Dialogue-sound processing apparatus and method

• US 5,761,637 A
• Filed: 08/02/1995
• Issued: 06/02/1998
• Est. Priority Date: 08/09/1994
• Status: Expired due to Term

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Chino, Tetsuro
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): MCFADDEN, SUSAN IRIS
• Application Number:
  US08/510,277
• Time in Patent Office: 1,035 Days
• Field of Search: 395/2.4-2.66, 395/2.79, 395/2.84
• US Class Current: 704/231
• CPC Class Codes: G10L 15/1807 using prosody or stress

Abstract

A dialogue-sound processing appratus of the present invention generates discourse structure representing the flow of dialogue from fragmentary spoken utterances. In the dialogue-sound processing apparatus, the speech fragments of the dialogue-sound is inputted through a sound input section. A clue extraction section extracts clue which is a word or prosodic feature representing flow of dialogue from the speech fragments. An utterance function rule memory section memorizes utterance function rule which is correspondence relation between the clue and the utterance function representing pragmatic effect for the flow of dialogue. An utterance function extraction section assigns the utterance function to the clue in accordance with the utterance function rule. A discourse structure generation section generates discourse structure representing the flow of dialogue from fragmentary spoken utterances in accordance with the utterance function corresponding to the clue assigned by the utterance function extraction section.

Claims

1. Dialogue-sound processing apparatus, comprising;

sound input means for inputting speech fragments of dialogue-sound in sequence;

clue extraction means for extracting a plurality of clues, each clue comprising a word or prosodic feature representing a flow of a dialogue from the speech fragments;

utterance function rule memory means for memorizing a plurality of utterance function rules, each rule defining a relation between one of the clues and an utterance function representing a pragmatic effect for the flow of the dialogue;

utterance function extraction means for assigning the utterance function to the clue extracted by said clue extraction means in accordance with the corresponding utterance function rule; and

discourse structure generation means for generating a discourse structure representing the flow of the dialogue of the speech fragments in accordance with the assigned utterance function.

2. Dialogue-sound processing apparatus according to claim 1,

wherein said discourse-structure generation means assigns constraint information to each speech fragment in accordance with the utterance function corresponding to the clue, and generates the discourse structure of a hierarchical tree by unification process or upward expanding process of each speech fragment in accordance with the constraint information.

3. Dialogue-sound processing apparatus according to claim 1,

wherein said discourse-structure of a hierarchical tree includes speech fragments located in order of time series along the flow of dialogue;

substantial utterance which is the speech fragment having substantial function corresponding to question, request, or acceptance of initiative-speaker;

the speech fragment including a communicative support having no substantial function;

the communication unit including one or more substantial utterances of one initiative-speaker and communicative support corresponding to the substantial utterance, which is a minimum unit of information between speakers of the dialogue.

4. Dialogue-sound processing apparatus according to claim 3,

wherein said discourse-structure of the hierarchical tree includes

a turn of initiation which is one or more communication units of one initiative-speaker;

a turn of response which is one or more communication units corresponding to the turn of initiation;

a turn of feedback which is one or more communication units corresponding to the turn of response; and

an exchange including at least the turn of initiation and the turn of response, which the one initiative-speaker has initiative of the exchange.

5. Dialogue-sound processing apparatus according to claim 4, wherein said discourse-structure of the hierarchical tree includes

discourse segments each of which is one or more exchanges of one initiative-speakers and one discourse which is one or more discourse segments.

6. Dialogue-sound processing apparatus according to claim 4, wherein said exchange includes at least the turn of initiation, the turn of response and an embedded level which is an embedded dialogue for correction of premise error or communicative support for resolving defect of dialogue-communication.

7. Dialogue-sound processing apparatus according to claim 4, wherein said exchange includes at least the turn of initiation, the turn of response, and a canceled level which is a rejection expression of transfer of initiative for utterance of opposite-speaker.

8. Dialogue-sound processing apparatus according to claim 6, wherein said discourse-structure generation means deletes or summarizes at least one of the speech fragment of the communication support and the speech fragment of the embedded level to simplify the discourse-structure.

9. Dialogue-sound processing method, comprising the steps of;

inputting speech fragments of the dialogue-sound in sequence;

extracting a plurality of clues,, each clue comprising a word or prosodic feature representing a flow of a dialogue from the speech fragment;

memorizing a plurality of utterance function rules, each rule defining a relation between one of the clues and an utterance function representing a pragmatic effect for the flow of the dialogue;

assigning the utterance function to the clue extracted at the extracting step in accordance with the corresponding utterance function rule; and

generating a discourse structure representing the flow of the dialogue of the speech fragments in accordance with the assigned utterance function.

10. Dialogue-sound processing apparatus, comprising;

sound input means for inputting user'"'"'s sound signal;

input analysis means for analysing the sound signal and for outputting input-intention information of the sound signal;

problem solution means for solving problem corresponding to the input-intention information and for outputting response-intention information as solution result;

output generation means for generating response information to the user in accordance with the response-intention information;

clue extraction means for extracting a plurality of clues, each clue comprising a word or prosodic feature representing a flow of a dialogue from speech fregments in the sound signal;

utterance function rule memory means for memorizing a plurality of utterance function rules, each rule defining a relation between one of the clues and an utterance function representing a pragmatic effect for the flow of the dialogue;

utterance function extraction means for assigning the utterance function to the clue extracted by said clue extraction means in accordance with the corresponding utterance function rule; and

discourse management means for generating a discourse structure representing the flow of the dialogue between the user'"'"'s sound signal and the response information in accordance with at least one of the assigned utterance function and the input-intention information, at least one of the response-intention information and the response information, and for controlling at least one of the analysis processing of said input analysis means, the solution processing of the problem solution means and the generation processing of the output generation means in accordance with the discourse structure.

11. Dialogue-sound processing method, comprising the steps of;

inputting user'"'"'s sound signal;

analysing the sound signal to output input-intention information of the sound-signal;

solving problem corresponding to the input-intention information to output response-intention information as solution result;

generating response information to the user in accordance with the response-intention information;

extracting a plurality of clues, each clue comprising a word or prosodic feature representing a flow of a dialogue from speech fragment in the sound signal;

memorizing a plurality of utterance function rules, each rule defining a relation between one of the clues and an utterance function representing a pragmatic effect for the flow of the dialogue;

assigning the utterance function to the clue extracted at the extracting step in accordance with the corresponding utterance function rule;

generating a discourse structure representing the flow of the dialogue between the user'"'"'s sound signal and the response information in accordance with at least one of the assigned utterance function and the input-intention information, at least one of the response-intention information and the response information; and

controlling at least one of the analysis processing at the analysing step, the solution processing at the solving step and the response information-generation processing at the response information-generating step in accordance with the discourse structure.