Automatic conversation system and conversation scenario editing device

1. A conversation scenario editing device for an automatic conversation system using a conversation scenario,

the automatic conversation system including a conversation device and a conversation server,

the conversation device for generating an input sentence through speech recognition of an utterance by a user and requesting the conversation server to send a reply sentence to the input sentence, and

the conversation server for determining the reply sentence based on the conversation scenario when requested to send the determined reply sentence to the conversation device for an output of the reply sentence to the user,

the conversation scenario editing device comprising:

an editor configured to check whether or not an appropriate state transition relation is maintained in the conversation scenario edited thereby for editing the conversation scenario;

a language model generator for calculating occurrence possibility of a word string of the input sentence within the conversation scenario edited by the editor and generating, based on a calculating results thereof, a language model to be used when the conversation device executes the speech recognition; and

a timer to check a state where a predefined time period has elapsed without an utterance by the user,

wherein the editor edits the conversation scenario to include a reply sentence corresponding to the state where a predefined time period has elapsed without an utterance by the user,

wherein the reply sentence corresponding to the state where a predefined time period has elapsed without an utterance by the user is a suggestion for leading the user to select a route in the conversation scenario and

wherein, in a case where a violation of the appropriate state transition relation occurs in the conversation scenario, the editor generates a message for notifying present of the violation and the input sentence or the reply sentence causing the violation to an operator and then displays the message on an output unit.

2. The conversation scenario editing device according to claim 1 , wherein

the conversation scenario includes an object and an morphism, and

the conversation scenario editing device further comprises an editor for generating the conversation scenario in which the input sentence is the morphism and the reply sentence is the object that corresponds to the morphism.

3. The conversation scenario editing device according to claim 2 , further comprises

a dynamic knowledge generator for generating, from the conversation scenario, dynamic knowledge so as to search the object that corresponds to the morphism.

4. The conversation scenario editing device according to claim 1 , wherein

the editor also has a semantic interpretation dictionary, and has a function to rearrange or integrate redundant contents in the conversation scenario by using the semantic interpretation dictionary.

5. The conversation scenario editing device according to claim 1 , wherein

the conversation device includes an input unit for receiving contents of the user's utterance and converting them into audio signals, a speech recognition unit connected with the input unit, a conversation process unit connected with the speech recognition unit, an operation controller connected with the conversation process unit, and an output unit connected with the conversation process unit and the operation controller, wherein

the speech recognition unit receives the language model generated by the language model generator and converts audio signals from the input unit into the input sentence based on the language model.

6. The conversation scenario editing device according to claim 5 , wherein the speech recognition unit includes a preprocess unit that receives the audio signal, a feature extracting unit connected with the preprocess unit, and a specifying unit connected with the feature extracting unit 120 , wherein

the specifying unit includes a specification operation unit, an acoustic model memory connected with the specification operation unit, and a language model memory connected with the specification operation unit for storing the language model from the language model generator, wherein

the content stored in the language model memory is rewritten by the language model from a language model generator.

7. An automatic conversation system comprising:

a conversation device for generating an input sentence through speech recognition of an utterance by a user and for requesting a conversation server to send a reply sentence to the input sentence;

the conversation server for determining the reply sentence based on a conversation scenario when being requested to send the determined reply sentence to the conversation device for an output of the reply sentence to the user; and

a conversation scenario editing device that includes

an editor configured to check whether or not an appropriate state transition relation is maintained in the conversation scenario edited thereby for editing the conversation scenario;

a language model generator for calculating occurrence possibility of a word string of the input sentence included in the conversation scenario edited by the editor and generating, based on a calculating results thereof, a language model to be used when the conversation device executes the speech recognition; and

a timer to check a state where a predefined time period has elapsed without an utterance by the user,

wherein the editor edits the conversation scenario to include a reply sentence corresponding to the state where a predefined time period has elapsed without an utterance by the users,

wherein the reply sentence corresponding to the state where a predefined time period has elapsed without an utterance by the user is a suggestion for leading the user to select a route in the conversation scenario and

wherein, in a case where a violation of the appropriate state transition relation occurs in the conversation scenario, the editor generates a message for notifying present of the violation and the input sentence or the reply sentence causing the violation to an operator and then displays the message on an output unit.

8. The automatic conversation system according to claim 7 , wherein

the conversation scenario includes an objects and morphisms, and

the system further comprises an editor for generating the conversation scenario in which the input sentence is the morphism and the reply sentence is the object that corresponds to the morphism.

9. The automatic conversation system according to claim 8 , further comprises

a dynamic knowledge generator for generating, from the conversation scenario, dynamic knowledge so as to search the object that corresponds to the morphism.

10. The automatic conversation system according to claim 7 , wherein

the conversation server is adapted to send, to the conversation device, operation control information in which an operation to be executed by the conversation device is described; and

the conversation device is adapted to execute the operation based on the operation control information along with the output of the reply sentence.

11. The automatic conversation system according to claim 10 , wherein

the conversation device is a terminal device that is controlled based on the operation control information.

12. The automatic conversation system according to claim 7 , wherein

the editor also has a semantic interpretation dictionary, and has a function to rearrange or integrate redundant contents in the conversation scenario by using the semantic interpretation dictionary.

13. The automatic conversation system according to claim 7 , wherein

the conversation device includes an input unit for receiving contents of the user's utterance and converting them into audio signals, a speech recognition unit connected with the input unit, a conversation process unit connected with the speech recognition unit, an operation controller connected with the conversation process unit, and an output unit connected with the conversation process unit and the operation controller, wherein

the speech recognition unit receives the language model generated by the language model generator and converts audio signals from the input unit into the input sentence based on the language model.

14. The automatic conversation system according to claim 13 , wherein

the speech recognition unit includes a preprocess unit that receives the audio signal, a feature extracting unit connected with the preprocess unit, and a specifying unit connected with the feature extracting unit 120 , wherein

the specifying unit includes a specification operation unit, an acoustic model memory connected with the specification operation unit, and a language model memory connected with the specification operation unit for storing the language model from the language model generator, wherein

the content stored in the language model memory is rewritten by the language model from a language model generator.