Library of existing spoken dialog data for use in generating new natural language spoken dialog systems

• US 8,478,589 B2
• Filed: 01/05/2005
• Issued: 07/02/2013
• Est. Priority Date: 01/05/2005
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: AT&T Intellectual Property II LP (AT&T, Inc.)
• Inventors: Shahraray, Behzad, Tur, Gokhan, Begeja, Lee, Di Fabbrizio, Giuseppe, Gibbon, David Crawford, Hakkani-Tur, Dilek Z., Liu, Zhu, Renger, Bernard S.
• Primary Examiner(s): Godbold, Douglas
• Assistant Examiner(s): Ortiz Sanchez, Michael
• Application Number:
  US11/029,319
• Publication Number:
  US 20060149554A1
• Time in Patent Office: 3,100 Days
• Field of Search: 704/257, 704/251, 704243-245, 704/231, 704/256, 704/232
• US Class Current: 704/231
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 15/28 Constructional details of speech recognition systems
• CPC Class Codes: G10L 25/48 specially adapted for particular use

Abstract

A machine-readable medium may include a group of reusable components for building a spoken dialog system. The reusable components may include a group of previously collected audible utterances. A machine-implemented method to build a library of reusable components for use in building a natural language spoken dialog system may include storing a dataset in a database. The dataset may include a group of reusable components for building a spoken dialog system. The reusable components may further include a group of previously collected audible utterances. A second method may include storing at least one set of data. Each one of the at least one set of data may include ones of the reusable components associated with audible data collected during a different collection phase.

Claims

1. A non-transitory computer-readable medium comprising:

a plurality of reusable components for building a natural language spoken dialog system, each of the plurality of reusable components comprising a plurality of groups of previously collected audible utterances and associated labels for call-types and named entities, wherein;

(1) the plurality of reusable components is organized into a plurality of datasets;

(2) each of the plurality of datasets comprises data pertaining to an industrial sector in a different task domain;

(3) data in the plurality of datasets is collected during a plurality of collection phases, each of the plurality of collection phases comprising a respective defined period of time;

(4) each group of the plurality of groups of previously collected audible utterances was collected in a separate spoken dialog system operating within a respective industry sector; and

(5) an annotation guide comprising guideline utterances and descriptions, the guideline utterances comprising both positive and negative utterances for an associated call-type category,

wherein the previously collected audible utterances are associated with an occurrence of utterance data comprising information indicating the associated call-type category, and wherein each respective industry sector is in a different task domain from other respective industry sectors.

2. The non-transitory computer-readable medium of claim 1, further comprising:

a first set of instructions and data for implementing a natural language understanding model based on at least one of the plurality of reusable components,

a second set of instructions and data for implementing an automatic speech recognition module based on at least one of the plurality of reusable components, and

a third set of instructions and data for implementing at least one of a named entity detection/extraction grammar and a model based on the at least one of the plurality of reusable components.

3. The non-transitory computer-readable medium of claim 1, wherein the occurrence of utterance data comprises at least one of information indicating to which of the previously collected audible utterances a label applies and information indicating to which of the previously collected audible utterances a label does not apply.

4. The non-transitory computer-readable medium of claim 3, wherein the occurrence of utterance data further comprises:

transcribed data associated with at least some of the previously collected audible utterances, and

labeled data providing information regarding a label associated with the occurrence of utterance data.

5. The non-transitory computer-readable medium of claim 1, wherein each of the plurality of datasets is stored in an XML database.

6. The non-transitory computer-readable medium of claim 1, wherein each of the plurality of datasets is stored in a relational database.

7. A method comprising:

storing via a processor a plurality of reusable components for building a natural language spoken dialog system, each of the plurality of reusable components comprising a plurality of groups of previously collected audible utterances and associated labels for call-types and named entities, wherein;

(1) the plurality of reusable components is organized into a plurality of datasets;

(2) each of the plurality of datasets comprises data pertaining to an industrial sector in a different task domain;

(3) data in the plurality of datasets is collected during a plurality of collection phases, each of the plurality of collection phases comprising a respective defined period of time;

(4) each group of the plurality of groups of previously collected audible utterances was collected in a separate spoken dialog system operating within a respective industry sector; and

(5) an annotation guide comprising guideline utterances and descriptions, the guideline utterances comprising both positive and negative utterances for an associated call-type category,

wherein the previously collected audible utterances are associated with an occurrence of utterance data comprising information indicating an associated call-type category, and wherein each respective industry sector is in a different task domain from other respective industry sectors.

8. The method of claim 7, further comprising:

storing the plurality of datasets, wherein each of the plurality of datasets further comprises a plurality of reusable components pertaining to one of a different industrial sector and a different task category.

9. The method of claim 7, wherein the plurality of datasets comprises a plurality of sets from the plurality of reusable components, each of the plurality of sets comprising data collected during a different data collection phase.

10. The method of claim 7, further comprising storing a set of data comprising at least one of the plurality of reusable components associated with data collected during a data collection phase.

11. The method of claim 7, further comprising assigning an attribute to one of the plurality of label information items.

12. The method of claim 11, wherein the attribute is at least one of a category attribute and a verb attribute.

13. The method of claim 7, wherein at least one of the plurality of guideline utterance items comprises transcription data.

14. The method of claim 11, wherein the attribute indicates whether the one of the plurality of labeled information items is at least one of generic, reusable, and specific to a given application.

15. The method of claim 7, wherein storing the plurality of reusable components further comprises:

storing information regarding a natural language understanding model;

storing information regarding an automatic speech recognition module; and

storing information regarding a named entity grammar.

16. The method of claim 7, wherein the plurality of reusable components are stored in a database.

17. The method of claim 16, wherein the database is at least one of an XML database and a relational database.

18. A system comprising:

a processor; and

a computer readable storage medium storing instructions for controlling the processor to perform steps comprising;

storing a plurality of reusable components for building a natural language spoken dialog system, each of the plurality of reusable components comprising a plurality of groups of previously collected audible utterances and associated labels for call-types and named entities, wherein;

(1) the plurality of reusable components is organized into a plurality of datasets;

(2) each of the plurality of datasets comprises data pertaining to an industrial sector in a different task domain;

(3) data in the plurality of datasets is collected during a plurality of collection phases, each of the plurality of collection phases comprising a respective defined period of time;

(4) each group of previously collected audible utterances was collected in a separate spoken dialog system operating within a respective industry sector; and

(5) an annotation guide comprising guideline utterances and descriptions, the guideline utterances comprising both positive and negative utterances for an associated call-type category,

wherein the previously collected audible utterances are associated with an occurrence of utterance data comprising information indicating an associated call-type category, and wherein each respective industry sector is in a different task domain from other respective industry sectors.

19. The system of claim 18, further comprising storing call-type information in each of the plurality of reusable components.

20. The system of claim 19, further comprising:

storing utterance data in each of the plurality of reusable components; and

associating the call-type information with the utterance data.

21. The system of claim 18, further comprising:

storing a plurality of sectors in a databases and

storing, in each of the plurality of sectors, a set of data comprising at least one of the plurality of reusable components associated with audible data collected during a different collection phase.

22. The system of claim 21, wherein each of the plurality of sectors comprises information pertaining to a different industrial sector.