database.

6. The method of claim 1, wherein the information database is a listings database.

7. The method of claim 1, wherein the grammars database is updated daily, weekly or monthly.

8. The method of claim 1, wherein the index database is updated daily, weekly or monthly.

9. The method of claim 1, wherein periodically updating the grammars database comprises:

processing a plurality of entries of the information database through a distortion model.

10. The method of claim 9, further comprising:

generating a transformation rule set for each entry from the plurality of entries; and

transforming each entry into a variation of the entry based on the rule set.

11. The method of claim 10, furthering comprising:

generating a probability associated with the variation of the entry.

12. The method of claim 11, further comprising:

creating a pseudo corpus including the variation and the associated probability.

13. The method of claim 11, further comprising:

generating a language model based on the variation and the associated probability using parameter estimation.

14. The method of claim 1, further comprising:

searching the listing database for the list of matching entries that matched the recognized result based on the updated index database.

15. The method of claim 1, wherein periodically updating the index database comprises:

processing a plurality of entries of the information database through a distortion model.

16. The method of claim 15, further comprising:

generating a transformation rule set for each entry from the plurality of entries; and

transforming each entry into a variation of the entry based on the rule set.

17. The method of claim 16, furthering comprising:

generating a probability associated with the variation of the entry.

18. A method comprising:

retrieving each entry of a plurality of entries contained in an informational database;

applying a transformation rule to each entry of the plurality of entries in the informational database;

generating a variation of each entry based on the applied transformation rule;

generating an associated probability for each variation; and

generating a stochastic language model for each variation and the associated probability based a parameter estimation technique.

19. The method of claim 18, wherein the informational database is a listings database.

20. The method of claim 18, wherein the variation of an entry is a distortion of the entry in the informational database.

21. The method of claim 18, further comprising:

outputting the generated stochastic language model into a grammar database.

22. The method of claim 18, further comprising:

outputting the generated stochastic language model into an index database.

23. The method of claim 18, wherein the transformation rule specify an alternate way of uttering an entry.

24. The method of claim 23, further comprising:

generating a pseudo corpus based on each variation and the associated probability.

25. The method of claim 18, wherein the variation includes an alternate word sequence representing an entry in the informational database.

26. An apparatus for providing a spoken language interface to an information database, comprising:

a grammar generator that is to periodically update a grammars database based on updated entries contained in an information database, wherein entries in the grammars database are a compact representation of the entries in the information database;

a index generator that is to periodically update an index database based on the updated entries contained in the information database;

a recognizer that is to generating a recognized result of a user'"'"'s communication based on the updated grammars database;

a matcher that is to search the updated index database for a list of matching entries that match the recognized result; and

an output manager to output the list of matching entries.

27. The apparatus of claim 26, wherein the grammar database is to periodically update entries in the grammars database based on the entries in the information database based on estimated N-gram statistics.

28. The apparatus of claim 26, wherein the entries in the grammars database do not directly correspond to entries in the listings database.

29. The apparatus of claim 26, wherein the grammar generator comprises:

a distortion model that is to generate a variation of an entry in the information database.

30. The apparatus of claim 29, wherein the distortion model comprises:

an analyzer that is to generate a transformation rule; and

an orthographies generator that is to generate the variation of the entry in the information database based on the generated transformation rule.

31. The apparatus of claim 30, wherein the distortion model to generate a probability associated with the variation.

32. The apparatus of claim 31, further comprising:

a pseudo corpus that is to store the variation and the associated probability.

33. The apparatus of claim 31, further comprising:

a parameter estimator that is to generate a language model based on the variation and the associated probability.

34. A machine-readable medium having stored thereon a plurality of executable instructions, the plurality of instructions comprising instructions to:

generate a grammars database based on the entries contained in the information database, wherein entries in the grammars database are a compact representation of the entries in the information database;

generate an index database based on the entries contained in the information database;

periodically update the grammars database based on updated entries contained in the information database;

periodically update the index database based on the updated entries contained in the information database;

generate a recognized result of a user'"'"'s communication based on the updated grammars database;

search the updated index database for a list of matching entries that match the recognized result; and

output the list of matching entries.

35. The machine-readable medium of claim 34 having stored thereon additional executable instructions, the additional instructions comprising instructions to:

generate entries in the grammars database based on the entries in the information database using estimated N-gram statistics.

36. The machine-readable medium of claim 34 having stored thereon additional executable instructions, the additional instructions comprising instructions to:

generate entries in the grammars database based on the entries in the information database using estimated N-gram statistics.

37. The machine-readable medium of claim 34 having stored thereon additional executable instructions, the additional instructions comprising instructions to:

process a plurality of entries of the information database through a distortion model.

38. The machine-readable medium of claim 37 having stored thereon additional executable instructions, the additional instructions comprising instructions to:

generate a transformation rule set for each entry from the plurality of entries; and

transform each entry into a variation of the entry based on the rule set.

39. The machine-readable medium of claim 38 having stored thereon additional executable instructions, the additional instructions comprising instructions to:

generate a probability associated with the variation of the entry.

40. The machine-readable medium of claim 39 having stored thereon additional executable instructions, the additional instructions comprising instructions to:

create a pseudo corpus including the variation and the associated probability.

41. The machine-readable medium of claim 39 having stored thereon additional executable instructions, the additional instructions comprising instructions to:

generate a language model based on the variation and the associated probability using parameter estimation.

42. The machine-readable medium of claim 34 having stored thereon additional executable instructions, the additional instructions comprising instructions to:

search the listing database for the list of matching entries that matched the recognized result based on the updated index database.

43. The machine-readable medium of claim 34 having stored thereon additional executable instructions, the additional instructions comprising instructions to:

process a plurality of entries of the information database through a distortion model.

44. The machine-readable medium of claim 43 having stored thereon additional executable instructions, the additional instructions comprising instructions to:

generate a transformation rule set for each entry from the plurality of entries; and

transform each entry into a variation of the entry based on the rule set.