JP2006251545A - Speech interaction system and computer program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a speech interaction system and a computer program that can maintain smooth interaction without causing an unnecessarily long waiting time for a user. <P>SOLUTION: The speech interaction system which receives and recognizes a speech on the basis of grammar for speech recognition to advance interaction on the basis of recognition results and interaction scenario information in which advance procedures of the interaction are described, and outputs an answer to the received speech analyzes the grammar for speech recognition and increases or decreases a wait time before a speech is received and/or a speech reception time for which a speech is continuously received on the basis of analysis results of the grammar for speech recognition. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a voice interaction system and a computer program capable of adjusting a response time of a user so that a dialogue automatically performed between the user and a computer in accordance with dialogue scenario information proceeds smoothly.

  In recent years, an interactive voice response (IVR) such as a voice portal using a speech recognition system (ASR: Auto Speech Recognition) has begun to spread. By using the voice dialogue system, it is possible to provide services such as a ticket reservation service and a courier re-delivery request service without assigning personnel to each service base. It is possible to enjoy great merits such as suppression of the above.

  On the other hand, since it responds automatically to a user's voice, it is an important subject to advance a smooth conversation. However, for example, if the volume of the user's voice is low and the voice dialogue system determines that it will not accept the voice from the user, the voice dialogue system will be in a standby state for voice input from the user, so the dialogue will be interrupted. It becomes difficult to maintain a smooth dialogue.

  Also, if the ambient noise volume is input in duplicate with the user's voice, the voice interaction system determines that the voice is continuously input by the user, and only for a predetermined time. There are cases where voice is continuously received. That is, for the user, although the voice is input, the voice dialogue system does not return a response, the dialogue is interrupted, and it is difficult to maintain a smooth dialogue.

  In order to solve such a problem, for example, in Patent Document 1, when voice by a user is not accepted, a waiting time for voice input is set for a certain period of time and surroundings that cannot be distinguished from voice input by the user Disclosed is a voice dialogue system that maintains a smooth dialogue by minimizing the dialogue interruption time by setting a maximum reception time for continuously accepting voice by the user when there is noise. .

Further, Patent Document 2 discloses a voice dialogue system that avoids interruption of a dialogue by advancing the dialogue according to a support scenario prepared in advance when voice input is not accepted for a certain period of time. .
Japanese Patent Application Laid-Open No. 8-076964 JP 2000-048038 A

  However, it is difficult for the voice dialogue system disclosed in Patent Document 1 to set an appropriate voice input waiting time and maximum reception time based on the progress of dialogue, and it is set assuming the worst case. By doing so, a time longer than necessary is often set. Therefore, the voice dialogue system often becomes unresponsive more than necessary, and there is a problem that it is difficult for the user to maintain a smooth dialogue. In addition, when a dialogue scenario is dynamically generated, there remains a problem that the waiting time for voice input and the maximum reception time cannot be set in advance.

  Furthermore, the speech dialogue system disclosed in Patent Document 2 determines the progress of the dialogue based on the presence or absence of voice input from the user, and it is necessary to create a dialogue scenario for support in every case. There was a problem that it was more difficult to create a dialogue scenario at the time of implementation.

  The present invention has been made in view of such circumstances, and it is an object of the present invention to provide a voice dialogue system and a computer program that can maintain a smooth dialogue without causing an excessive waiting time for a user. And

  In order to achieve the above object, the speech dialogue system according to the first aspect of the invention describes means for accepting speech, means for recognizing the speech received based on the speech recognition grammar, the recognition result and the progress of the dialogue. In a spoken dialogue system comprising means for advancing a dialogue based on dialogue scenario information and a means for outputting a response to the accepted speech, an analysis means for analyzing the speech recognition grammar, and an analysis of the speech recognition grammar A waiting time adjusting means for increasing or decreasing the waiting time until the voice is received based on the result.

  In the voice dialogue system according to the second invention, in the first invention, the voice recognition grammar specifies a voice pattern to be accepted and a keyword to be included in the voice, and the analysis means is for the voice recognition. Based on the grammar, the number of speech patterns that can be accepted and the number of speech keywords that are accepted are determined.

  The voice interaction system according to the third invention is characterized in that, in the first or second invention, the voice interaction system further comprises means for increasing / decreasing a voice reception time during which voice can be received based on an analysis result of the voice recognition grammar. And

  In the speech dialogue system according to the fourth invention, in the third invention, the speech recognition grammar specifies a speech pattern to be accepted, and the analysis means accepts based on the speech recognition grammar. The maximum number of characters or the maximum number of mora included in the character string when the speech pattern capable of being recognized is recognized is determined.

  The computer program according to the fifth aspect of the present invention receives speech, recognizes the received speech based on the speech recognition grammar, and advances the dialogue based on the recognition scenario and dialogue scenario information describing the progress of the dialogue. In the computer program executable by the computer that outputs a response to the received speech, the computer is allowed to analyze the speech recognition grammar until the speech is received based on the analysis result of the speech recognition grammar. It functions as a waiting time adjusting means for increasing or decreasing the waiting time.

  In the first invention and the fifth invention, the speech recognition grammar used in accordance with the dialogue scenario information is analyzed, and the waiting time until the speech is received is increased or decreased based on the analysis result of the speech recognition grammar. Thus, based on the speech recognition grammar, for example, for questions that require considerable think time for the user to respond, for questions that require a long waiting time and do not require so much thought time to respond The waiting time can be set short, and a smooth dialogue can be maintained without setting a dialogue interruption time unnatural for the user.

  In the second invention, the number of speech patterns that can be accepted and the number of speech keywords that are accepted are determined based on the speech recognition grammar. Thereby, the number of voice patterns expected as a response by the user and the number of keywords included in the response voice can be determined from the speech recognition grammar. For example, when the number of voice patterns and the number of keywords are large, the user Is a question that requires a considerable amount of thinking time to respond, and sets a long waiting time. When the number of voice patterns and the number of keywords are small, it is determined that the question does not require so much thought time for the user to respond, and the waiting time is set short. Therefore, it is possible to maintain a smooth conversation without setting a conversation interruption time unnatural for the user.

  In the third invention, the voice reception time during which voice can be received is increased or decreased based on the analysis result of the voice recognition grammar. Thus, based on the speech recognition grammar, for example, for a question with a large number of characters included in the utterance pattern necessary for the user to respond, the speech reception time during which speech can be received is set long and the number of characters is small. For questions, the voice reception time during which voice can be received can be set short, and smooth conversation can be maintained without setting an unnatural conversation interruption time for the user.

  In the fourth invention, the grammar for speech recognition is analyzed, and the maximum number of characters or the maximum number of mora included in the character string when speech patterns that can be accepted are recognized is determined. Thereby, based on the grammar for speech recognition, for example, the maximum number of characters or the maximum number of mora included in the utterance pattern necessary for the user to respond can be specified. For questions that have a long voice reception time that can receive voice and the maximum number of characters or maximum number of mora is small, the voice reception time that can receive voice can be set short, which is unnatural for the user. It is possible to maintain a smooth dialogue without setting a simple dialogue interruption time.

  According to the first and fifth inventions, based on the speech recognition grammar, for example, for a question that requires a considerable thinking time for the user to respond, a long waiting time is set and the thinking time is required to respond. For questions that do not require so much, the waiting time can be set short, and a smooth dialog can be maintained without setting an uninterrupted dialog interruption time for the user.

  In the second invention, the number of speech patterns expected as a response by the user and the number of keywords included in the response speech can be determined from the speech recognition grammar. For example, when the number of speech patterns and the number of keywords are large, It is determined that the question requires a considerable amount of thinking time for the user to respond, and the waiting time is set longer. When the number of voice patterns and the number of keywords are small, it is determined that the question does not require so much thought time for the user to respond, and the waiting time is set short. Therefore, it is possible to maintain a smooth conversation without setting a conversation interruption time unnatural for the user.

  In the third invention, based on the speech recognition grammar, for example, for a question with a large number of characters included in the utterance pattern necessary for the user to respond, the speech reception time during which speech can be received is set long, and the number of characters For questions with a small number of questions, the voice reception time during which voice can be received can be set short, and smooth conversation can be maintained without setting an unnatural conversation interruption time for the user. .

  In the fourth invention, based on the speech recognition grammar, for example, the maximum number of characters or the maximum number of mora included in the utterance pattern necessary for the user to respond can be specified. Can set a long voice reception time that can receive voice, and for questions with a small maximum number of characters or a maximum number of mora, the voice reception time that can receive voice can be set short. A smooth conversation can be maintained without setting an unnatural conversation interruption time.

  Hereinafter, the present invention will be specifically described with reference to the drawings showing embodiments thereof.

(Embodiment 1)
Hereinafter, the voice interaction system according to Embodiment 1 of the present invention will be specifically described with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a voice interaction system according to Embodiment 1 of the present invention. As shown in FIG. 1, the voice dialogue system according to the first embodiment includes a dialogue control device 10 including a voice input / output unit 20 that accepts a user's voice and outputs a response voice to the user. ing.

  The dialogue control device 10 includes at least a CPU (central processing unit) 11, a recording unit 12, a RAM 13, a communication interface 14 connected to an external communication unit, and an auxiliary recording unit 15 using a portable recording medium 16 such as a DVD or a CD. Consists of.

  The CPU 11 is connected to the above-described hardware units of the dialog control apparatus 10 via the internal bus 17, and controls the above-described hardware units and processes programs recorded in the recording unit 12, for example, use Various software functions are executed in accordance with a program for receiving a voice of a person, a program for recognizing the received voice, a program for reading out dialogue scenario information and generating a response to the recognition result, a program for reproducing and generating the generated response, and the like.

  The recording unit 12 includes a built-in fixed recording device (hard disk), a ROM, and the like, and has a dialog acquired from an external computer or a portable recording medium 16 such as a DVD or CD-ROM via the communication interface 14. A processing program necessary to function as the control device 10 is recorded. The recording unit 12 records not only the processing program but also dialogue scenario information 121 describing a dialogue scenario for performing an automatic response.

  The RAM 13 is composed of a DRAM or the like, and records temporary data generated when the software is executed. The communication interface 14 is connected to the internal bus 17 and can transmit and receive data necessary for processing by connecting so as to be able to communicate with an external network.

  The voice input / output unit 20 receives a user's voice by a voice input device such as a microphone, converts the voice data into voice data and sends the voice data to the CPU 11, and a synthesized voice corresponding to the generated response according to an instruction of the CPU 11. The function of reproducing and outputting from the audio output device is provided.

  The auxiliary recording unit 15 uses a portable recording medium 16 such as a CD or a DVD, and downloads a program, data, or the like processed by the CPU 11 to the recording unit 12. It is also possible to write the data processed by the CPU 11 for backup.

  In the first embodiment, the case where the dialogue control apparatus 10 and the voice input / output unit 20 are integrated will be described. However, the present invention is not particularly limited thereto, and one voice input / output unit 20 is provided. Alternatively, it may be connected to a plurality of dialogue control apparatuses 10 via a network or the like.

  The dialogue control apparatus 10 of the voice dialogue system according to the first embodiment of the present invention uses a command of the CPU 11 in accordance with the dialogue scenario information 121 stored in the recording means 12 in order to prompt the user to input voice. Audio output is performed from the audio input / output unit 20. For example, a question that can limit the next expected voice to the user to “Yes” or “No”, such as “Is your business XX”, is output as a voice. When recognizing a voice by a user, the voice is recognized based on a voice recognition grammar generated at the time of voice output or recorded as a part of the dialogue scenario information 121 in advance.

  The dialogue scenario information 121 is described so as to be able to accept a user's voice in the dialogue by, for example, VoiceXML (hereinafter referred to as VXML) scenario description language. That is, the dialogue scenario information 121 describes the contents of the output from the computer, the transition of the dialogue according to the voice uttered by the user, the processing to be performed next according to the voice content, and the like.

  When the user's voice is received from the voice input / output unit 20 with respect to the output voice, the recording means 12 is used as waveform data of the received voice or data indicating a feature value as a result of acoustic analysis of the received voice. In addition, voice recognition is performed on the voice recorded in the RAM 13 and in accordance with an instruction from the CPU 11. The speech recognition engine used for speech recognition processing is not particularly limited, and any speech recognition engine that is generally used may be used.

  The recording means 12 is not limited to a built-in hard disk, and records a large amount of data such as a hard disk built in another computer connected via the communication interface 14. Any recording medium can be used.

  The CPU 11 analyzes the speech recognition grammar used for the speech recognition process, and determines the number x of speech patterns that can be accepted and the number y of keywords to be included in the speech to be accepted. FIG. 2 is an exemplary diagram of voice patterns and keywords in the voice that can be received when a question such as “Is your business OO?” Is output by voice, for example.

  As shown in FIG. 2, in the case of a question that restricts the voice expected to the user to “Yes” or “No” next, the voice patterns that can be accepted are two patterns of “Yes” and “No”. It is. Therefore, it is possible to determine that the number of sound patterns that can be accepted x = 2.

  Further, since the number of keywords in the received voice is one, it can be determined that the number of keywords y = 1 to be included in the received voice.

  FIG. 3 is a diagram showing examples of voice patterns and keywords in the voice that can be received when a question such as “Please tell me the desired conditions for tickets” is outputted as voice. In this case, the next voice expected from the user is a combination of the departure / arrival station and seating conditions such as “from Tokyo”, “to Shin-Osaka”, “no smoking”,.

  As shown in FIG. 3, if the acceptable voice pattern is “from Tokyo”, “arriving at Shin-Osaka”, “no smoking”,... The number of possible voice patterns x = n can be determined. The number y of keywords in the received voice is the maximum number of words included in the voice pattern, and can be determined as y = 3 in the example of FIG.

  The CPU 11 analyzes the speech recognition grammar for each question, and determines the number x of speech patterns that can be accepted and the number y of keywords in the speech to be accepted for each question as the analysis result of the speech recognition grammar. Obtaining and setting a waiting time T1 for voice input from the user as a function of x and y as needed for each question. For example, the voice input waiting time T1 can be calculated as shown in (Expression 1), where a, b, and c are constants.

(Equation 1)
T1 = ax + by + c

  That is, the larger the number x of sound patterns that can be accepted, the greater the number of items to be selected for the user. Therefore, it is considered that the user's thinking time becomes longer. In addition, it is considered that the thinking time for organizing, classifying, and the like becomes longer for the user as the number y of keywords in the received voice is larger. Therefore, by changing the waiting time T1 for voice input from the user according to the speech recognition grammar, it is possible to continue a natural conversation for the user.

  FIG. 4 is a flowchart showing a processing procedure of the CPU 11 of the dialogue control apparatus 10 of the voice dialogue system according to Embodiment 1 of the present invention. The CPU 11 of the dialogue control apparatus 10 acquires the speech recognition grammar included in the dialogue scenario information 121 (step S401), and calculates the number of acceptable voice patterns (step S402). The CPU 11 calculates the number of keywords included for each acceptable voice pattern (step S403), and determines whether or not the calculated number of keywords is the maximum value (step S404). Specifically, the calculated number of keywords is recorded in the RAM 13, and the determination is made by comparing with the number of keywords calculated at any time.

  When the CPU 11 determines that the calculated keyword count is the maximum value (step S404: YES), the CPU 11 records the calculated keyword count as the maximum keyword count in the RAM 13 (step S405). The CPU 11 repeatedly executes the processing from step S403 on for all voice patterns (step S406: NO), and when the CPU 11 determines that the processing has been completed for all voice patterns (step S406: YES), the CPU 11 Based on the calculated number x of acceptable voice patterns and the calculated number y of keywords, a waiting time T1 for voice input by the user is calculated (step S407).

  The CPU 11 outputs a synthesized voice based on the dialogue scenario information (step S408), and determines whether or not a voice from the user has been received (step S409). When the CPU 11 determines that the user's voice has been received (step S409: YES), the CPU 11 recognizes the received voice (step S410) and proceeds to the next dialogue scenario. When the CPU 11 determines that the user's voice is not received (step S409: NO), the CPU 11 determines whether or not the calculated waiting time T1 has elapsed (step S411).

  The CPU 11 waits for voice input by the user until the calculated waiting time T1 elapses (step S411: NO), and when the CPU 11 determines that the calculated waiting time T1 has elapsed (step S411: YES), By executing predetermined error processing (step S412), the CPU 11 maintains a smooth dialogue without causing the user to feel that the dialogue has been interrupted unnaturally.

  The waiting time for each question in the dialogue scenario may be set in advance or may be set dynamically during the dialogue. If the speech recognition grammar is fixed, it is better to set it in advance, and if the speech recognition grammar is dynamically changed or added, the waiting time should also be set dynamically. become. Even if the recognition grammar is dynamically changed or added, a waiting time is set in advance, and the waiting time is recalculated when the speech recognition grammar is dynamically changed or added. You may make it do.

  Note that a correction coefficient for correcting the calculated speech input waiting time T1 may be recorded in the recording unit 12 in advance according to the content of the speech recognition grammar. FIG. 5 is a diagram schematically showing the procedure of the correction coefficient inquiry when the correction coefficient for correcting the voice input waiting time is recorded in the recording means 12.

  FIG. 5 shows three cases (a), (b), and (c) as examples of the speech recognition grammar. When the speech recognition grammar is FIG. 5A, the CPU 11 inquires the recording means 12 using the grammar content “destination response” as key information, and extracts the correction coefficient “1.8”. Similarly, when the speech recognition grammars are FIG. 5B and FIG. 5C, the CPU 11 inquires the recording means 12 using the grammar contents “order content confirmation” and “number confirmation” as key information, respectively. Correction coefficients “1.2” and “1.0” are extracted.

  In this way, the length of the thinking time by the user can be set according to the grammatical content, and even in the case of the same voice pattern (FIG. 5 (b), FIG. 5 (c)) is assumed. A correction coefficient that takes into account the length of thinking time by the user can be obtained, and by multiplying the obtained correction coefficient by the waiting time T1 for voice input, a waiting time that is more suitable for actual conversation can be set. A more natural dialogue can be realized.

  As described above, according to the first embodiment, the number of speech patterns assumed as a response by the user and the number of keywords included in the response speech can be determined from the speech recognition grammar. If the number and the number of keywords are large, it is determined that the question requires a considerable thinking time for the user to respond, and the waiting time is set longer. When the number of voice patterns and the number of keywords are small, it is determined that the question does not require so much thought time for the user to respond, and the waiting time is set short. Therefore, it is possible to maintain a smooth dialogue without causing an uninterrupted dialogue interruption time for the user.

  Also, by changing the correction coefficient for correcting the waiting time to be set according to the content of the recognition grammar, it is possible to set the waiting time according to the actual dialogue and maintain a more natural dialogue. It becomes possible. The correction coefficient may be changed according to the response speed to the question. For example, it can be estimated that the faster the response to the question, the more familiar the user is with the operation of the system, and the waiting time can be shortened by reducing the correction coefficient.

  In the first embodiment, an example is described in which the waiting time for speech input by the user based on the speech recognition grammar is calculated every time the synthesized speech is output from the speech dialogue system. However, at the start or before the start of the dialogue according to the dialogue scenario information, all grammars for speech recognition are analyzed, and the waiting time for speech input by the user is calculated based on the analysis result. It may be recorded in the recording means 12.

(Embodiment 2)
Hereinafter, the voice interaction system according to the second embodiment of the present invention will be specifically described with reference to the drawings. Since the configuration of the voice interaction system according to the second embodiment of the present invention is the same as that of the first embodiment, detailed description thereof will be omitted by attaching the same reference numerals. The second embodiment is different from the first embodiment in that an upper limit value of a time for continuously inputting a voice by a user is determined.

  The dialogue control apparatus 10 of the voice dialogue system according to the second embodiment of the present invention uses a command from the CPU 11 in accordance with the dialogue scenario information 121 stored in the recording means 12 in order to prompt the user to input voice. Audio output is performed from the audio input / output unit 20. For example, a question that can limit the next expected voice to the user to “Yes” or “No”, such as “Is your business XX”, is output as a voice. When recognizing a voice by a user, the voice is recognized based on a voice recognition grammar generated at the time of voice output or recorded as a part of the dialogue scenario information 121 in advance.

  The dialogue scenario information 121 including the speech recognition grammar is described so as to be able to accept the voice of the user in the dialogue by, for example, VoiceXML (hereinafter, VXML) scenario description language. That is, the dialogue scenario information 121 describes the content of the output from the computer, the transition of the dialogue according to the voice uttered by the user, the processing to be performed next according to the content of the voice, and the like.

  When the user's voice is received from the voice input / output unit 20 with respect to the output voice, the recording means 12 is used as waveform data of the received voice or data indicating a feature value as a result of acoustic analysis of the received voice. In addition, voice recognition is performed on the voice recorded in the RAM 13 and in accordance with an instruction from the CPU 11. The speech recognition engine used for speech recognition processing is not particularly limited, and any speech recognition engine that is generally used may be used.

  The recording means 12 is not limited to a built-in hard disk, and records a large amount of data such as a hard disk built in another computer connected via the communication interface 14. Any recording medium can be used.

  The CPU 11 analyzes the speech recognition grammar used for the speech recognition processing, and determines the maximum number of characters or the maximum number of mora included in the speech pattern that can be accepted. FIG. 6 is an exemplary diagram of speech patterns that can be accepted when a question such as “Is your business XX” is output by voice and the number of recognized characters that are included when the speech pattern is voice-recognized, for example. .

  As shown in FIG. 6, in the case of a question that restricts the voice expected to the user to “Yes” or “No” next, the voice patterns that can be accepted are two patterns of “Yes” and “No”. It is. Accordingly, since the number of recognized characters when the speech pattern that can be accepted is recognized is 2, 3 respectively, the maximum number of characters z = 3 can be determined.

  FIG. 7 is an illustration of a voice pattern that can be received when a question such as “which bullet train is used” is output by voice and the number of recognized characters included when the voice pattern is voice-recognized. In this case, the voice expected for the user is, for example, “Tohokushinkansen”, “Tokaishinkansen”, “Sanyoushinkansen”. Therefore, since the recognized character numbers when the speech patterns that can be accepted are recognized are 10, 12, and 10, respectively, the maximum number of characters z = 12 can be determined.

  The CPU 11 obtains the maximum number of characters z when speech patterns that can be accepted are recognized as speech recognition grammar analysis results, and the speech reception time T2 for continuously receiving speech from the user as a function of z. Is set at any time. For example, the voice reception time T2 can be calculated as (Equation 2), where d and e are constants.

(Equation 2)
T2 = dz + e

  That is, the larger the maximum number of characters z when a speech pattern that can be accepted is recognized, the longer the speech to be input for the user. Therefore, when the reception time of the voice input by the user is not set unnecessarily long and the ambient noise is erroneously detected, the user knows that the erroneous detection has been made in the same amount of time as a normal conversation. It is possible to maintain a more natural dialogue. Note that z may be the maximum number of mora included in a character string obtained by voice recognition of an acceptable voice. When a plurality of keywords are included in one utterance, z may be calculated in consideration of the time interval between keywords.

  FIG. 8 is a flowchart showing a processing procedure of the CPU 11 of the dialogue control apparatus 10 of the voice dialogue system according to Embodiment 2 of the present invention. The CPU 11 of the dialogue control apparatus 10 acquires the speech recognition grammar included in the dialogue scenario information 121 (step S801), and calculates the number of recognized characters when speech recognition is performed for each acceptable speech pattern (step S801). S802), it is determined whether or not the calculated number of recognized characters is the maximum value (step S803). Specifically, the calculated number of recognized characters is recorded in the RAM 13, and the determination is made by comparing with the number of recognized characters calculated as needed.

  If the CPU 11 determines that the calculated number of recognized characters is the maximum value (step S803: YES), the CPU 11 calculates the maximum number of characters or the maximum number of recognized mora when the speech pattern that can be accepted is recognized. Is recorded in the RAM 13 (step S804). The CPU 11 repeatedly executes the processing from step S802 on for all voice patterns (step S805: NO), and when the CPU 11 determines that the processing has been completed for all voice patterns (step S805: YES), the CPU 11 Based on the maximum number of characters or the maximum number of mora z when the recognized speech pattern that can be received is recognized, a speech reception time T2 for continuously receiving the speech from the user is calculated (step S806).

  The CPU 11 outputs a synthesized voice based on the dialogue scenario information (step S807), and determines whether or not the calculated voice reception time T2 has elapsed (step S808). The CPU 11 continues to accept the user's voice until it is determined that the voice reception time T2 has elapsed (step S808: NO). If the CPU 11 determines that the voice reception time T2 has elapsed (step S808: YES), the CPU 11 The input voice is recognized (step S809), and the process proceeds to the next dialogue scenario.

  As described above, according to the second embodiment, it is possible to determine the maximum number of characters or the maximum number of mora when the speech pattern assumed as a response by the user is recognized from the speech recognition grammar, By setting the voice reception time that can continuously accept the user's voice according to the maximum number of characters or the maximum number of mora, it is unnatural for the user even if the surrounding noise is falsely detected. It is possible to maintain a smooth dialogue without causing a dialogue interruption time.

  In the second embodiment, an example is described in which the voice reception time for continuously receiving the voice by the user based on the voice recognition grammar is calculated every time the synthesized voice is output from the voice dialogue system. However, the present invention is not limited to this. All speech recognition grammars are analyzed at the start or before the start of the dialog according to the dialog scenario information, and the user's voice is analyzed based on the analysis result. The voice reception time continuously received may be calculated and recorded in the recording unit 12.

  In addition, when used in combination with the first embodiment rather than the second embodiment alone, the interruption time of the conversation becomes a time that is more natural, and it is possible to maintain a smoother conversation. Become.

  Regarding the above first and second embodiments, the following additional notes are disclosed.

(Appendix 1)
Means for receiving audio;
Means for recognizing the received speech based on the speech recognition grammar;
Means for proceeding the dialogue based on the recognized scenario and the dialogue scenario information describing the procedure of the dialogue;
A voice dialogue system comprising: means for outputting a response to the received voice;
Analyzing means for analyzing the speech recognition grammar;
Waiting time adjusting means for increasing or decreasing the waiting time until receiving a voice based on the analysis result of the speech recognition grammar.

(Appendix 2)
The speech recognition grammar specifies a speech pattern to be accepted and a keyword to be included in the speech,
2. The spoken dialogue system according to claim 1, wherein the analysis means determines the number of voice patterns that can be accepted and the number of keywords of the voice to be accepted based on the grammar for voice recognition. .

(Appendix 3)
The spoken dialogue system according to claim 1 or 2, further comprising means for increasing or decreasing a voice reception time during which voice can be received based on an analysis result of the voice recognition grammar.

(Appendix 4)
The speech recognition grammar specifies a pattern of speech to be accepted,
The analyzing means is configured to determine a maximum number of characters or a maximum number of mora included in a character string when a recognizable speech pattern is speech-recognized based on the speech recognition grammar. The voice interaction system according to appendix 3.

(Appendix 5)
Accept audio,
Recognize the received speech based on the speech recognition grammar,
Based on the recognized results and dialog scenario information that describes the progress of the dialog, the dialog proceeds.
In the voice interaction method for outputting a response to the received voice,
Analyzing the speech recognition grammar;
A voice interaction method characterized by increasing or decreasing a waiting time until a voice is received based on an analysis result of the voice recognition grammar.

(Appendix 6)
The speech recognition grammar specifies a speech pattern to be accepted and a keyword to be included in the speech,
The voice interaction method according to claim 5, wherein the number of voice patterns that can be accepted and the number of voice keywords that are accepted are determined based on the voice recognition grammar.

(Appendix 7)
The voice interaction method according to appendix 5 or 6, wherein the voice reception time during which voice can be received is increased or decreased based on the analysis result of the voice recognition grammar.

(Appendix 8)
The speech recognition grammar specifies a pattern of speech to be accepted,
8. The voice interaction method according to appendix 7, wherein the maximum number of characters or the maximum number of mora included in a character string when speech patterns that can be accepted are recognized based on the speech recognition grammar is determined.

(Appendix 9)
A computer that accepts speech, recognizes the received speech based on the speech recognition grammar, advances the dialogue based on the recognized scenario and dialogue scenario information describing the progress of the dialogue, and outputs a response to the received speech In an executable computer program,
The computer,
A computer program that functions as an analysis unit that analyzes the speech recognition grammar, and a waiting time adjustment unit that increases or decreases a waiting time until a speech is received based on an analysis result of the speech recognition grammar.

(Appendix 10)
The speech recognition grammar specifies a speech pattern to be accepted and a keyword to be included in the speech,
The computer program according to appendix 9, wherein the analyzing means determines the number of speech patterns that can be accepted and the number of speech keywords to be accepted based on the speech recognition grammar.

(Appendix 11)
The computer,
The computer program according to appendix 9 or 10, wherein the computer program functions as means for increasing or decreasing a voice reception time during which a voice can be received based on an analysis result of the voice recognition grammar.

(Appendix 12)
The speech recognition grammar specifies a pattern of speech to be accepted,
The analyzing means is configured to determine a maximum number of characters or a maximum number of mora included in a character string when a recognizable speech pattern is recognized based on the speech recognition grammar. The computer program according to appendix 11.

It is a block diagram which shows the structure of the speech dialogue system which concerns on Embodiment 1 of this invention. It is an illustration figure of the keyword in the sound pattern and sound which can be accepted. It is an illustration figure of the keyword in the sound pattern and sound which can be accepted. It is a flowchart which shows the process sequence of CPU of the dialogue control apparatus of the voice dialogue system which concerns on Embodiment 1 of this invention. It is a figure which shows typically the procedure of a correction coefficient inquiry in case the correction coefficient which corrects the waiting time of audio | voice input is recorded on the recording means. It is an illustration figure of the number of recognition characters contained when carrying out the speech recognition of the speech pattern and speech pattern which can be received. It is an illustration figure of the number of recognition characters contained when carrying out the speech recognition of the speech pattern and speech pattern which can be received. It is a flowchart which shows the process sequence of CPU of the dialogue control apparatus of the voice dialogue system which concerns on Embodiment 2 of this invention.

Explanation of symbols

10 Dialogue control device 11 CPU
12 recording means 13 RAM
14 Communication Interface 15 Auxiliary Recording Unit 16 Portable Recording Medium 20 Audio Input / Output Unit 121 Dialog Scenario Information

Claims (5)

Means for receiving audio;
Means for recognizing the received speech based on the speech recognition grammar;
Means for proceeding the dialogue based on the recognized scenario and the dialogue scenario information describing the procedure of the dialogue;
A voice dialogue system comprising means for outputting a response to the received voice,
Analyzing means for analyzing the speech recognition grammar;
Waiting time adjusting means for increasing or decreasing the waiting time until receiving a voice based on the analysis result of the speech recognition grammar. The speech recognition grammar specifies a speech pattern to be accepted and a keyword to be included in the speech,
2. The voice dialogue according to claim 1, wherein the analyzing means determines the number of voice patterns that can be accepted and the number of voice keywords to be accepted based on the grammar for voice recognition. system.   3. The voice dialogue system according to claim 1, further comprising means for increasing / decreasing a voice reception time during which voice can be received based on an analysis result of the voice recognition grammar. The speech recognition grammar specifies a pattern of speech to be accepted,
The analyzing means is configured to determine a maximum number of characters or a maximum number of mora included in a character string when a recognizable speech pattern is recognized based on the speech recognition grammar. The voice interaction system according to claim 3. A computer that accepts speech, recognizes the received speech based on the speech recognition grammar, advances the dialogue based on the recognized scenario and dialogue scenario information describing the progress of the dialogue, and outputs a response to the received speech In an executable computer program,
The computer,
A computer program that functions as an analysis unit that analyzes the speech recognition grammar, and a waiting time adjustment unit that increases or decreases a waiting time until a speech is received based on an analysis result of the speech recognition grammar.

Images