Voice processing device and method, and program

1. A voice processing device comprising:

score calculation means for calculating a score indicating compatibility of a voice signal input on the basis of an utterance of a user with each of plural pieces of intention information indicating each of a plurality of intentions;

pre-score calculation means for calculating a pre-score, for each of the plural pieces of intention information, that specifies an occurrence probability for each of the plural pieces of intention information, the pre-score being calculated based on transitioning through a state transition model based on words of the utterance of the user, a plurality of states in the state transition model corresponding to a predetermined conversation, the words of the utterance of the user being a condition that results in the transitioning, the score being based on the pre-score;

intention selection means for selecting the intention information indicating the intention of the utterance of the user among the plural pieces of intention information on the basis of the score calculated by the score calculation means; and

intention reliability calculation means for calculating the reliability of the intention information selected by the intention selection means on the basis of the score calculated by the score calculation means.

2. The voice processing device according to claim 1 , wherein the score calculation means includes:

acoustic score calculation means for calculating an acoustic score as the score indicating an acoustic compatibility of the voice signal with intention information, for each of the plural pieces of intention information; and

language score calculation means for calculating a language score as the score indicating a language compatibility of the voice signal with intention information, for each of the plural pieces of intention information.

3. The voice processing device according to claim 1 , wherein the pre-score calculation means adjusts the pre-score for each of the plural pieces of intention information in response to a context at the time when the voice signal is input.

4. The voice processing device according to claim 3 , wherein the score calculation means calculates the score of a voice zone in the voice signal.

5. The voice processing device according to claim 3 , wherein the score calculation means further calculates the score of the intention information which does not indicate a specific intention among the plural pieces of intention information, and

wherein the intention selection means determines, in a case where the intention information which does not indicate the specific intention is selected, that the utterance of the user corresponding to the voice signal is a meaningless utterance which does not indicate a specific intention.

6. The voice processing device according to claim 3 , wherein the score calculation means gives a weight to each of the acoustic score, the language score and the pre-score, and

wherein the intention reliability calculation means adjusts each weight, and calculates the reliability on the basis of the scores each given the adjusted weight.

7. The voice processing device according to claim 3 , further comprising management means for managing an output based on the intention information selected by the intention selection means.

8. The voice processing device according to claim 1 , wherein the score calculation means includes:

acoustic score calculation means for calculating an acoustic score as the score indicating an acoustic compatibility of the voice signal with intention information, for each of the plural pieces of intention information, the acoustic score calculating means including a grammar database including a concatenation probability of at least one of a bigram grammar or a trigram grammar.

9. The voice processing device according to claim 1 ,

wherein each of the plural pieces of intention information corresponds to one or more example sentences, the example sentences of each intention information being used to generate a dictionary and a grammar rule for the intention information, and the dictionary and the grammar rule being used by the score calculation means for calculating the score corresponding to the intention information.

10. A voice processing method comprising the steps of:

calculating a score indicating compatibility of a voice signal input on the basis of an utterance of a user with each of plural pieces of intention information indicating each of a plurality of intentions;

calculating a pre-score, for each of the plural pieces of intention information, that specifies an occurrence probability for each of the plural pieces of intention information, the pre-score being calculated based on transitioning through a state transition model based on words of the utterance of the user, a plurality of states in the state transition model corresponding to a predetermined conversation, the words of the utterance of the user being a condition that results in the transitioning, the score being based on the pre-score;

selecting the intention information indicating the intention of the utterance of the user among the plural pieces of intention information on the basis of the score calculated by a process of the score calculation step; and

calculating the reliability of the intention information selected by a process of the intention selection step on the basis of the score calculated by the process of the score calculation step.

11. A non-transitory computer-readable storage medium having computer readable program codes embodied in the computer readable storage medium that, when executed cause a computer to execute:

calculating a score indicating compatibility of a voice signal input on the basis of an utterance of a user with each of plural pieces of intention information indicating each of a plurality of intentions;

calculating a pre-score, for each of the plural pieces of intention information, that specifies an occurrence probability for each of the plural pieces of intention information, the pre-score being calculated based on transitioning through a state transition model based on words of the utterance of the user, a plurality of states in the state transition model corresponding to a predetermined conversation, the words of the utterance of the user being a condition that results in the transitioning, the score being based on the pre-score;

selecting the intention information indicating the intention of the utterance of the user among the plural pieces of intention information on the basis of the calculated score; and

calculating the reliability of the intention information selected by a process of the intention selection step on the basis of the calculated score.

12. A voice processing device comprising:

at least one hardware processor;

a score calculation unit, implemented by the at least one hardware processor, configured to calculate a score indicating compatibility of a voice signal input on the basis of an utterance of a user with each of plural pieces of intention information indicating each of a plurality of intentions;

a pre-score calculation unit for calculating a pre-score, for each of the plural pieces of intention information, that specifies an occurrence probability for each of the plural pieces of intention information, the pre-score being calculated based on transitioning through a state transition model based on words of the utterance of the user, a plurality of states in the state transition model corresponding to a predetermined conversation, the words of the utterance of the user being a condition that results in the transitioning, the score being based on the pre-score;

an intention selection unit, implemented by the at least one hardware processor, configured to select the intention information indicating the intention of the utterance of the user among the plural pieces of intention information on the basis of the score calculated by the score calculation unit; and

an intention reliability calculation unit, implemented by the at least one hardware processor, configured to calculate the reliability of the intention information selected by the intention selection unit on the basis of the score calculated by the score calculation unit.