JP2000293194A - Voice interactive device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a voice interactive device in which interactive procedures are determined to achieve an interactive objective most efficiently in accordance with a user. SOLUTION: The device is provided with a voice recognition section 1 which outputs a voice recognition result, an interactive procedure storage section 2 which holds a voice recognition object vocabulary, the voice recognition result and interactive procedures that define transition destination interactive states corresponding to the number of malrecognitions, a voice recognition correct/ error number storage section 3 which holds the number of correct/errors in voice recognition, a transition destination interactive state determining section 4 which determines and outputs a transition destination interactive state while referring to the interactive procedures held in the section 2, and an interactive control section 5 which transitions an interactive state to the transition destination interactive state outputted by the section 4. Thus, interactive procedures, in which an interactive objective is most efficiently achieved in accordance with the user.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a spoken dialogue apparatus used for a natural language man-machine interface.

[0002]

2. Description of the Related Art Speech dialogue with a device has increased the importance of a voice dialogue device for obtaining information required by a user. In such a spoken dialogue device,
It is important to control the dialogue to obtain the information that the user needs efficiently. Conventionally, for such a purpose, the average number of spoken dialogues is estimated, and the dialogue procedure is performed based on the estimated value. A method of setting has been proposed.

[0003] A conventional voice interaction apparatus will be described with reference to the drawings. FIG. 18 shows, for example, Japanese Patent Application Laid-Open No. 10-091.
BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a diagram showing a configuration of a conventional voice interaction procedure generation device disclosed in Japanese Patent Publication No. 188.

[0004] In the conventional speech dialogue procedure generating apparatus configured as described above, in the overall dialogue repeat count evaluation processor, the basic dialogue decomposer decomposes the dialogue procedure into basic dialogues, and the basic dialogue repeater evaluation processor treats the phoneme. The number of repetitions of each basic dialogue is evaluated using the estimated recognition rate obtained from the misrecognition matrix and the vocabulary, and a basic dialogue repetition totaling unit sums and outputs the number of repetitions of each basic dialogue. The minimum selection output unit selects the minimum value from the outputs of the evaluation units for the total number of repetitions of each dialog and determines the dialog procedure.

[0005]

However, in the above-described conventional speech dialogue procedure generation apparatus, the estimated recognition rate used for estimating the number of times of dialogue is determined by the phoneme misrecognition matrix obtained in advance from the actual utterance. It is obtained from a predetermined vocabulary, and does not represent the recognition rate of a user who is inputting voice to the device. Therefore, since the estimated number of repetitions of the dialogue is not the number of repetitions reflecting the speech recognition rate of a specific user, the determined dialogue procedure does not necessarily ensure that the user achieves the purpose of the dialogue most efficiently. There was a problem.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems, and has as its object to provide a voice interactive device capable of determining a dialog procedure for achieving a dialog purpose most efficiently according to a user. I do.

[0007]

According to a first aspect of the present invention, there is provided a speech dialogue apparatus for performing a recognition process on an input speech and outputting a speech recognition result, and a speech recognition target in each dialogue state. A dialogue procedure storage unit that holds a dialogue procedure that defines a transition destination dialogue state according to a vocabulary, a speech recognition result, and the number of erroneous recognitions; a speech recognition true / false number storage unit that holds the number of correctness / errors of speech recognition; Based on the number of correct / incorrect speech recognition held in the number-of-times storage unit and the speech recognition result output by the voice recognition unit, determine the transition destination dialog state with reference to the dialog procedure held in the dialog procedure storage unit. A transition destination dialog state determination unit to output, and a dialogue tube that outputs a correct / incorrect result with respect to the speech recognition result output by the voice recognition unit and transitions the dialog state to the transition destination dialog state output by the transition destination dialog state determination unit It is obtained by a part.

According to a second aspect of the present invention, there is provided a voice interaction apparatus for performing a recognition process on an input voice and outputting a voice recognition result, and a voice recognition target vocabulary, a voice recognition result, and a voice recognition result in each dialogue state. An interaction procedure storage unit that holds an interaction procedure that defines a transition destination interaction state according to an assumed recognition rate, a speech recognition accuracy / error number storage unit that holds the number of times of speech recognition error, and an audio recognition accuracy / number of error storage unit that is held An expected speech recognition rate test unit that performs a test on an assumed recognition rate defined in the current conversation state based on the number of correct / incorrect speech recognitions performed, and outputs all of the not-rejected assumed recognition rates; With reference to the dialogue procedure held in the above, from the speech recognition result outputted by the speech recognition unit and the transition destination conversation state corresponding to the assumed recognition rate output by the assumed speech recognition rate test unit, the transition destination dialog state A transition destination dialog state determining unit that determines and outputs one, and outputs a correct / incorrect result for the speech recognition result output by the speech recognition unit, and sets the dialog state to the transition destination dialog state output by the transition destination dialog state determination unit And a dialogue management unit that transitions to

According to a third aspect of the present invention, in the speech dialogue apparatus, the dialogue management unit determines that the destination dialogue state output by the destination dialogue state determination unit is a dialogue end state, and that the user's dialogue purpose is If not achieved, the dialog with the user is terminated and switched to the operator.

According to a fourth aspect of the present invention, in the voice dialogue apparatus, the dialogue procedure storage unit holds a dialogue procedure defining an average number of dialogues up to an end dialogue state in each dialogue state, and determines the transition destination dialogue state. The unit refers to the dialogue procedure stored in the dialogue procedure storage unit, and recognizes a speech recognition result output by the speech recognition unit and a transition destination dialogue corresponding to the assumed recognition rate output by the assumed speech recognition rate test unit. The transition destination conversation state is determined to be one based on the average number of conversations from the state to the end conversation state, and is output.

According to a fifth aspect of the present invention, in the speech dialogue apparatus, the dialogue procedure storage unit holds a dialogue procedure defining a speech recognition rate distribution in each dialogue state, and is held in the speech recognition correct / incorrect count storage unit. Using the number of correct and incorrect speech recognition
A speech recognition rate estimator for estimating and outputting a user's speech recognition rate up to the current conversation state, a speech recognition rate output by the speech recognition rate estimator, and a speech recognition rate distribution in the current conversation state. A voice recognition success possibility determining unit that determines the possibility that the user's input is correctly recognized and outputs a determination result, wherein the dialog management unit determines the possibility of the speech recognition success possibility determining unit. Based on the result, the dialog with the user is terminated and switched to the operator.

According to a sixth aspect of the present invention, there is provided a speech recognition apparatus for storing an estimated speech recognition rate of a user in each conversation state up to the conversation state and a history of correctness of speech recognition results in the conversation state. A speech recognition rate distribution for calculating a speech recognition rate distribution in each dialogue state with reference to the true / false history accumulating section and the speech recognition true / false history accumulating section, and updating the speech recognition rate distribution held in the dialogue procedure storage section. An update unit is further provided.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiment 1 A voice interactive device according to Embodiment 1 of the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing a configuration of a voice interaction device according to Embodiment 1 of the present invention. In the drawings, the same reference numerals indicate the same or corresponding parts.

In FIG. 1, reference numeral 1 denotes a speech recognition unit which performs a recognition process on an input speech and outputs a speech recognition result, and 2 denotes a speech recognition target vocabulary, a speech recognition result, and the number of erroneous recognitions in each dialogue state. A dialogue procedure storage unit that holds a dialogue procedure that defines a transition destination dialogue state, 3 is a speech recognition correct / error count storage unit that holds the number of correct / incorrect counts of voice recognition, and 4 is a speech recognition A transition destination dialog state determination unit that determines and outputs a transition destination dialog state with reference to the dialog procedure stored in the dialog procedure storage unit 2 based on the number of correct / errors and the speech recognition result output by the speech recognition unit 1, 5 Outputting a correct or incorrect result with respect to the recognition result output by the voice recognition unit 1;
This is a dialog management unit that changes the dialog state to the dialog state output by the transition destination dialog state determination unit 4.

Next, the operation of the voice interaction apparatus according to the first embodiment will be described with reference to the drawings. FIG.
FIG. 3 and FIG. 3 are diagrams showing an example of a dialogue procedure stored in a dialogue procedure storage unit of the voice dialogue apparatus according to Embodiment 1 of the present invention.

Hereinafter, a specific operation of the case where the voice interaction apparatus is used for telephone number guidance will be described. A telephone number guidance voice interactive device is a device in which a user interacts with the device by voice and inputs item information such as an address and a target name necessary for telephone number guidance, and the device makes a telephone number based on the input items. Is performed, and a telephone number is provided to the user.

[0017] For example, in the dialog state S ₁₀ shown in the upper part of FIG. 2, all the prefecture name in Japan as the voice recognition target words V _10, table T ₁₀ transition destination dialog state in accordance with the speech recognition result and misrecognition number Is stipulated. Table T ₁₀ transition destination conversational state, the transition destination dialog state regardless of the misrecognition number when the speech recognition result is, for example, "Kanagawa" S
It is ₃₅ .

The table T ₃₅ of the transition destination conversation state shown in the lower part of FIG. 2 shows that the speech recognition result is “Yes”. For example, when the number of erroneous recognitions is two or less, the transition destination conversation state is S.
₁₂₀ , the speech recognition result is "yes" and the number of false recognitions is 3
If the number is not less than 5 and not more than 5 times, it indicates that the transition destination conversation state is _S121 .

[0019] Each conversation state, the voice recognition target words, it is possible to describe the interactive control information other than the transition destination dialog state, in the upper dialog state S ₁₀ in FIG. 2, for example, in response to the user response sentence a ₁₀ "please enter the prefecture name" is defined as.

FIG. 4 shows an example of the number of correct / incorrect voice recognitions held in the voice recognition correct / incorrect number storage unit 3. As shown in FIG. From the start of the dialogue with the user to the current state of the dialogue, the number of times that the speech recognition result was correct is “7” times, and the number of times that the speech recognition result was incorrect is “2” times. I have.

The user right or wrong number is 4 for speech recognition held in the speech recognition errata count storage unit 3, dialog state S ₁₀
Will be described.

[0022] Upon reaching the dialogue state S _10, dialogue management unit 5
Refers to the dialogue procedure for dialog state S ₁₀ shown in FIG. 2 held dialogue procedure storage unit 2 responds with "Please enter a prefecture name" to the user. When the user inputs "Kanagawa", the voice recognition unit 1 performs voice recognition on the input voice and outputs a recognition result "Kanagawa".

The destination dialog state determination unit 4 refers to the table T ₁₀ of the destination dialog state in the dialog state S ₁₀ shown in FIG. a speech recognition result "Kanagawa" to be the misrecognition number held in the speech recognition errata frequency storage section 3 "2", a transition destination dialog state and determines that S ₃₅ output.

The dialogue management unit 5 is a transition destination dialogue state determination unit 4
The current dialog state is transited to the transition destination dialog state S ₃₅ output by the user, and the user refers to the dialog procedure in the dialog state S ₃₅ shown in the lower part of FIG. "Is Kanagawa".

When the user inputs "yes", the voice recognition unit 1 performs voice recognition on the input voice and outputs a voice recognition result "yes".

The dialogue management unit 5 determines that the recognition result "Kanagawa" is a correct recognition result based on the speech recognition result "Yes" for the acknowledgment "Kanagawa is", and determines that correct recognition has occurred. The number of correct recognitions output to the storage unit 3 and stored in the number of correctness / unreliability counts of speech recognition 3 is updated to “8”.

The destination dialog state determination unit 4 refers to the table T ₃₅ of the destination dialog state in the dialog state S ₃₅ shown in the lower part of FIG.
There the speech recognition result "Yes" to be output, from the misrecognition number held in the speech recognition errata frequency storage section 3 "2", and outputs a transition destination dialog state to determine the S _120.

The dialog management unit 5 is a transition destination dialog state determination unit 4
There shifts the current dialog state to transition destination dialog state S ₁₂₀ to output, with reference to the dialogue procedure in dialog state S ₁₂₀ shown in the middle of FIG 3 held in the interactive procedure storage unit 2, to the user And reply, "Please enter the address below the prefecture name." On the other hand, the user inputs, for example, “It is Ofuna in Kamakura” and continues the dialogue.

On the other hand, the user correctness number of speech recognition held in the speech recognition errata count storage unit 3 is the number of times shown in FIG. 4, type "Kanagawa" in the dialog state S _10, the speech recognition unit 1 A case in which “Kagawa” is erroneously recognized will be described.

The transition destination dialog state determination unit 4 refers to the table T ₁₀ transition destination dialog state in a dialog state S ₁₀ shown in the upper part of FIG. 2 held in the interactive procedure storage unit 2, the speech recognition unit 1
Based on the voice recognition result “Kagawa” output by the device and the number of incorrect recognitions “2” held in the voice recognition correct / incorrect number storage unit 3, the transition destination dialog state is determined to be _S53 and output.

The dialog management unit 5 includes a transition destination dialog state determination unit 4
The current dialog state is transited to the transition destination dialog state S ₅₃ output by the user, and the user refers to the dialog procedure in the dialog state S ₅₃ shown in the upper part of FIG. And responds, "It's Kagawa."

When the user inputs "No", the voice recognition unit 1 performs voice recognition on the input voice and outputs a voice recognition result "No".

The dialogue management unit 5 confirms the response "I'm Kagawa."
Based on the speech recognition result “No”, the recognition result “Kagawa” is determined to be a recognition error, and the fact that erroneous recognition has occurred is output to the speech recognition correct / incorrect number storage unit 3. The held false recognition count is updated to “3”.

The transition destination dialog state determination unit 4 refers to the table T ₅₃ transition destination dialog state in a dialog state S ₅₃ shown in the upper part of FIG. 3 held in the interactive procedure storage unit 2, the speech recognition unit 1
There the speech recognition result "No" to be output, from the misrecognition number held in the speech recognition errata frequency storage unit 3 "3", and outputs a transition destination dialog state to determine the S _10.

Dialogue state S_TenAgain, the user
And input "Kanagawa".
When the river is recognized, the transition destination dialog state determination unit 4
State S _TenTable T of the transition destination conversation state in_TenSee
Transition from the voice recognition result "Kanagawa" and the number of incorrect recognition "3"
Set the destination dialog state to S₃₅Is determined and output.

The dialogue management unit 5, to the transition destination dialogue state S ₃₅ to transition the current dialog state, with reference to the dialogue procedure in the interactive state S _35, responds that "It is Kanagawa" to the user ,
When the user inputs “Yes”, the voice recognition unit 1 outputs a voice recognition result “Yes”.

Based on the speech recognition result "Yes" for the acknowledgment "Kanagawa is", the dialog management unit 5 judges that the recognition result "Kanagawa" is a correct recognition result, and determines that correct recognition has occurred. The number of correct recognitions output to the storage unit 3 and stored in the number of correctness / unreliability counts of speech recognition 3 is updated to “8”.

The transition destination dialog state determination unit 4 determines the dialog state S ₃₅
With reference to the table T ₃₅ transition destination dialog state in a speech recognition result "yes" output by the speech recognition unit 1, the misrecognition number held in the speech recognition errata frequency storage unit 3 "3", the transition The previous conversation state is determined as _S121 and output.

The dialogue management unit 5, the current dialog state to transition from S ₃₅ to S _121, with reference to the dialogue procedure in the dialog state S ₁₂₁ shown in the lower part of FIG. 3, "a city or to the user Please enter a county name. " On the other hand, the user inputs, for example, “Kamakura” and continues the dialogue.

By the above operation, a user who has a small number of times of occurrence of erroneous recognition can select a dialogue procedure such as "dialogue state _S120 " in which the vocabulary to be recognized is enlarged and the number of times of dialogue is reduced. For a user who has a large number of times of recognition, a dialogue procedure such as “dialogue state S ₁₂₁ ” can be selected in which the number of dialogues increases but the vocabulary to be recognized is reduced to reduce erroneous recognition. Therefore, since an optimal dialog procedure according to the user's voice recognition rate can be selected, the purpose of the dialog can be achieved most efficiently according to the user.

Embodiment 2 Embodiment 2 A voice dialogue apparatus according to Embodiment 2 of the present invention will be described with reference to the drawings. FIG. 5 is a diagram showing a configuration of a voice interaction device according to Embodiment 2 of the present invention.

In FIG. 5, 1 is a voice recognition unit, 2 is a dialogue procedure storage unit, 3 is a voice recognition correct / incorrect count storage unit, 4 is a transition destination dialogue state determination unit, 5 is a dialogue management unit, and 6 is an assumed voice recognition rate. It is a test section.

Next, the operation of the voice interaction apparatus according to the second embodiment will be described with reference to the drawings. FIG.
And FIG. 7 are diagrams showing an example of a dialogue procedure of the voice dialogue device according to Embodiment 2 of the present invention.

The operation of the dialogue procedure storage unit 2, the transition destination dialogue state determination unit 4, and the assumed speech recognition rate test unit 6 will be described. The operations of the voice recognition unit 1, the voice recognition correct / incorrect number storage unit 3, and the dialogue management unit 5 are the same as those in the first embodiment, and a description thereof will be omitted.

[0045] For example, in the dialog state S ₁₀ shown in the upper part of FIG. 6, all the prefecture name in Japan as the voice recognition target words V _10, table T ₁₀ transition destination dialog state in accordance with the speech recognition result and the expected recognition rate Is stipulated. Table T ₁₀ transition destination conversational state, the speech recognition result is a transition destination dialog state regardless assumed recognition rate in the case of "Kanagawa" indicates that the S _35. In the table T ₃₅ of the transition destination dialog state shown in the lower part of FIG. 6, the speech recognition result is “Yes”, and when the assumed recognition rate for the user is 90%, the transition destination dialog state is S ₁₂₀ , and the voice is If the recognition result is "yes",
If the assumed recognition rate for the user is 80%, this indicates that the transition destination dialog state is _S121 .

The user errata number of speech recognition held in the speech recognition errata count storage unit 3 is the number of times shown in FIG. 4, the operation when it reaches the dialog state S _10.

[0047] Upon reaching the dialogue state S _10, dialogue management unit 5
Refers to the dialogue procedure for dialog state S ₁₀ shown in the upper part of FIG. 6, which is held in the interactive procedure storage unit 2 responds with "Please enter a prefecture name" to the user. When the user inputs “Kanagawa”, the voice recognition unit 1 performs voice recognition on the input voice and outputs a recognition result “Kanagawa”.

The assumed speech recognition rate test section 6 does not perform a test because the assumed recognition rate for the speech recognition result "Kanagawa" is arbitrary.

The transition destination dialog state determination unit 4 refers to the table T ₁₀ transition destination dialog state in a dialog state S ₁₀ shown in the upper part of FIG. 6, which is held in the interactive procedure storage unit 2, the speech recognition unit 1
But the transition destination dialogue state determines and outputs the S ₃₅ from the output to the speech recognition result "Kanagawa".

[0050] for the response "It is Kanagawa" in the dialogue state S ₃₅ shown in the lower part of FIG. 6, when the user inputs "yes", dialogue management unit 5 speech recognition errata number of times that a correct recognition occurs The number of correct recognitions output to the storage unit 3 and stored in the number of correctness / unreliability counts of speech recognition 3 is updated to “8”.

The assumed speech recognition rate test unit 6 sets the dialogue state S ₃₅
The hypothesis test is performed on the number of correct and incorrect voice recognitions held in the voice recognition correct and incorrect number storage unit 3 at a predetermined risk rate with the assumed recognition rates of 90% and 80% as the hypothesis with reference to the dialog procedure in.

[0052] The hypothesis tests, determined u for observations by the formula shown in FIG. 8, obtained using normal distribution table to u ₀ for hazard ratio, determines rejection of the hypothesis by comparing the u and u ₀ Since there is a known means for performing the above, it is used. In addition,
In FIG. 8, p is a hypothesis, k is the number of correct recognitions, and n is the total number of speech recognitions, that is, the sum of the number of correct recognitions and the number of incorrect recognitions.

When the total number of times of recognition is 10 and the number of times of correct answer recognition is 8 and a test is performed on a hypothesis of 90% with a risk rate of 10%, u = 1.154 and u ₀ = 1.282.
u <u _{0 and the} hypothesis is not rejected. When a test is performed on the hypothesis 80%, u = 0, so that u <u _{0 and the} hypothesis is not rejected. Therefore, the assumed speech recognition rate test unit 6
Outputs 90% and 80% as test results.

The transition destination dialog state determination unit 4 selects, for example, 90%, which is the largest of the assumed recognition rates of 90% and 80% output from the assumed speech recognition rate test unit 6. The criteria for selection are:
The designer presupposes that the user is assumed to be a user having the highest recognition rate as much as possible and selects the largest assumed recognition rate in order to complete the conversation with a small number of conversations without limiting the voice input as much as possible.

The destination dialog state determination unit 4 refers to the table T ₃₅ of the destination dialog state in the dialog state S ₃₅ shown in the lower part of FIG.
There the output to the speech recognition result "Yes", from the determined assumed recognition rate of 90%, and outputs the transition destination dialog state to determine the S _120.

The dialog management unit 5 includes a transition destination dialog state determination unit 4
There shifts the current dialog state to transition destination dialog state S ₁₂₀ to output, with reference to the dialogue procedure in dialog state S ₁₂₀ shown in the middle of FIG 7 held in the interactive procedure storage unit 2, to the user And reply, "Please enter the address below the prefecture name." On the other hand, the user inputs, for example, “It is Ofuna in Kamakura” and continues the dialogue.

Meanwhile, user correctness number of speech recognition held in the speech recognition errata count storage unit 3 is the number of times shown in FIG. 4, type "Kanagawa" in the dialog state S _10, the speech recognition unit 1 A case in which “Kagawa” is erroneously recognized will be described.

As in the first embodiment, the dialogue state S
_{If the} user inputs “Kanagawa” again as the prefecture name in ₁₀ and the voice recognition unit 1 correctly recognizes “Kanagawa”,
Transition destination dialog state determination unit 4 refers to the table T ₁₀ transition destination dialog state in a dialog state S _10, a transition destination dialog state is determined as S ₃₅ from the speech recognition result "Kanagawa", the dialog management unit 5
Makes the current conversation state transition to the transition destination conversation state S ₃₅ ,
When the user responds "I am Kanagawa" and the user inputs "Yes", the voice recognition unit 1 outputs the voice recognition result "Yes".

Based on the speech recognition result "Yes" for the acknowledgment "Kanagawa is", the dialogue management unit 5 determines that the recognition result "Kanagawa" is a correct recognition result, and determines that correct recognition has occurred. The number of correct recognitions output to the storage unit 3 and stored in the number of correctness / unreliability counts of speech recognition 3 is updated to “8”. At this point, the number of erroneous recognitions is “3”.

The assumed speech recognition rate test unit 6 tests the hypotheses 90% and 80% at a risk rate of 10% for a total recognition count of 11 and a correct recognition count of 8 times. For 90%, u = 1.910> u ₀ = 1.282 and the hypothesis is rejected. For 80%, u = 0.6 <u ₀
= 1.282 and the hypothesis is not rejected. Therefore,
The assumed speech recognition rate test unit 6 outputs 80% as a test result.

The destination dialog state determination unit 4 refers to the table T ₃₅ of the destination dialog state in the dialog state S ₃₅ shown in the lower part of FIG.
There the output to the speech recognition result "Yes", from the determined assumed recognition rate of 80%, and outputs the transition destination dialog state to determine the S _121.

[0062] dialogue management unit 5, the current dialog state to transition from S ₃₅ to S _121, with reference to the dialogue procedure in the dialog state S ₁₂₁ shown in the lower part of FIG. 7, "city or to the user Please enter a county name. " On the other hand, the user inputs, for example, “Kamakura” and continues the dialogue.

By the above operation, the dialogue procedure is changed based on the test result of the assumed speech recognition based on the number of correct / incorrect speech recognitions of the user. increased to be selected interactive procedure such as a dialog state S ₁₂₀ the interaction number is reduced, the relative the assumed recognition rate is poor user interaction number erroneously that becomes more reduced the recognition target vocabulary recognition A dialog procedure such as the dialog state _S121 to be reduced can be selected. Therefore, since an optimal dialog procedure according to the user's voice recognition rate can be selected, the purpose of the dialog can be achieved most efficiently according to the user.

Embodiment 3 Embodiment 3 A voice interactive device according to Embodiment 3 of the present invention will be described with reference to the drawings. FIG. 9 is a diagram showing a configuration of a voice interaction device according to Embodiment 3 of the present invention.

In FIG. 9, 1 is a voice recognition unit, 2 is a dialogue procedure storage unit, 3 is a voice recognition correct / incorrect number-of-times storage unit, 4 is a transition destination dialogue state determination unit, and 5 is a dialogue management unit.

Next, the operation of the voice interaction apparatus according to the third embodiment will be described with reference to the drawings.

The operation of the dialog management unit 5 will be described. Note that the operations of the voice recognition unit 1, the dialogue procedure storage unit 2, the voice recognition correct / incorrect number-of-times storage unit 3, and the transition destination dialogue state determination unit 4 are the same as those in the first embodiment, and a description thereof will be omitted.

When the number of correct / incorrect voice recognitions held in the voice recognition correct / incorrect number storage unit 3 is 10 times of correct answer recognition and 7 times of false recognition, the user enters the dialogue state S shown in the upper part of FIG.
_{When the} user arrives at ₁₀ and inputs "Kanagawa" for "Please enter the prefecture name" as in the first embodiment, the operation performed when the voice recognition unit 1 misrecognizes "Kagawa" is explain.

[0069] transition destination dialog state determination unit 4 refers to the table T ₁₀ transition destination dialog state, the transition destination dialog state and outputs the determined S ₅₃ from the speech recognition result "Kagawa", the dialog management unit 5
But when to transition the conversation state to the S ₅₃ to respond "It is Kagawa", the user inputs "No".

The dialog management section 5 outputs that erroneous recognition has occurred, and the number of erroneous recognitions held in the voice recognition correct / incorrect number storage section 3 is updated to “8”.

[0071] transition destination dialog state determination unit 4 refers to the table T ₅₃ transition destination dialog state shown in the upper part of FIG. 3, erroneously held speech recognition result "No" to the speech recognition errata count storage unit 3 Based on the number of times of recognition “8”, the transition destination conversation state is determined as S _end which is the end conversation state, and is output.

The dialogue management unit 5 is provided with a transition destination dialogue state determination unit 4
When the dialogue state _Send is input from, it is determined whether or not the user has been guided to the telephone number. If the telephone number has not been guided, the dialogue with the device is terminated and the dialogue is switched to the operator.

Whether the telephone number has been guided or not is determined by, for example, assigning “0” as an initial value to the dialogue management unit 5 and setting a counter that changes the value to “1” when a guidance response is executed by 1 And the counter may be checked.

With the above operation, for a user who has a low recognition rate and is unlikely to achieve the dialogue purpose, the dialogue can be switched to the operator, and the user can efficiently achieve the dialogue purpose.

Embodiment 4 Embodiment 4 A speech dialogue apparatus according to Embodiment 4 of the present invention will be described with reference to the drawings. FIG. 10 is a diagram showing a configuration of a voice interaction device according to Embodiment 4 of the present invention.

In FIG. 10, 1 is a speech recognition unit, 2 is a dialogue procedure storage unit, 3 is a speech recognition correct / incorrect number storage unit, 4 is a transition destination dialogue state determination unit, 5 is a dialogue management unit, and 6 is an assumed speech recognition rate. It is a test section.

Next, the operation of the voice interaction apparatus according to the fourth embodiment will be described with reference to the drawings. FIG.
FIG. 1 is a diagram showing an example of a dialogue procedure of a voice interactive device according to Embodiment 4 of the present invention.

The operation of the dialog procedure storage unit 2 and the transition destination dialog state determination unit 4 will be described. The operations of the speech recognition unit 1, the speech recognition correct / incorrect number storage unit 3, the dialogue management unit 5, and the assumed speech recognition rate test unit 6 are the same as those in the second embodiment, and therefore will not be described.

For example, the dialogue state S ₁₀ shown in the upper part of FIG.
, A table T ₁₀ of transition destination conversation states according to all prefecture names, speech recognition results, and assumed recognition rates as speech recognition target vocabulary V ₁₀ , and an average number of conversations up to the end dialog state for each assumed speech recognition rate table N ₁₀ is defined.

The average number of conversations up to the end conversation state in the conversation state S ₁₀ is, for example, assuming that the assumed speech recognition rate is constant and no erroneous recognition occurs, and the conversation state S ₁₀
The average value of the number of state transitions from to all reachable end dialogue states is used approximately.

[0081] illustrating the operation when the user right or wrong number is the number of times shown in FIG. 4 of the speech recognition held in the speech recognition errata count storage unit 3 has reached the dialog state S _10.

The user inputs "Kanagawa" in the response "Enter the prefecture name" of the dialog management unit 5, and the user inputs "Yes" in the response "Kanagawa is" of the dialog management unit 5. The operations up to this point are the same as in the second embodiment. The assumed speech recognition rate test unit 6 operates in the same manner as in the second embodiment, and outputs 90% and 80% as test results.

The transition destination conversation state determination unit 4 refers to the table N ₃₅ of the average number of conversations for each assumed speech recognition in S ₃₅ shown in the lower part of FIG. From the speech recognition rates of 90% and 80%, 90% with the smallest average number of conversations is selected, and the transition destination conversation state is set to S.
Determined as ₁₂₀ and output.

According to the above operation, in addition to the assumed speech recognition rate for the user, the conversation procedure is changed using the average number of conversations according to the assumed speech recognition rate, so that the user can achieve the conversation purpose most efficiently. Can be.

Embodiment 5 FIG. Embodiment 5 A speech dialogue apparatus according to Embodiment 5 of the present invention will be described with reference to the drawings. FIG. 12 is a diagram showing a configuration of a voice interaction device according to Embodiment 5 of the present invention.

In FIG. 12, 1 is a voice recognition unit, 2 is a dialogue procedure storage unit, 3 is a voice recognition correct / incorrect count storage unit, 4 is a transition destination dialogue state determination unit, 5 is a dialogue management unit, and 7 is a voice recognition rate estimation. A unit 8 is a speech recognition success possibility determining unit.

Next, the operation of the voice interaction apparatus according to the fifth embodiment will be described with reference to the drawings. FIG.
FIG. 3 is a diagram showing an example of a dialogue procedure of the voice interactive device according to Embodiment 5 of the present invention.

The operation of the dialogue procedure storage unit 2, the dialogue management unit 5, the speech recognition rate estimation unit 7, and the speech recognition success possibility determination unit 8 will be described. Note that the operations of the voice recognition unit 1, the voice recognition correct / incorrect number storage unit 3, and the transition destination dialog state determination unit 4 are the same as those in the first embodiment, and thus description thereof is omitted.

[0089] For example, in an interactive state S ₁₀ shown in FIG. 13, all of the prefecture name of Japan as the voice recognition target vocabulary V _10,
Table T ₁₀ transition destination dialog state in accordance with the speech recognition result and misrecognition number, as a distribution of the speech recognition accuracy for speech recognition target vocabulary V _10, average value 85, normal distribution D of the dispersion 10
₁₀ : N (85, ₁₀ ) is specified.

[0090] illustrating the operation when the user right or wrong number is the number of times shown in FIG. 4 of the speech recognition held in the speech recognition errata count storage unit 3 has reached the dialog state S _10.

The speech recognition rate estimating section 7 refers to the speech recognition correct / incorrect number storage section 3 and, based on the number of correct recognitions “7” and the number of incorrect recognitions “2”, estimates the user using, for example, the maximum likelihood estimation method. The recognition rate Ru = 7 / _9.times.100 = 78% is calculated and output.

The speech recognition success possibility judging section 8 outputs the estimated user recognition rate R _u = 78% output from the speech recognition rate estimating section 7.
If, it determines the distribution of the speech recognition rate which is defined in dialog state S _10, whether the user is included in a predetermined reference or more portions of the speech recognition rate distribution.

For example, if the reference is 50%, the recognition rate section including 50% of the normal distribution N (85, 10) is R _L = 7.
8.2 ≦ a R ≦ 91.8, estimated recognition rate of the user R _u
Is not more than the lower limit _RL of the section. Therefore, the speech recognition success possibility determination unit 8 determines that the user has no possibility of successful speech recognition.

Since the result of the speech recognition success possibility judging section 8 is that there is no possibility of speech recognition, the dialog management section 5 terminates the dialog with the user and switches to the operator.

By the above operation, the dialogue procedure is changed based on the user's voice recognition possibility determined by the voice recognition success possibility determination section 8, so that the user who has a low possibility of successful voice recognition can communicate with the device. Switching to the operator is performed without performing useless conversation, and the user can efficiently achieve the purpose of the conversation.

Embodiment 6 FIG. Embodiment 6 A speech dialogue apparatus according to Embodiment 6 of the present invention will be described with reference to the drawings. FIG. 14 is a diagram showing a configuration of a voice interaction device according to Embodiment 6 of the present invention.

In FIG. 14, 1 is a voice recognition unit, 2 is a dialogue procedure storage unit, 3 is a voice recognition correct / incorrect number storage unit, 4 is a transition destination dialogue state determination unit, 5 is a dialogue management unit, and 7 is a voice recognition rate estimation. And 8, a speech recognition success possibility determining unit, 9 a speech recognition rate correct / error history accumulation unit, and 10 a speech recognition rate distribution updating unit.

Next, the operation of the voice interaction apparatus according to the sixth embodiment will be described with reference to the drawings.

The operation of the speech recognition rate correct / incorrect history storage section 9 and the speech recognition rate distribution updating section 10 will be described. Note that the speech recognition unit 1, the dialogue procedure storage unit 2, the speech recognition correct / incorrect count storage unit 3, the transition destination dialogue state determination unit 4, the dialogue management unit 5, the speech recognition rate estimation unit 7, and the speech recognition success possibility determination unit 8 The operation is
The description is omitted because it is the same as in the fifth embodiment.

The dialog procedure stored in the dialog procedure storage unit 2 is shown in FIG. 13, and the number of correct / errors of speech recognition stored in the voice recognition correct / error count storage unit 3 is eight times of correct answer and two times of false recognition. for times, explaining the operation when the user has reached the dialog state S _10.

The speech recognition rate estimating section 7 is the same as in the fifth embodiment.
User's estimated speech recognition rate R _u= 80%
Output.

The speech recognition correct / error history accumulation unit 9 stores the speech recognition rate.
The estimated speech recognition rate R of the user output by the estimation unit 7_uTo
And the current conversation state S_TenIs obtained from the dialog management unit 5 and FIG.
Dialogue state S shown in 5_TenCreates a speech recognition error / correction history table for
To achieve. Note that the conversation state S _TenTable exists for
If this is the case, add it to the end of the table and accumulate it.

The speech recognition success possibility judging section 8 operates in the same manner as in the fifth embodiment, and distributes the speech recognition rate distribution N (85, 1).
At 0), it is determined that the user has a possibility of successful voice recognition.

The user inputs "Kanagawa" in the response "Please enter the prefecture name" of the dialog management unit 5, and the user inputs "Yes" in the response "Kanagawa is" of the dialog management unit 5. The operations up to this point are the same as in the fifth embodiment.

Based on the speech recognition result “Yes” for the acknowledgment “Kanagawa is”, the dialog management unit 5 determines that the recognition result “Kanagawa” is a correct recognition result, and determines that correct recognition has occurred. Output to the storage unit 3 and also to the speech recognition correct / error history accumulation unit 9.

[0106] Voice recognition errata history storage unit 9, a correct recognition determination output from the dialogue management unit 5, the speech recognition errata history table for dialog state S ₁₀ shown in FIG. 15, the speech recognition of the estimated speech recognition rate of 80% It is recorded in the correct / incorrect column as shown in FIG.

By continuing the dialogue, a speech recognition correct / error history table for each dialogue state is created, and each time a dialogue with a plurality of users is performed, the voice recognition corrective / error history storage unit 9 stores each dialogue state. , And the correctness of the speech recognition in the dialogue state is accumulated.

The speech recognition rate distribution updating unit 10 uses the speech recognition correct / error history table for each conversation state stored in the speech recognition correct / error history storage unit 9 to perform speech recognition in each dialog state held by the dialog procedure storage unit 2. Update the rate distribution.

[0109] For example, if the speech recognition errata history table dialog state S ₁₀ stored in the speech recognition errata history storage unit 9, is an extract only the audio recognition rate for correct recognition is shown as in FIG. 17, for example Average 8 using maximum likelihood estimation
2.63 and variance 14.25 are obtained as estimates.

[0110] The speech recognition rate distribution updating unit 10 sets the dialogue state S
₁₀ is N (82.63, 14.
Update to 25).

With the above operation, the speech recognition correct / error history table including the estimated speech recognition rate and the voice recognition correct / false judgment is stored in the voice recognition correct / error history storage unit 9, and the recognition target in each dialog state is stored from the stored voice recognition correct / error history table. Since the distribution of the speech recognition rate for the vocabulary can be learned, the accuracy of the speech recognition possibility determination is improved, and the user can efficiently achieve the purpose of the conversation.

[0112]

As described above, the speech dialogue apparatus according to the first aspect of the present invention includes a speech recognition unit that performs a recognition process on an input speech and outputs a speech recognition result, and a speech recognition unit in each dialogue state. A dialogue procedure storage unit that holds a dialogue procedure that defines a transition destination dialogue state according to a target vocabulary, a speech recognition result, and the number of erroneous recognitions; a speech recognition true / false count storage unit that holds the number of times of speech recognition errors; Based on the number of correct / incorrect voice recognitions held in the correct / incorrect number storage unit and the speech recognition result output by the voice recognition unit, a transition destination dialog state is determined with reference to the dialog procedure stored in the dialog procedure storage unit. A transition destination dialog state determining unit that outputs a correct / incorrect result for the speech recognition result output by the speech recognition unit, and transitions the dialog state to a transition destination dialog state output by the transition destination dialog state determination unit. Because a talk management unit, an effect that can determine the interaction steps to achieve the most efficient interaction object according to the user.

As described above, the speech dialogue apparatus according to claim 2 of the present invention performs a recognition process on an input speech and outputs a speech recognition result, and a speech recognition target vocabulary in each dialogue state. A dialogue procedure storage unit that holds a dialogue procedure that defines a transition destination dialogue state according to a speech recognition result and an assumed recognition rate, a speech recognition correct / error count storage unit that holds the number of correctness / errors of speech recognition, Based on the number of correct and incorrect voice recognitions held in the storage unit, a test is performed on the assumed recognition rate specified in the current conversation state,
An assumed speech recognition rate test unit that outputs all assumed recognition rates that are not rejected; and a speech recognition result output by the speech recognition unit and the assumed speech recognition rate test with reference to the dialogue procedure held in the dialogue procedure storage unit. A transition destination dialog state determining unit that determines and outputs one transition destination dialog state from a transition destination dialog state corresponding to an assumed recognition rate output by the unit, and a correct / incorrect result for a speech recognition result output by the voice recognition unit And a dialogue management unit that transitions the dialogue state to the transitional dialogue state output by the transitional dialogue state determination unit, so that a dialogue procedure for achieving the dialogue purpose most efficiently according to the user is provided. Is determined.

As described above, in the voice interaction apparatus according to a third aspect of the present invention, the dialog management unit determines that the transition destination dialog state output by the transition destination dialog state determination unit is a dialog end state, and If the user's dialogue purpose is not achieved, the dialogue with the user is terminated and the operator is switched to the operator, so that there is an effect that the dialogue procedure for achieving the dialogue purpose most efficiently can be determined according to the user.

As described above, in the voice interaction apparatus according to a fourth aspect of the present invention, the interaction procedure storage unit holds an interaction procedure defining an average number of interactions up to an end interaction state in each interaction state. The transition destination dialog state determination unit
With reference to the interaction procedure held in the interaction procedure storage unit,
From the speech recognition result output by the speech recognition unit and the transition destination conversation state corresponding to the assumed recognition rate output by the assumed speech recognition rate test unit, the transition destination dialog state is determined based on the average number of conversations until the end conversation state. Since the information is determined and output as one, it is possible to determine an interactive procedure for achieving the interactive purpose most efficiently according to the user.

As described above, in the speech dialogue apparatus according to claim 5 of the present invention, the dialogue procedure storage unit holds a dialogue procedure defining a speech recognition rate distribution in each dialogue state, and A voice recognition rate estimating unit that estimates and outputs a user's voice recognition rate up to the current dialogue state using the number of times of voice recognition correct / error stored in the storage unit; and a voice recognition output by the voice recognition rate estimating unit. And a speech recognition success possibility determining unit that determines a possibility that the input of the user is correctly recognized based on the speech recognition rate distribution in the current dialogue state and outputs a determination result. The management unit terminates the dialogue with the user based on the determination result of the speech recognition success possibility determination unit and switches to the operator, so that the dialogue method for achieving the dialogue purpose most efficiently according to the user is performed. There is an effect that can be determined.

As described above, the speech dialogue apparatus according to claim 6 of the present invention provides, in each dialogue state, the estimated speech recognition rate of the user up to the dialogue state and the history of correctness of the speech recognition result in the dialogue state. With reference to the voice recognition correct / error history storage unit that stores
A speech recognition rate distribution updating unit that calculates a speech recognition rate distribution in each dialogue state and updates the speech recognition rate distribution held in the dialogue procedure storage unit; This has the effect that a dialog procedure for achieving the purpose can be determined.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of a voice interaction device according to Embodiment 1 of the present invention.

FIG. 2 is a diagram showing an example of a dialogue procedure of the voice dialogue device according to Embodiment 1 of the present invention.

FIG. 3 is a diagram showing an example of a dialogue procedure of the voice dialogue device according to the first embodiment of the present invention.

FIG. 4 is a diagram showing storage contents of a voice recognition correct / incorrect number-of-times storage unit of the voice interactive device according to Embodiment 1 of the present invention;

FIG. 5 is a diagram showing a configuration of a voice interaction device according to a second embodiment of the present invention.

FIG. 6 is a diagram showing an example of a dialogue procedure of the voice dialogue device according to Embodiment 2 of the present invention.

FIG. 7 is a diagram showing an example of a dialogue procedure of the voice dialogue device according to Embodiment 2 of the present invention.

FIG. 8 is a diagram showing an example of a test expression of the voice interaction device according to the second embodiment of the present invention.

FIG. 9 is a diagram showing a configuration of a voice interaction device according to a third embodiment of the present invention.

FIG. 10 is a diagram showing a configuration of a voice interaction device according to a fourth embodiment of the present invention.

FIG. 11 is a diagram showing an example of a dialogue procedure of the voice dialogue device according to Embodiment 4 of the present invention.

FIG. 12 is a diagram showing a configuration of a voice interaction device according to a fifth embodiment of the present invention.

FIG. 13 is a diagram showing an example of a dialogue procedure of the voice dialogue device according to Embodiment 5 of the present invention.

FIG. 14 is a diagram showing a configuration of a voice interaction device according to a sixth embodiment of the present invention.

FIG. 15 is a diagram showing a speech recognition correct / incorrect history table of the speech interactive device according to Embodiment 6 of the present invention;

FIG. 16 is a diagram showing a speech recognition correct / incorrect history table of the speech interaction apparatus according to Embodiment 6 of the present invention.

FIG. 17 is a diagram showing a speech recognition rate for correct answer recognition of the speech interaction apparatus according to Embodiment 6 of the present invention.

FIG. 18 is a diagram illustrating a configuration of a conventional voice interaction device.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 Speech recognition part, 2 dialogue procedure storage part, 3 speech recognition correct / incorrect number storage part, 4 destination dialog state determination part, 5 dialogue management part, 6 assumed speech recognition rate test part, 7 speech recognition rate estimation part, 8 speech recognition A success possibility determination unit, 9 a speech recognition rate correct / error history storage unit, and 10 a speech recognition rate distribution updating unit.

 ────────────────────────────────────────────────── ─── Continued on the front page (72) Inventor Yasushi Ishikawa 2-3-2 Marunouchi, Chiyoda-ku, Tokyo F-term (reference) in Mitsubishi Electric Corporation 5D015 LL00 LL10

Claims (6)

[Claims] 1. A speech recognition unit that performs a recognition process on an input speech and outputs a speech recognition result, and a speech recognition target vocabulary, a speech recognition result, and a transition destination conversation state according to the number of times of erroneous recognition in each conversation state. A dialogue procedure storage unit for holding a specified dialogue procedure; a speech recognition error / correction count storage unit for holding the number of times of speech recognition error / error; a speech recognition error / error count held in the speech recognition error / error count storage unit; and the speech recognition unit A destination dialog state determining unit that determines and outputs a transition destination dialog state by referring to the dialog procedure stored in the dialog procedure storage unit based on the speech recognition result output by the voice recognition unit; A speech dialogue device, comprising: a dialogue management unit that outputs a correct / false result with respect to a speech recognition result and transitions a dialogue state to a transitional dialogue state output by the transitional dialogue state determination unit. 2. A speech recognition unit that performs recognition processing on an input speech and outputs a speech recognition result, and a speech recognition target vocabulary, a speech recognition result, and a transition destination conversation state corresponding to an assumed recognition rate in each conversation state. A dialogue procedure storage unit for holding a specified dialogue procedure, a speech recognition accuracy / error count number storage unit for holding the number of times of speech recognition error, and a current A test is performed on the assumed recognition rate defined in the dialog state, and an assumed speech recognition rate test unit that outputs all the assumed recognition rates that are not rejected, with reference to the dialog procedure stored in the dialog procedure storage unit,
A transition destination dialog state that determines and outputs one transition destination conversation state from the transition destination conversation state corresponding to the speech recognition result output by the speech recognition unit and the assumed recognition rate output by the assumed speech recognition rate test unit. A determining unit, and a dialog managing unit that outputs a correct / incorrect result with respect to the voice recognition result output by the voice recognizing unit, and transitions a dialog state to a transition destination dialog state output by the transition destination dialog state determining unit. A spoken dialogue device. 3. The dialog management unit, if the transition destination dialog state output by the transition destination dialog state determination unit is a dialog end state and the user's dialog purpose is not achieved, 3. The voice dialogue device according to claim 1, wherein the dialogue is terminated and switched to an operator. 4. The dialogue procedure storage unit holds a dialogue procedure defining an average number of dialogues up to an end dialogue state in each dialogue state, and the transition destination dialogue state determination unit is held in the dialogue procedure storage unit. Referring to the dialogue procedure described above, the average number of dialogues from the speech recognition result output by the voice recognition unit and the transition destination dialogue state corresponding to the assumed recognition rate output by the assumed speech recognition rate test unit to the end dialogue state The destination dialog state based on the
4. The voice interaction device according to claim 1, wherein the voice interaction device is determined and output. 5. The dialogue procedure storage unit holds a dialogue procedure defining a speech recognition rate distribution in each dialogue state, and uses a speech recognition correct / error count stored in the speech recognition correct / false count storage unit to store a current speech recognition correct / error count. A voice recognition rate estimating unit that estimates and outputs the user's voice recognition rate up to the dialogue state, and a voice recognition rate output by the voice recognition rate estimating unit, based on the voice recognition rate distribution in the current dialogue state, A voice recognition success possibility determining unit that determines a possibility that the input of the user is correctly recognized and outputs a determination result, wherein the dialogue management unit includes a determination result of the voice recognition success possibility determining unit. 5. The system according to claim 1, wherein the dialogue with the user is switched to an operator based on the termination.
The voice interaction device according to any one of the above. 6. A speech recognition correct / error history storage unit for storing an estimated speech recognition rate of a user up to the dialogue state and a history of correctness / error of speech recognition results in the dialogue state in each dialogue state, A speech recognition rate distribution updating unit that calculates a speech recognition rate distribution in each dialogue state with reference to the storage unit, and updates the speech recognition rate distribution held in the dialogue procedure storage unit. The voice interaction device according to claim 5, wherein