Systems and methods for determining predictive models of discourse functions

• US 20050182625A1
• Filed: 02/18/2004
• Published: 08/18/2005
• Est. Priority Date: 02/18/2004
• Status: Active Grant

• Current Assignee: Fuji Xerox, Fuji Xerox Company Limited (Xerox Holdings Corp.)
• Original Assignee: Fuji Xerox Company Limited (Xerox Holdings Corp.)
• Inventors: Polanyi, Livia, Van Den Berg, Martin H., Thione, Giovanni L., Azara, Misty
• Granted Patent: US 7,542,903 B2
• Time in Patent Office: Days
• US Class Current: 704/236
• CPC Class Codes: G06F 40/35 Discourse or dialogue representation
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 25/51 for comparison or discrimination
• CPC Class Codes: G10L 25/90 Pitch determination of speech signals

Abstract

Techniques are provided for determining predictive models of discourse functions based on prosodic features of natural language speech. Inter and intra sentential discourse functions in a training corpus of natural language speech utterances are determined. The discourse functions are clustered. The exemplary prosodic features associated with each type of discourse function are determined. Machine learning, observation and the like are used to determine a subset of prosodic features associated with each type of discourse function useful in predicting the likelihood of each type of discourse function.

Claims

1. A method of determining a predictive model for discourse functions comprising the steps of:

determining a corpus of speech utterances;

determining a least one discourse function associated with at least one speech utterance;

determining at least one prosodic feature associated with the at least one discourse function; and

determining at least one predictive model of discourse functions based on the prosodic features and the discourse functions.

2. The method of claim 1, in which the discourse functions are determined based on a theory of discourse analysis.

3. The method of claim 2, in which the theory of discourse analysis is at least one of:

the Linguistic Discourse Model, the Unified Linguistic Discourse Model, Rhetorical Structure Theory, Discourse Structure Theory and Structured Discourse Representation Theory.

4. The method of claim 1, in which the predictive models are determined based on at least one of:

machine learning, rules.

5. The method of claim 4, in which the machine learning based predictive models are determined based on at least one of:

statistics, decision trees, Naï

ve Bayes.

6. The method of claim 1, in which the prosodic features occur in at least one of a location:

preceding, within and following the associated discourse function.

7. The method of claim 1, in which the prosodic features are encoded within a prosodic feature vector.

8. The method of claim 7, in which the prosodic feature vector is a multimodal feature vector.

9. The method of claim 1, in which the discourse function is an intra-sentential discourse function.

10. The method of claim 1, in which the discourse function is an inter-sentential discourse function.

11. A system for determining predictive discourse function models comprising:

an input/output circuit for retrieving a corpus of at least one speech utterance;

a processor for determining prosodic features associated with the at least one speech utterance, and which determines at least one discourse function associated with the corpus of at least one speech utterance and determines at least one prosodic feature associated with the at least one discourse function and determines a predictive model for discourse functions based on the prosodic features and the discourse function.

12. The system of claim 11, in which the discourse functions are determined based on a theory of discourse analysis.

13. The system of claim 12, in which the theory of discourse analysis is at least one of:

the Linguistic Discourse Model, the Unified Linguistic Discourse Model, Rhetorical Structure Theory, Discourse Structure Theory and Structured Discourse Representation Theory.

14. The system of claim 11, in which the predictive models are determined based on at least one of:

machine learning, rules.

15. The system of claim 14, in which the machine learning based predictive models are determined based on at least one of:

statistics, decision trees, Naï

ve Bayes.

16. The system of claim 11, in which the prosodic features occur in at least one of a location:

preceding, within and following the associated discourse function.

17. The system of claim 11, in which the prosodic features are encoded within a prosodic feature vector.

18. The system of claim 17, in which the prosodic feature vector is a multimodal feature vector.

19. The system of claim 11, in which the discourse function is an intra-sentential discourse function.

20. The system of claim 11, in which the discourse function is an inter-sentential discourse function.

21. A carrier wave encoded to transmit a control program, useable to program a computer to determine a predictive model for discourse functions, to a device for executing the program, the control program comprising:

instructions for determining a corpus of speech utterances;

instructions for determining a least one discourse function associated with at least one speech utterance;

instructions for determining at least one prosodic feature associated with the at least one discourse function;

instructions for determining at least one predictive model of discourse functions based on the prosodic features and the discourse functions.

22. Computer readable storage medium comprising:

computer readable program code embodied on the computer readable storage medium, the computer readable program code usable to program a computer to determine a predictive model for discourse functions comprising the steps of;

determining a corpus of speech utterances;

determining a least one discourse function associated with at least one speech utterance;

determining at least one prosodic feature associated with the at least one discourse function; and

determining at least one predictive model of discourse functions based on the prosodic features and the discourse functions.