DIALOG PROCESSING SYSTEM, DIALOG PROCESSING METHOD AND COMPUTER PROGRAM

• US 20090012787A1
• Filed: 07/03/2008
• Published: 01/08/2009
• Est. Priority Date: 07/03/2007
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Takeuchi, Hironori, Itoh, Nobuyasu, Negishi, Shiho
• Granted Patent: US 8,060,365 B2
• Time in Patent Office: Days
• US Class Current: 704/235
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 2015/088 Word spotting

Abstract

A dialog processing system which includes a target expression data extraction unit for extracting a plurality of target expression data each including a pattern matching portion which matches an utterance pattern, which are inputted by an utterance pattern input unit and is an utterance structure derived from contents of field-independent general conversations, among a plurality of utterance data which are inputted by an utterance data input unit and obtained by converting contents of a plurality of conversations in one field; a feature extraction unit for retrieving the pattern matching portions, respectively, from the plurality of target expression data extracted, and then for extracting feature quantity common to the plurality of pattern matching portions; and a mandatory data extraction unit for extracting mandatory data in the one field included in the plurality of utterance data by use of the feature quantities extracted.

Claims

1. A dialog processing system comprising:

an utterance pattern input unit for inputting an utterance pattern which is an utterance structure derived from contents of field-independent general conversations;

an utterance data input unit for inputting a plurality of utterance data obtained by converting contents of a plurality of conversations in one field into a text format;

a target expression data extraction unit for extracting a plurality of target expression data from the plurality of utterance data inputted by the utterance data input unit, the target expression data each including a pattern matching portion which matches any of the utterance pattern inputted by the utterance pattern input unit;

a feature extraction unit for retrieving each of the pattern matching portions from the plurality of target expression data extracted by the target expression data extraction unit, and then for extracting a feature quantity common to some of the plurality of pattern matching portions; and

a mandatory data extraction unit for extracting mandatory data in the one field by use of the feature quantities extracted by the feature extraction unit, the mandatory data included in the plurality of utterance data inputted by the utterance data input unit by use of the feature quantities extracted by the feature extraction unit.

2. The dialog processing system according to claim 1, further comprising:

a relaxed pattern generation unit for generating a relaxed pattern by use of the feature quantity extracted by the feature extraction unit; and

an utterance data evaluation unit for evaluating the plurality of utterance data by use of the relaxed pattern generated by the relaxed pattern generation unit, wherein

the conversations are through voice communications,

the utterance data input unit inputs the plurality of utterance data obtained by converting the conversations into a text format by use of speech recognition, and

the mandatory data extraction unit extracts the mandatory data in the one field from the plurality of utterance data according to the evaluation results obtained by the utterance data evaluation unit.

3. The dialog processing system according to claim 2, wherein the utterance data evaluation unit includes:

a calculation unit for calculating the numbers of appearances of the relaxed pattern by applying the relaxed pattern to the plurality of pattern matching portions, the plurality of target expression data except the plurality of pattern matching portions and the plurality of utterance data except the plurality of target expression data and

a mandatory pattern generation unit for generating a mandatory pattern from the relaxed patterns according to the numbers of appearances calculated by the calculation part; and

wherein the mandatory data extraction unit extracts the mandatory data in the one field from the plurality of utterance data according to the mandatory patterns generated by the mandatory pattern generation part.

4. The dialog processing system according to claim 3, wherein the mandatory pattern generation unit generates the mandatory patterns from the relaxed patterns according to an expected value of the number of errors calculated by the formula,

5. The dialog processing system according to claim 1, wherein the feature extraction unit extracts, as the feature quantity, at least any one of a word and an expression including the word, the word and expression included in the plurality of pattern matching portions and appearing a certain number of times or more.

6. The dialog processing system according to claim 1, wherein the feature extraction unit extracts the feature quantity by using a distribution of conversation times of the plurality of generated pattern matching portions.

7. The dialog processing system according to claim 2, wherein the relaxed pattern generation unit generates the relaxed pattern by use of the feature quantity and a component word that constitutes the utterance patterns.

8. The dialog processing system according to claim 1, wherein the feature extraction unit extracts a feature quantity common to the mandatory data in the one field by using, as the plurality of pattern matching portions, the mandatory data in the one field extracted by the mandatory data extraction unit.

9. The dialog processing system according to claim 1, wherein the mandatory data in the one field extracted by the mandatory data extraction unit is data that can be verified by an F measure calculated by the formula,

10. A computer program embodied in computer-readable memory comprising computer code for allowing a computer to execute the steps of:

inputting an utterance pattern which is an utterance structure derived from contents of field-independent, general conversations;

inputting a plurality of utterance data obtained by converting contents of a plurality of conversations in one field into a text format;

extracting, from the plurality of utterance data inputted, a plurality of target expression data each including a pattern matching portion which matches the inputted utterance patterns;

retrieving the pattern matching portions, respectively, from the plurality of target expression data extracted, and then extracting a feature quantity common to the plurality of pattern matching portions; and

extracting mandatory data in the one field included in the plurality of utterance data inputted by use of the feature quantities extracted.

11. A dialog processing method comprising the steps of:

inputting utterance patterns which are utterance structures derived from contents of field-independent, general conversations;

inputting a plurality of utterance data obtained by converting contents of a plurality of conversations in one field into a text format;

extracting a plurality of target expression data including pattern matching portions which match the inputted utterance patterns from the plurality of utterance data inputted;

retrieving the pattern matching portions, respectively, from the plurality of target expression data extracted so as to extract feature quantities common to a plurality of the pattern matching portions; and

extracting mandatory data in the one field included in the plurality of utterance data inputted by use of the feature quantities extracted.