System and method of providing a spoken dialog interface to a website

• US 7,373,300 B1
• Filed: 12/18/2003
• Issued: 05/13/2008
• Est. Priority Date: 12/18/2002
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: AT&T Corporation (AT&T, Inc.)
• Inventors: Bangalore, Srinivas, Feng, Junlan, Rahim, Mazin G.
• Primary Examiner(s): Smits, Talivaldis Ivars
• Application Number:
  US10/740,186
• Time in Patent Office: 1,608 Days
• Field of Search: None
• US Class Current: 704/270.1
• CPC Class Codes: G10L 13/08 Text analysis or generation of parameters for speech synthesis out of text, e.g. grapheme to phoneme translation, prosody generation or stress or intonation determination
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: H04M 3/4936 Speech interaction details speech recognition per se G10L15/00

Abstract

Disclosed is a system and method for generating a spoken dialog service from website data. Spoken dialog components typically include an automatic speech recognition module, a language understanding module, a dialog management module, a language generation module and a test-to-speech module. These components are capable of being automatically trained from processed website data. A website analyzer converts a website into structured text data set and a structured task knowledge base. The website analyzer further extracts linguistic items from the website data. The dialog components are automatically trained from the structured text data set, structured task knowledge base and linguistic items.

Claims

1. A method of generating data for use by a plurality of components in a spoken dialog service, the method comprising:

converting data from a structured database to a structured text data set and a structured task knowledge base for use by at least a portion of a plurality of spoken dialog service components;

extracting linguistic items from the structured database for use by at least another portion of the plurality of spoken dialog service components; and

automatically generating a spoken dialog service by training each of the plurality of spoken dialog service components using the structured text data, the structured task knowledge base or the linguistic items.

2. The method of claim 1, wherein the structured text data set and the structured task knowledge base are used by a language understanding module, a language generation module and a dialog manager module.

3. The method of claim 1, where the linguistic items from the data are used for building a language model used for training an automatic speech recognition module.

4. The method of claim 3, wherein the linguistic items comprise named-entities.

5. The method of claim 4, wherein the structured text data further comprises nominal, verbal and adjectival phrases.

6. The method of claim 1, wherein the structured text data comprises a hierarchical tree.

7. The method of claim 6, wherein the hierarchical tree comprises a plurality of non-leaf tree nodes having a node alias that is human understandable.

8. The method of claim 7, wherein each non-leaf tree node further comprises a verbose description document and a concise summary.

9. The method of claim 8, wherein the verbose description document and concise summary are used for information matching and help prompt construction during dialog execution.

10. The method of claim 6, wherein the hierarchical tree further comprises at least one leaf node corresponding to a webpage.

11. The method of claim 10, wherein each leaf node further contains a list of information units associated with the webpage.

12. The method of claim 11, wherein the list of information units further comprises product and service names, product descriptions and service definitions.

13. The method of claim 12, wherein the list of information units further comprises explanations of acronyms and frequently asked questions.

14. An analyzer module for generating a database of information used to build a spoken dialog service, the analyzer module comprising:

means for generating a structured text data set and a structured task knowledge base from website data for use by at least one spoken dialog component;

means for extracting linguistic items from the website data for use by at least one spoken dialog component; and

means for automatically training each component of a spoken dialog service using the structured text data set, structured task knowledge base and linguistic items.

15. The analyzer module of claim 14, wherein the structured text data set and the structured task knowledge base are used by a language understanding module, language generation module and dialog manager module.

16. The analyzer module of claim 14, where the linguistic items from web data are used for building a language model used for training an automatic speech recognition module.

17. The analyzer module of claim 16, wherein the linguistic items comprise named entities.

18. The analyzer module of claim 15, wherein the structured text data further comprises nominal, verbal and adjectival phrases.

19. The analyzer module of claim 14, wherein the structured text data comprises a hierarchical tree.

20. The analyzer module of claim 19, wherein the hierarchical tree comprises a plurality of non-leaf tree nodes having a node alias that is human understandable.

21. The analyzer module of claim 20, wherein each non-leaf tree node further comprises a verbose description document and a concise summary.

22. The analyzer module of claim 21, wherein the hierarchical tree further comprises leaf tree node that each correspond to a webpage.

23. The analyzer module of claim 22, wherein each leaf tree node further includes a list of information units.

24. The analyzer module of claim 23, wherein the list of information units comprises product names and product or service descriptions.

25. The analyzer module of claim 24, wherein the list of information units further comprises service names and definitions.

26. The analyzer module of claim 25, wherein the list of information units further comprises explanations of acronyms and frequently asked questions.

27. A system for generating a spoken dialog service from website data, the system comprising:

A. a plurality of spoken dialog components comprising an automatic speech recognition module, a language understanding module, a dialog management module, a language generation module and a text-to-speech module; and

B. a website analyzer having;

1. means for generating a structured text data set and a structured task knowledge base from website data for use by at least one of the spoken dialog components;

2. means for extracting linguistic items from the website data for use by at least one of the spoken dialog components; and

3. means for automatically training each component of a spoken dialog service using the structured text data set, structured task knowledge base and linguistic items.

28. The system for generating a spoken dialog service from website data of claim 27. wherein the structured text data set and the structured task knowledge base are used by a language understand module, language generation module and dialog manager module.

29. The system for generating a spoken dialog service from website data of claim 27, where the linguistic items for web data are used for building a language model used for training an automatic speech recognition module.

30. The system for generating a spoken dialog service from website data of claim 29, wherein the linguistic items comprise named-entities.

31. The system for generating a spoken dialog service from website data of claim 28, wherein the structured text data further comprises nominal, verbal and adjectival phrases.

32. The system for generating a spoken dialog service from website data of claim 27, wherein the structured text data comprises a hierarchical tree.

33. The system of generating a spoken dialog service from website data of claim 32, wherein the hierarchical tree comprises a plurality of non-leaf tree nodes having a node alias that is human understandable.

34. The system for generating a spoken dialog service from website data of claim 33, wherein each non-leaf tree node further comprises a verbose description document and a concise summary.

35. The system for generating a spoken dialog service from website data of claim 34, wherein the hierarchical tree further comprises a leaf tree node that corresponds to a webpage.

36. The system for generating a spoken dialog service from website data of claim 35, wherein each leaf tree node further includes a list of information units.

37. The system of generating a spoken dialog service from website data of claim 36, wherein the list of information units comprises product names and product or service descriptions.

38. The system for generating a spoken dialog service from website data of claim 37, wherein the list of information units further comprises service names and definitions.

39. The system for generating a spoken dialog service from website data of claim 38, wherein the list of information units further comprises explanations of acronyms and frequently asked questions.

40. A method of segmenting webpages from a website for use in automatically generating a spoken dialog interface to the website, the method comprising:

representing each webpage in a website by at least one set of vectors;

encoding each webpage as a sequence of features vectors;

detecting boundaries between information units on each webpage; and

classifying information blocks organized according to the detected boundaries between information units into information class symbols, wherein the classified information blocks are utilized for generating a spoken dialog interface to the website.

41. The method of claim 40, wherein detecting boundaries is performed using a binary information retrieval classifier.

42. The method of claim 40, wherein detecting boundaries is performed using a binary support vector machine classifier to classify difference vectors.