JP2009198614A - Interaction device and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To properly correspond to an optional input for continuing interaction. <P>SOLUTION: A no-input time counter 22 measures a no-input time. When the measured no-input time is a predetermined time or more, an interaction type determination part 34 determines whether an interaction type is an information providing type or an information acquiring type based on the probability of the information providing type obtained from a determination result in the past and the transition probability and the probability of the information acquiring type. When a voice signal fetching part 20 fetches a voice signal, the interaction type determination part 34 determines a query answering type when it is determined that an input voice represents a request/query. If it is determined that the input voice does not represent a request/query, either of an information receiving type and the information acquiring type is determined based on the probability of the information acquiring type and that of the information receiving type. A correspondence generation part 38 generates a correspondence sentence complying with the determination result to utterance from a user. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a dialog device and a program, and more particularly to a dialog device and a program for dialog in response to a user's utterance or input sentence.

2. Description of the Related Art Conventionally, there has been known a voice interaction device that can accurately grasp a user's request contents and smoothly exchange with the user without restricting the user's voice (Patent Document 1). ). This voice interaction device mixes a system-driven dialogue system and a user-driven dialogue system, and appropriately switches between the two types of dialogue systems according to recognition results and dialogue history. It has the flexibility of user-driven dialogue while maintaining robustness.
JP 2003-228393 A

  However, the technique described in Patent Document 1 has a problem that it cannot respond appropriately to an unexpected input. For example, by using the reliability output together with the recognition result of speech recognition, when it seems to be an unexpected input, it can respond with a fixed phrase such as “Please say again”, but the user is a system Since it is determined that the dialogue cannot be performed and the dialogue is interrupted, the dialogue cannot be continued.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a dialog device and a program that can appropriately respond to an arbitrary input and can continue the dialog. To do.

  In order to achieve the above object, an interactive apparatus according to a first aspect of the present invention measures an input means for inputting at least one of an utterance and an input sentence by a user and a time during which there is no input to the input means. Non-input time measurement means, request question determination means for determining whether or not at least one of the utterance and the input sentence input to the input means represents the user's request or question, and the non-input time measurement When the time measured by the means is a predetermined time or more, it is determined that the interaction with the user is an information providing type in which the device itself provides information voluntarily, and the request question determining means determines the user If it is determined that the user's request or question is expressed, it is determined that the dialogue with the user is a question answering type in which the device responds to the user's request or question, When it is determined that the user's request or question is not represented by the question determination unit, the dialogue with the user is an information reception type in which the device itself receives information that the user voluntarily provides. An interactive discriminating unit for discriminating; a response generating unit for generating a response sentence corresponding to a discrimination result by the interactive discriminating unit for at least one of the utterance and the input sentence; and a response generated by the response generating unit Output means for outputting a sentence.

  According to a second aspect of the present invention, there is provided a non-input time measuring unit that measures a time during which a computer is not input to an input unit that inputs at least one of an utterance and an input sentence by a user. When the time measured by the requested question determination means and the non-input time measurement means for determining whether at least one of the uttered speech and the input sentence represents the user's request or question is a predetermined time or more In the case where it is determined that the dialogue with the user is an information providing type in which the device itself provides information voluntarily, and it is determined by the request question determination means that it represents the user's request or question Determining that the interaction with the user is a question answering type in which the device responds to the user's request or question, and the request question determining means If it is determined that it does not represent the user's request or question, the interactive determination for determining that the dialogue with the user is an information acceptance type in which the device itself accepts information that the user voluntarily provides And a program for functioning as at least one of the utterance and the input sentence as a response generation unit that generates a response sentence according to the determination result by the interactive determination unit.

  According to the first and second inventions, at least one of the user's utterance and the input sentence is input to the input means. Further, the non-input time measuring means measures the time during which the state where there is no input to the input means continues.

  Further, the request question determination unit determines whether at least one of the utterance and the input sentence input to the input unit represents a user request or a question.

  Then, when the time measured by the no-input time measuring means is equal to or longer than the predetermined time by the interactive discrimination means, the dialogue with the user is an information providing type in which the device itself provides information voluntarily. Determine. Further, when it is determined by the request question determination means that it represents a user request or question, it is determined that the dialogue with the user is a question response type in which the own device answers the user request or question. To do. Further, when it is determined by the request question determination means that it does not represent the user's request or question, the interaction with the user is an information reception type in which the device itself receives information that the user voluntarily provides Determine.

  Then, the response generation means generates a response sentence corresponding to the determination result by the interactive determination means for at least one of the utterance and the input sentence. The response means generated by the response generation means is output by the output means.

  Thus, based on whether there is no input time and whether the input utterance or input sentence represents a request or a question, the interactive classification of the dialog with the user is an information providing type, a question answering type, and an information type. It is possible to appropriately respond to an arbitrary input and continue the dialogue by determining which type is an acceptance type and generating a response sentence.

  An interactive apparatus according to a third aspect of the present invention is an input unit that inputs at least one of a user's utterance and an input sentence, a non-input time measuring unit that measures a time during which there is no input to the input unit, The request question determination means for determining whether or not at least one of the utterance and the input sentence input to the input means represents the user's request or question, and the time measured by the non-input time measurement means is predetermined. If it is more than the time, it is determined that the dialogue with the user is an information acquisition type in which the device voluntarily asks a question, and it is determined that the request question determination means represents the user's request or question. If it is determined that the dialogue with the user is a question answering type in which the user's own device answers the user's request or question, the request question determining means determines the user. If it is determined that the request or question is not represented, the dialog with the user is determined to be an information reception type in which the user apparatus accepts information that the user voluntarily provides; Response generation means for generating a response sentence corresponding to the determination result by the interactive determination means for at least one of the utterance and the input sentence; and output means for outputting the response sentence generated by the response generation means; It is comprised including.

  According to a fourth aspect of the present invention, there is provided a non-input time measuring unit that measures a time during which a computer is not input to an input unit that inputs at least one of an utterance and an input sentence by a user. When the time measured by the requested question determination means and the non-input time measurement means for determining whether at least one of the uttered speech and the input sentence represents the user's request or question is a predetermined time or more In the case where it is determined that the dialogue with the user is an information acquisition type in which the device itself asks a question voluntarily, and the request question determination unit determines that the request or the question represents the user The dialogue with the user is determined to be a question response type in which the own device answers the user's request or question, and the user is asked by the request question determination means. If it is determined that it does not represent a request or a question, the interactive type discriminating means for discriminating that the dialogue with the user is an information accepting type in which the device itself accepts information that the user voluntarily provides; and It is a program for functioning as a response generation unit that generates a response sentence corresponding to a determination result by the interactive determination unit for at least one of the utterance and the input sentence.

  According to the third and fourth inventions, at least one of a user's utterance and an input sentence is input to the input means. Further, the non-input time measuring means measures the time during which the state where there is no input to the input means continues.

  Further, the request question determination unit determines whether at least one of the utterance and the input sentence input to the input unit represents a user request or a question.

  Then, when the time measured by the no-input time measuring means is equal to or longer than the predetermined time by the interactive discrimination means, it is determined that the dialogue with the user is an information acquisition type in which the device spontaneously asks a question. . Further, when it is determined by the request question determination means that it represents a user request or question, it is determined that the dialogue with the user is a question response type in which the own device answers the user request or question. To do. Further, when it is determined by the request question determination means that it does not represent the user's request or question, the interaction with the user is an information reception type in which the device itself receives information that the user voluntarily provides Determine.

  Then, the response generation means generates a response sentence corresponding to the determination result by the interactive determination means for at least one of the utterance and the input sentence. The response means generated by the response generation means is output by the output means.

  Thus, based on whether there is no input time and whether the input utterance or input sentence represents a request or question, the interactive classification of user interaction is classified as an information acquisition type, a question response type, and an information type. It is possible to appropriately respond to an arbitrary input and continue the dialogue by determining which type is an acceptance type and generating a response sentence.

  An interactive apparatus according to a fifth aspect of the present invention is an input unit that inputs at least one of an utterance and an input sentence by a user, a non-input time measuring unit that measures a time during which there is no input to the input unit, In the dialogue between the user and his / her own device, the own device spontaneously makes a decision as to whether or not at least one of the utterance and the input sentence input to the input device represents the user's request or question. Information provision type that provides information automatically, information acquisition type that self device voluntarily asks, question response type that self device answers user's request or question, and information that user provides voluntarily A transition history storage means for storing a transition history of an interactive classification consisting of an information acceptance type accepted by the apparatus, and a past time when the time measured by the no-input time measurement means is a predetermined time or more, Based on the determination result and the transition history, it is determined that the interaction with the user is either the information provision type or the information acquisition type, and the request question determination unit represents the request or question of the user. If it is determined that the dialogue with the user is the question answering type, and if it is determined by the request question determination means that it does not represent the user's request or question, An interactive type discriminating unit that discriminates that the dialogue with the user is the information acceptance type, and a response that generates a response sentence according to the discrimination result by the interactive type discriminating unit for at least one of the utterance and the input sentence The generating unit includes an output unit that outputs the response sentence generated by the response generating unit.

  According to a sixth aspect of the present invention, there is provided a non-input time measuring unit that measures a time during which a computer is not input to an input unit that inputs at least one of an utterance and an input sentence by a user. Request question determination means for determining whether or not at least one of the uttered speech and input sentence represents the user's request or question, and the device itself provides information spontaneously in the dialogue between the user and the device Information-providing type, information-acquiring type in which the device itself asks questions, question-answering type in which the device responds to user requests or questions, and information that the device accepts information that the user voluntarily provides Transition history storage means for storing the transition history of the interactive type consisting of acceptance type, and when the time measured by the non-input time measuring means is a predetermined time or more, Based on the result and the transition history, it is determined that the interaction with the user is one of the information provision type and the information acquisition type, and the request question determination unit represents the user's request or question. If determined, it is determined that the dialogue with the user is the question answering type, and if it is determined by the request question determination means that it does not represent the user's request or question, the user Generating a response sentence according to a determination result by the interactive determination means for at least one of the utterance and the input sentence, and an interactive determination means for determining that the dialogue with the information acceptance type It is a program for functioning as a means.

  According to the fifth and sixth inventions, at least one of the user's utterance and the input sentence is input to the input means. Further, the non-input time measuring means measures the time during which the state where there is no input to the input means continues.

  Further, the request question determination unit determines whether at least one of the utterance and the input sentence input to the input unit represents a user request or a question.

  Then, when the time measured by the no-input time measuring means is equal to or longer than the predetermined time by the interactive discrimination means, the dialogue with the user is based on the past discrimination result and the transition history. It is determined that it is one of acquisition type. Further, when it is determined by the request question determination means that it represents the user's request or question, it is determined that the dialogue with the user is a question response type. Further, when it is determined by the request question determination means that the request or question from the user is not represented, it is determined that the dialogue with the user is an information acceptance type.

  Then, the response generation means generates a response sentence corresponding to the determination result by the interactive determination means for at least one of the utterance and the input sentence. The response means generated by the response generation means is output by the output means.

  Thus, based on the no-input time, the transition history of the interactive classification, and whether or not the input utterance or input sentence represents a request or question, the interactive classification of the dialog with the user is an information provision type. It is possible to respond appropriately to any input and continue the dialogue by discriminating whether it is information acquisition type, question answering type, or information receiving type and generating a response sentence it can.

  In the interactive discrimination means according to the fifth invention, when it is determined that the request question determination means does not represent the user's request or question, the dialogue with the user is performed based on the past determination result and the transition history. Therefore, it can be determined that the information reception type or the information acquisition type. As a result, when the input utterance or input sentence does not represent a request or a question, it is more appropriate for any input by further determining whether it is an information acceptance type or an information acquisition type. Can respond.

  The response generating means according to the first and fifth inventions, when determined to be of the information providing type, has selected from a plurality of information stored in an information database storing a plurality of information for providing to the user A response sentence can be generated using the information.

  When it is determined that the response generation means is a question response type, the question sentence or request stored in the question answer database in which the question sentence or request sentence and the answer sentence to the question sentence or request sentence are stored in association with each other Searching a question sentence or request sentence corresponding to a request or question represented by at least one of an utterance and an input sentence from a sentence, and generating a response sentence using an answer sentence to the searched question sentence or request sentence it can.

  The corresponding device further includes an analysis unit that analyzes the structure of at least one of the utterance and the input sentence input by the input unit, and the response generation unit determines that the structure and the structure are in a case where the response generation unit is determined to be an information reception type. The response sentence can be generated using the response sentence stored in the response database stored in correspondence with the response sentence and corresponding to the structure analyzed by the analyzing means.

  When it is determined that the response generation means according to the third and fifth inventions is an information acquisition type, a plurality of question sentences stored in a question request database storing a plurality of question sentences or request sentences to the user A response sentence can be generated using the question sentence selected from the above.

  The input means inputs an utterance by the user, and the request question determination means extracts a voice feature amount for the utterance input from the input means, and extracts the extracted feature amount and a previously obtained question or request. It is possible to determine whether or not at least one of the utterance and the input sentence represents the user's request or question by comparing with the feature amount of the expressed utterance.

  As described above, according to the dialogue apparatus and program of the present invention, the interactive type of dialogue with the user is based on the no-input time and whether or not the inputted utterance or input sentence represents a request or a question. By discriminating the classification and generating a response sentence, it is possible to appropriately respond to an arbitrary input and to obtain an effect that the dialogue can be continued.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the present embodiment, a case will be described in which the present invention is applied to an interactive device that responds to a user's utterance using speech.

  As shown in FIG. 1, an interactive apparatus 10 according to the first embodiment is configured by a microphone, and is input by an input unit 12 that collects a user's utterance and generates an audio signal. A computer 14 that generates a response utterance based on the received audio signal, and a voice output unit 16 that includes a speaker and outputs the generated response utterance.

  The computer 14 includes a CPU, a RAM, and a ROM that stores a program for executing a response generation processing routine to be described later, and is functionally configured as follows. The computer 14 continues from the signal input from the input unit 12 to a voice cutout unit 20 that cuts out a voice signal indicating the input voice and a state in which no input voice is input based on the signal input from the input unit 12. A non-input time counter 22 that measures a non-input time, a word recognition unit 24 that recognizes a user utterance based on a voice signal with reference to a word registered in a recognition dictionary database (not shown), and a voice Based on the signal, a feature amount extraction unit 26 that extracts a feature amount (for example, duration or fundamental frequency) of speech, and, as a transition history, transition probabilities of four types of interactive classifications described later in a dialog with the user are obtained. The stored transition probability storage unit 28, the determination result storage unit 30 that stores the result of the previous determination by the interactive classification determination unit 34 described later, the stored transition probability and the previous time Based on the determination result, the probability calculation unit 32 that calculates the probability that the current dialogue is each interactive classification, the no-input time measured by the no-input time counter 22, the recognition result by the voice recognition unit 24, and the feature amount An interactive classification discriminating unit 34 for discriminating a dialogue with the user into one of four types of interactive classifications based on the feature amount extracted by the extracting unit 26 and the probability calculated by the probability calculating unit 32; I have.

  Here, four types of interactive classification will be described. The dialogue with the user is an information providing dialogue in which the dialogue device 10 voluntarily provides information to the user, an information acquisition dialogue in which the dialogue device 10 voluntarily asks the user, a user request or a question On the other hand, the interactive device 10 is classified into four types of interactive classifications: a question-and-answer type dialogue in which the interactive device 10 answers, and an information reception type dialogue in which the interactive device 10 accepts information that the user voluntarily provides.

  The information provision type dialogue is a dialogue on the assumption that the dialogue device 10 has the initiative of the dialogue and the dialogue device 10 provides information, and the state of the dialogue device 10 and the information stored in the dialogue device 10. Is a dialogue that the dialogue device 10 voluntarily provides to the user. For example, when the user approaches the interactive device 10 or when the user is silent in front of the interactive device 10, information held by the interactive device 10 (news, weather information, how to handle the interactive device 10) Information).

  The information acquisition type dialogue is a dialogue on the assumption that the dialogue device 10 has the initiative of the dialogue and the user provides information to the dialogue device 10, and the user's own information and requests, or the user This is a dialogue in which the dialogue device 10 voluntarily asks about information known to the user. For example, a dialogue (such as “What do you want to know?” Or “What did you come to see today?”) For the user to acquire user information necessary for the dialogue device 10 to narrow down information to be provided to the user. Or a dialogue to continue the dialogue.

  The question answering type dialogue is a dialogue on the assumption that the user has the initiative of the dialogue and the dialogue device 10 provides information to the user, and the dialogue device 10 answers the question from the user. It is a dialogue to search and answer. For example, in response to a question or request from the user (such as “Where is the toilet?” Or “What time are you doing?”), Information that is the answer to the question or request is retrieved from the database and provided to the user.

  The information-accepting dialogue is a dialogue on the assumption that the user has the initiative of the dialogue and the user provides information to the dialogue device 10, and the user's own information, information that the user knows, etc. In this dialogue, the dialogue apparatus 10 accepts information that the user voluntarily provides. For example, in response to an input other than a question or a request from the user, for example, a response is given (for example, “I am so happy”) or a question (for example, “play” "Where did you come to?"), Spread the topic (for example, "The weather is nice today" and "The laundry dries well." ), And wait for the input that is not assumed in the above information-providing dialog, information-acquisition-type dialog, and question-and-answer-type dialog as much as possible, and conduct a dialog to continue the dialog .

  The voice recognition unit 24 recognizes the user utterance based on the input voice and outputs the reliability for each recognition result. For example, when the input utterance is “What is today's weather?”, As recognition results, “What is today's weather?”, “Today's weather.”, And “Today's weather is good.” A reliability “0.8” for “What is the weather today?”, A reliability “0.5” for “Today's weather.”, And a reliability “0.3” for “Good weather today” are obtained.

  The feature quantity extraction unit 25 extracts a fundamental frequency as a voice feature quantity from the input voice signal. Based on the fundamental frequency extracted by the feature amount extraction unit 25, it can be determined whether or not the prosody of the speech is raised.

  In the transition probability storage unit 28, as shown in FIG. 2, the transition probabilities of the above four types of interactive classifications are stored as information indicating the transition history. Here, the transition probability a_xy is calculated as follows.

First, in the WOZ (Wizard of OZ) experiment or the like, correct dialogue transition history data as shown below is created.
User: I came to play.
The interactive device: Hello, Who did you come? (Information-accepting dialogue)
User: I came with a friend.
Dialogue device: Yeah. (Information-accepting dialogue)
Dialogue device: XX is on display today. (Information-providing dialogue)
User: What is XX?
Dialogue device: It's the topic of xx now. That famous △△ made it. (Question answering dialogue)
Dialogue device: Are you interested in XX? (Information acquisition dialogue)
From the above dialog transition history data (initial state → information acceptance type dialogue → information acceptance type dialogue → information provision type dialogue → question answer type dialogue → information acquisition type dialogue) A transition probability a_xy for transition from the type classification x to the interactive classification y is calculated.
a_xy =
(Number of times of transition from interactive classification x to interactive classification y) / (Number of times of interactive classification x)
... (1)
However, the information provision type dialogue is assumed to be interactive category 1, the information acquisition type dialogue is assumed to be interactive type 2, the question answering type dialogue is assumed to be interactive type 3, and the information acceptance type dialogue is assumed to be interactive type 4.

  From the plurality of dialog transition history data, the transition probability is calculated for all combinations of transitions of the four types of interactive classification using the above equation (1), and stored in the transition probability storage unit 28.

  Based on the transition probability stored in the transition probability storage unit 28 and the interactive classification as the previous determination result stored in the determination result storage unit 30, the probability calculation unit 32 uses the interactive classification of the previous determination result. The transition probability from 1 to the predetermined interactive classification is calculated as the probability that the interactive classification of the current discrimination target dialog is the predetermined interactive classification.

  When the no-input time measured by the no-input time counter 22 is 5 seconds or more, the interactive classification discriminating unit 34 performs the probability and information acquisition type dialogue that is the information provision type dialogue calculated by the probability calculation unit 32. Based on a certain probability, it is determined that the current interaction with the user is either an information providing interaction or an information acquisition interaction. When the input voice is cut out by the voice cutout unit 20, the utterance from the user is asked or asked based on the result of the voice recognition unit 24 and the voice feature amount extracted by the feature amount extraction unit 26. It is determined whether or not.

  In the determination as to whether or not it is a question or a request, the structure at the end of the speech recognition result with the highest reliability obtained by the speech recognition unit 24 is the structure of the question sentence or the request sentence (for example, “??” If it is the same as “˜tell me” or “want to do”), it is determined that the utterance from the user is a question or request. In addition, the fundamental frequency as the feature amount of the voice is compared with the fundamental frequency at the time of the question or request obtained in advance (the fundamental frequency that becomes the upper right at the end of the speech), and the extracted fundamental frequency and the question Alternatively, if the fundamental frequency obtained in advance is similar to the request, it is determined that the utterance from the user is a question or a request.

  As described above, when it is determined that the utterance from the user is a question or a request, it is determined that the current interaction with the user is a question answering-type interaction, and is determined not to be a question or a request. In this case, based on the probability of the information acquisition type dialogue and the probability of the information reception type dialogue calculated by the probability calculation unit 32, the current user dialogue is either the information acquisition type dialogue or the information acceptance type dialogue. It is determined that

  In addition, the computer 14 of the dialogue apparatus 10 includes a dialogue history storage unit 36 that stores, as a dialogue history, an utterance from the user recognized by the voice recognition unit 24 and a response sentence generated by a response generation unit 38 to be described later. Based on the interactive classification determined by the type classification determination unit 34, the utterance from the user recognized by the voice recognition unit 24, and the conversation history stored in the conversation history storage unit 36, a response sentence to the utterance from the user is obtained. And a response generation unit 38 that generates the response.

  As illustrated in FIG. 3, the response generation unit 38 generates an information provision type response generation that generates a response sentence corresponding to the information provision type dialogue when the interactive type classification determination unit 34 determines that the information provision type dialogue is performed. An information acquisition type response generation unit 44 that generates a response sentence corresponding to the information acquisition type dialogue when it is determined by the unit 42 and the interactive type classification determination unit 34, and an interactive classification determination unit When it is determined by the question answering type dialogue 34 that the question answering type response generating unit 46 generates a response sentence corresponding to the question answering type dialogue and the interactive classification discriminating unit 34 is the information acceptance type dialogue. An information acceptance type response generation unit 48 that generates a response sentence corresponding to the information acceptance type dialogue when it is determined.

  Based on the information database 42A storing a plurality of pieces of information to be provided to the user, the dialogue history stored in the dialogue history storage unit 36, and the information in the information database 42A, the information provision type response generation unit 42 performs an information provision type dialogue. A response sentence generation unit 42B that generates a response sentence corresponding to the response sentence. The response sentence generation unit 42B selects information that has not yet been output from the information database 42A by using the conversation history, and includes words included in the utterance from the previous user obtained from the conversation history from the selected information. The most relevant information is selected, and a response sentence is generated using the selected information.

  For example, information that has not yet been output, and the like "I today are exhibited is ○○○", "weather It would be a good today" and, if the utterance from the previous user is "Hello" Selects “I like the weather today” and generates a response sentence.

  The information acquisition type response generation unit 44 is based on the question database 44A storing a plurality of question sentences to the user, the dialog history stored in the dialog history storage unit 36, and the question text in the question database 44A. And a response sentence generation unit 44B that generates a response sentence corresponding to the above. The response sentence generation unit 44B selects a question sentence that has not yet been output from the question database 44A using the conversation history, and is included in the utterance from the previous user obtained from the conversation history from the selected question sentence. A question sentence most closely related to the word is selected, and a response sentence is generated using the selected question sentence. When there is a question sentence indicating the question that the dialogue apparatus 10 wants to know most in the question database 44A, the question sentence indicating the question that the dialogue apparatus 10 wants to know most is selected, and a response sentence is selected using the selected question sentence. Is generated.

  For example, a question sentence that has not been output yet is “What do you want to know?” Or “Is you interested in XX (for example, the name of an exhibit?)? If it is “Is interested in the name of the exhibit?”, The user selects “Are you interested in the name of the exhibit?” And generates a response sentence.

  The question response type response generation unit 46 includes a question answer database 46A in which a plurality of question sentences or request sentences from the user and answer sentences to the question sentences or request sentences are associated with each other, and a dialogue stored in the dialogue history storage unit 36. A response sentence generation unit 46B that generates a response sentence corresponding to the question-response dialogue based on the history and the answer sentence of the question answer database 46A is provided. The response sentence generation unit 46B uses the dialogue history to select an answer sentence that has not been output from the question answer database 46A, and is obtained from the voice recognition unit 24 from the question sentence corresponding to the selected answer sentence. A question sentence that most closely matches the utterance from the user is selected, and a response sentence is generated using an answer sentence for the selected question sentence.

  For example, the answer sentences that have not yet been output are “I am in the right back of the guide counter” and “I am in the 2F special exhibition room”, and the question sentences “Where is the toilet?” And “○” ○ From "Where is the name of the exhibit?", Select the question sentence "Where is the toilet?" That is most similar to the voice recognition result "Where is the toilet?", And the answer sentence corresponding to the selected question sentence A response sentence is generated using “I'm right behind the guide counter”.

  The information reception type response generation unit 48 is a structure of a response database 48A that stores a plurality of sentence structures and response sentences corresponding to the sentence structure, and a structure analysis unit that analyzes the structure of the utterance from the user obtained from the speech recognition unit 24. 48C, and a response sentence generation unit 48B that generates a response sentence corresponding to the information acceptance type dialogue based on the structure of the utterance from the user analyzed by the structure analysis unit 48C and the response sentence of the response database 48A. Yes.

  The structure analysis unit 48C performs morphological analysis on the utterance from the user obtained from the speech recognition unit 24, and analyzes the structure (part of speech and meaning) of the utterance from the user using a word dictionary. The response sentence generation unit 48B generates a plurality of response sentence candidates for the structure analyzed by the structure analysis unit 48C using the response database 48A, and selects the most natural response sentence candidate using the dialogue history. Then, a response sentence to the user utterance is generated.

  For example, the response sentence for the keyword “XX (person name)” is “Yes”, the response sentence for the keyword “weather” or “good” is “the selection is dry well” If the response to the predicate “Tayo” is “Who are you with?” And the response to the predicate “I saw you” is “When did you see it?”, The user's utterance “Akkun When the name “Aku” and the predicate “I came” are obtained as a structure from “I came to play,” “Who are you?” And “Hey, are n’t you?” Is done. Then, using the conversation history, “Hey is so” is selected as the most natural response sentence, and a response sentence is generated.

  Next, the principle of this embodiment will be described. There are two types of dialogues between the user and the dialogue device, depending on whether the user has the initiative of dialogue or the dialogue device type, the dialogue device-led dialogue and the user-led dialogue. The interactive device initiative type interactive device can show the user a response that the interactive device can respond to, but has a problem that the user cannot directly input what he / she wants to input. Therefore, in the interactive apparatus 10 according to the present embodiment, when the user directly inputs what he / she wants to input, a response sentence corresponding to the question answering type dialogue is generated and responded to the user's utterance.

  In addition, the user-initiated interactive device can wait for a user input and return a response according to the input, but there is a problem that a dialog is not established if there is no user input. Therefore, in this embodiment, a response sentence corresponding to the information provision type dialogue is generated, and the dialogue is established even when there is no user input.

  In both interactive device-driven and user-driven interactive devices, anthropomorphic interactive devices (two-dimensional characters, animal / humanoid robots, etc., which generate human language) ), The user may make an input equivalent to that of a person to the dialog device even though the dialog device cannot accept any input. For example, an in-house guidance system installed in an exhibition hall that accepts voice input (waiting for utterances such as “Where is the toilet?” Or “What time are you waiting?”) (For example, “I'm so happy.”, “I'm coming to play.”, And “I like the weather today.”) There is a problem that the system erroneously responds due to an unexpected input or is discarded (rejected) as an unacceptable utterance. If the system responds incorrectly or returns a fixed phrase at the time of rejection such as “Please say in another way”, the user decides that he / she cannot interact with the dialog device and stops the dialog.

  Since many of the unexpected utterances are often information provision from the user, the dialog device 10 according to the present embodiment responds to the information acquisition type dialog or the information reception type dialog when an arbitrary input is made from the user. The response sentence is generated and the user's utterance is responded.

  As described above, in the present embodiment, for an arbitrary input, an interaction with the user is determined as an interactive classification, and a response is generated according to the determined interactive classification. Can be accepted properly.

  Hereinafter, the response generation processing routine executed by the computer 22 will be described with reference to FIG. First, in step 100, measurement of the no-input time is started, and in the next step 102, it is determined whether or not an audio signal indicating the input audio has been cut out from the signal input from the input unit 12. When the voice signal corresponding to the user's utterance is generated by the input unit 12 and the voice signal indicating the input voice is cut out, in step 104, the input voice signal is referenced with reference to the word registered in the recognition dictionary database. Based on the above, the user utterance is recognized, and the reliability is calculated for each recognized user utterance.

  In step 106, the fundamental frequency of the input audio signal is extracted as the audio feature quantity. In step 108, the user utterance with the highest reliability recognized in step 104 and extracted in step 106. Based on the fundamental frequency of the voice signal, it is determined whether or not the utterance from the user represents a question or a request.

  If it is determined in step 108 that the utterance from the user represents a question or a request, it is determined that the dialog with the user is a question answering type dialog, and in step 110, it is recognized in step 104. Based on the utterance from the user and the question answering database 46A, a response sentence corresponding to the question answering type dialogue is generated for the utterance from the user and output to the voice output unit 16, and a response generation processing routine is executed. finish.

  If it is determined in step 108 that the utterance from the user does not represent a question or a request, in step 112, based on the transition probability stored in the transition probability storage unit 28 and the previous determination result. The probability that the dialogue with the current user is an information acquisition dialogue and the probability that the dialogue with the current user is an information reception dialogue is calculated. In step 114, it is determined whether or not the probability of the information acceptance type dialogue calculated in step 112 is greater than the probability of the information acquisition type dialogue. If the probability of the information acceptance type dialogue is greater than the probability of the information acquisition type dialogue at step 114, it is determined that the dialogue with the user is an information acceptance type dialogue, and at step 116, the recognition at step 104 is recognized. The structure of the utterance from the user is analyzed, and a response sentence corresponding to the information reception type dialogue is generated for the utterance from the user based on the structure of the utterance from the analyzed user and the response database 48A. Then, the generated response sentence is synthesized with speech and output to the speech output unit 16, and the response generation processing routine is terminated.

  If the probability of the information acquisition dialogue is equal to or less than the probability of the information acceptance dialogue in step 114, it is determined that the dialogue with the user is an information acquisition dialogue. In step 118, the stored dialogue is determined. Based on the history and the previous user's utterance, a response sentence corresponding to the information acquisition type dialogue is generated, the generated response sentence is synthesized by voice, and output to the voice output unit 16 to finish the response generation processing routine To do.

  If the audio signal is not cut out from the signal input from the input unit 12 in step 102, it is determined in step 120 whether or not the measured no-input time is 5 seconds or more. If the time during which no voice is input from the user is less than 5 seconds, the process returns to step 102. On the other hand, if it is determined in step 120 that the measured no-input time is 5 seconds or longer, in step 122, based on the transition probability stored in the transition probability storage unit 28 and the previous determination result. Each of the probability that the dialogue with the current user is an information provision type dialogue and the probability that the dialogue with the current user is an information acquisition type dialogue is calculated. Then, in step 124, it is determined whether or not the probability of the information provision type dialogue calculated in step 122 is larger than the probability of the information acquisition type dialogue. If the probability of the information providing dialogue is greater than the probability of the information acquiring dialogue in step 124, it is determined that the dialogue with the user is an information providing dialogue, and the stored dialogue history is stored in step 126. Based on the previous utterance from the user and the information database 42A, a response sentence corresponding to the information providing type dialogue is generated for the utterance from the user, and the generated response sentence is synthesized by speech and output as voice It outputs to the part 16, and a response generation process routine is complete | finished.

  On the other hand, if the probability of the information acquisition type dialogue is greater than or equal to the probability of the information provision type dialogue in step 124, it is determined that the dialogue with the user is the information acquisition type dialogue, and the process proceeds to step 118. Then, a response sentence corresponding to the information acquisition type dialogue is generated, the generated response sentence is synthesized by voice, and output to the voice output unit 16, and the response generation processing routine is terminated.

  Then, the response sentence generated as described above is output as voice by the voice output unit 16.

  As described above, according to the dialogue apparatus according to the first embodiment, based on the no-input time, the transition probability of the interactive classification, and whether the utterance from the user represents a request or a question, It is determined whether the interactive classification of the dialogue with the user is an information providing dialogue, an information acquisition dialogue, a question answering dialogue, or an information acceptance dialogue, and the response sentence is determined according to the identified interactive classification. By generating, it is possible to appropriately respond to any input and continue the dialogue.

  In addition, by switching the response sentence generation method according to the determined interactive classification, it is possible to realize a dialog that accepts any input from the user and does not fail.

  Next, a second embodiment will be described. In addition, about the part which has the structure similar to 1st Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  In the second embodiment, an image is captured in front of the interactive device by the imaging unit, and it is determined whether or not the user existing in front of the interactive device is watching the interactive device. The difference from the first embodiment is that the measurement of the no-input time is started when it is determined that the user is watching.

  As illustrated in FIG. 5, the interactive apparatus 210 according to the second embodiment includes an input unit 12, an imaging unit 212 that captures an image of the front of the device, a signal input by the input unit 12, and the imaging device 212. The computer 214 that generates a response utterance based on the image captured by the computer and the audio output unit 16 are provided.

  The imaging unit 212 is arranged to capture an area including the user's face when a user is present in front of the device. When the user is looking at his / her own device in front of his / her own device, a front face image in which the user's face is facing the front is captured.

  The computer 214 includes a front face detection unit 220 that detects a front face image of the user from images picked up by the sound extraction unit 20 and the image pickup unit 212, and a time when the front face image is detected by the front face detection unit 220. A no-input time counter 222 that measures the no-input time of the input voice, a voice recognition unit 24, a feature amount extraction unit 26, a transition probability storage unit 28, a discrimination result storage unit 30, a probability calculation unit 32, The interactive classification discriminating unit 34, the dialogue history storage unit 36, and the response generation unit 38 are provided.

  The front face detection unit 220 detects the front face image of the user existing in front of the user's own device from the image captured by the imaging unit 212 using the learning pattern of the face image facing the front.

  When the front face image of the user is detected by the front face detection unit 220, the no-input time counter 222 determines that the user is looking at the interactive device 210, and the state where no sound is input from the user continues. Start measuring time.

  In the response generation processing routine according to the second embodiment, it is determined whether or not the front face image of the user is detected from the captured image, and measurement of the no-input time is started when the front face image is detected. To do. Then, it is determined whether or not the audio signal is cut out from the signal input from the input unit 12, and if the audio signal is not cut out from the input signal, the measured no-input time is 5 It is determined whether or not it is equal to or longer than second. If it is determined that the measured no-input time is 5 seconds or longer, it is determined that the voice input is interrupted for a long time even though the user is watching the dialog device 210, and the current dialog Based on the probability that is an information-providing dialogue and the probability that it is an information-acquiring dialogue, it is determined whether the interactive classification of the dialogue with the user is an information-providing dialogue or an information-acquisition dialogue. A response sentence is generated according to the interactive classification.

  On the other hand, when a speech signal is cut out from the input signal, it is determined whether or not the speech from the user represents a question or a request based on the speech recognition result and the fundamental frequency of the speech signal. If it is determined that it represents a question or a request, it is determined that the dialogue with the user is a question-response dialogue, and a response sentence corresponding to the question-response dialogue is generated. Further, when it is determined that the utterance from the user does not represent a question or a request, the interactive classification of the dialog with the user is based on the probability of the information acquisition dialog and the probability of the information acceptance dialog. It is determined whether the dialogue is an information reception type dialogue or an information acquisition type dialogue, and a response sentence is generated according to the discriminated dialogue type.

  Thus, it is possible to accurately measure the time during which there is no input from the user by starting the measurement of the no-input time from when it is determined that the user is looking at the interactive device. In addition, based on the measured no-input time, it is possible to accurately determine that the interaction with the user is an information providing interaction.

  In the above embodiment, the case where the response to the utterance from the user is performed by the sound output by the speaker is described as an example, but the present invention is not limited to this, and the response sentence is displayed on the display. Also good.

  Next, a third embodiment will be described. In addition, about the part which has the structure similar to 1st Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  The third embodiment is mainly different from the first embodiment in that text data is input from the user.

  As shown in FIG. 6, the dialogue apparatus 310 according to the third embodiment is configured with a keyboard, and an input unit 312 that inputs an input sentence from the user as text data, and is input by the input unit 312. A computer 314 that generates a response sentence based on the text data, and a display unit 316 that includes a display and displays the generated response sentence are provided.

  The computer 314 includes a no-input time counter 322 that measures a no-input time based on an input from the input unit 312, a transition probability storage unit 28, a discrimination result storage unit 30, a probability calculation unit 32, and a no-input time. Based on the no-input time measured by the counter 322, the text data input from the input unit 312, and the probability calculated by the probability calculation unit 32, the dialogue with the user is classified into one of four types of interactive classifications. An interactive classification discriminating unit 334 for discriminating, a dialogue history storage unit 36, and a response generating unit 38 are provided.

  Next, a response generation processing routine according to the third embodiment will be described with reference to FIG. In addition, about the process similar to 1st Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  First, in step 100, measurement of no-input time is started, and in the next step 350, it is determined whether or not an input sentence is input as text data from the input unit 312. When the user inputs an input sentence using the input unit 312, in step 352, it is determined whether or not the input sentence input in step 350 represents a question or a request.

  If it is determined in step 352 that the input sentence from the user represents a question or a request, it is determined that the dialogue with the user is a question answering dialogue, and in step 110, the input sentence from the user is On the other hand, a response sentence corresponding to the question response type dialogue is generated and output to the display unit 316, and the response generation processing routine is terminated.

  If it is determined in step 352 that the input sentence from the user does not represent a question or a request, in step 112, the probability of an information acquisition type dialogue and the probability of an information acceptance type dialogue are calculated. To do. In step 114, it is determined whether or not the probability of the information acceptance type dialogue is greater than the probability of the information acquisition type dialogue. If the probability of the information acceptance type dialogue is greater than the probability of the information acquisition type dialogue, In step 116, it is determined that the dialogue with the user is an information reception type dialogue. Based on the structure of the input sentence from the user and the response database 48A, the input dialogue from the user is changed to the information reception type dialogue. A corresponding response sentence is generated and output to the display unit 316, and the response generation processing routine ends.

  If the probability of the information acquisition dialogue is equal to or less than the probability of the information acceptance dialogue in step 114, it is determined that the dialogue with the user is an information acquisition dialogue. In step 118, the stored dialogue is determined. Based on the history and the previous user input sentence, a response sentence corresponding to the information acquisition type dialogue is generated and output to the display unit 316, and the response generation processing routine is terminated.

  If there is no input from the input unit 312 in step 350, it is determined in step 120 whether or not the measured no-input time is 5 seconds or longer, and there is a state in which no input sentence is input from the user. If the duration is less than 5 seconds, the process returns to step 350. On the other hand, if it is determined in step 120 that the measured no-input time is 5 seconds or longer, each of the probability of the information providing dialogue and the probability of the information acquisition dialogue is calculated in step 122. . In step 124, it is determined whether or not the probability of being an information providing dialogue is greater than the probability of being an information acquiring dialogue, and if the probability of being an information providing dialogue is greater than the probability of being an information obtaining dialogue. It is determined that the dialogue with the user is an information provision type dialogue, and in step 126, information provision is provided for the input sentence from the user based on the dialogue history, the previous input sentence from the user, and the information database 42A. A response sentence corresponding to the type dialogue is generated and output to the display unit 316, and the response generation processing routine is terminated.

  On the other hand, if the probability of the information acquisition type dialogue is greater than or equal to the probability of the information provision type dialogue in step 124, it is determined that the dialogue with the user is the information acquisition type dialogue, and the process proceeds to step 118. Then, a response sentence corresponding to the information acquisition type dialogue is generated and output to the display unit 16, and the response generation processing routine is terminated.

  Then, the response sentence generated as described above is displayed on the display unit 16.

  Thus, based on the no-input time, the transition probability of the interactive classification, and whether the input sentence from the user represents a request or a question, the interactive classification of the interaction with the user is an information providing conversation, Appropriate response to any input by discriminating between information acquisition type dialogue, question answering type dialogue and information acceptance type dialogue, and generating a response sentence according to the discriminated dialogue type Can continue the dialogue.

  In the first to third embodiments, the case where the user interaction is determined from the four types of interactive classifications has been described as an example. However, the present invention is not limited to this. Instead, the determination may be made from three types of interactive classifications, that is, an information provision type dialogue, a question answering type dialogue, and an information reception type dialogue. In this case, when the non-input time is equal to or longer than the predetermined time, it is determined that the dialogue is an information provision type dialogue. When it is determined that the utterance from the user is not a question or a request, the information acceptance type dialogue is performed. What is necessary is just to discriminate | determine. Further, it may be determined from three types of interactive classifications, that is, an information acquisition type dialogue, a question answering type dialogue, and an information acceptance type dialogue. In this case, if the non-input time is equal to or longer than the predetermined time, it is determined that the dialogue is an information acquisition type. If it is determined that the utterance from the user is not a question or a request, the calculated information acquisition is performed. What is necessary is just to discriminate | determine that it is either an information acquisition type | mold dialog and an information reception type | mold dialog based on the probability of a type | mold dialog, and the probability of an information reception type | mold dialog.

  In addition, the case where the probability of each interactive classification is calculated using the previous discrimination result and the transition probability has been described as an example, but the present invention is not limited to this. The probability of each interactive classification may be calculated using the probability.

  In addition, the case where the information database, the question database, the question answer database, and the response database are provided inside the dialogue apparatus has been described as an example. However, the information database, the question database, and the question answer are not limited thereto. The database and the response database may be provided outside the interactive apparatus, and these databases and the interactive apparatus may be connected via a network so that the interactive apparatus accesses these databases via the network.

It is the schematic which shows the structure of the dialogue apparatus which concerns on the 1st Embodiment of this invention. It is a state transition diagram of interactive classification. It is a block diagram which shows the structure of the response production | generation part of the dialogue apparatus which concerns on the 1st Embodiment of this invention. It is a flowchart which shows the content of the response generation process routine of the dialogue apparatus which concerns on the 1st Embodiment of this invention. It is the schematic which shows the structure of the dialogue apparatus which concerns on the 2nd Embodiment of this invention. It is the schematic which shows the structure of the dialogue apparatus which concerns on the 3rd Embodiment of this invention. It is a flowchart which shows the content of the response generation process routine of the dialogue apparatus which concerns on the 3rd Embodiment of this invention.

Explanation of symbols

10, 210 Dialogue device 12 Microphone 14 Speaker 16 Recognition dictionary database 18 Concept word dictionary database 20, 220 Concept expression database 21 Unknown response database 22, 222 Computer 218 Adjective dictionary database

Claims (12)

An input means for inputting at least one of an utterance and an input sentence by the user;
No-input time measuring means for measuring the time during which no input to the input means continues,
Request question determination means for determining whether at least one of the utterance and the input sentence input to the input means represents the request or question of the user;
When the time measured by the non-input time measuring means is a predetermined time or more, it is determined that the dialogue with the user is an information providing type in which the device itself provides information voluntarily, and the request question When it is determined by the determination means that the user's request or question is represented, it is determined that the dialogue with the user is a question response type in which the own device answers the user's request or question, When it is determined that the request question determination unit does not represent the user's request or question, the user's device is an information reception type in which the user's own device accepts information that the user voluntarily provides. Interactive discrimination means for discriminating;
Response generation means for generating a response sentence according to a determination result by the interactive determination means for at least one of the utterance and the input sentence;
Output means for outputting a response sentence generated by the response generation means;
Interactive device including An input means for inputting at least one of an utterance and an input sentence by the user;
No-input time measuring means for measuring the time during which no input to the input means continues,
Request question determination means for determining whether at least one of the utterance and the input sentence input to the input means represents the request or question of the user;
When the time measured by the no-input time measuring means is a predetermined time or more, it is determined that the dialogue with the user is an information acquisition type in which the own device asks a question spontaneously, and the request question determining means When it is determined that the user's request or question is represented by the user's request or question, it is determined that the user's request or question is a question answering type in which the device answers the request. When it is determined by the question determination means that the request or question of the user is not represented, it is determined that the interaction with the user is an information reception type in which the device itself receives information that the user voluntarily provides Interactive discrimination means,
Response generation means for generating a response sentence according to a determination result by the interactive determination means for at least one of the utterance and the input sentence;
Output means for outputting a response sentence generated by the response generation means;
Interactive device including An input means for inputting at least one of an utterance and an input sentence by the user;
No-input time measuring means for measuring the time during which no input to the input means continues,
Request question determination means for determining whether at least one of the utterance and the input sentence input to the input means represents the request or question of the user;
In the dialogue between the user and the user device, the information providing type in which the user device voluntarily provides information, the information acquisition type in which the user device voluntarily asks a question, the question response in which the user device answers the user request or question A transition history storage means for storing a transition history of an interactive classification consisting of a type and an information acceptance type in which the device accepts information that the user voluntarily provides;
When the time measured by the non-input time measuring means is a predetermined time or more, based on the past determination result and the transition history, the interaction with the user is either the information providing type or the information acquiring type. If it is determined by the request question determination means that it represents the user's request or question, it is determined that the interaction with the user is the question response type, and the request If it is determined by the question determination means that it does not represent the user's request or question, the interactive determination means for determining that the interaction with the user is the information acceptance type,
Response generation means for generating a response sentence according to a determination result by the interactive determination means for at least one of the utterance and the input sentence;
Output means for outputting a response sentence generated by the response generation means;
Interactive device including   When the interactive determination means determines that the request question determination means does not represent the user's request or question, based on the past determination result and the transition history, the interaction with the user is: The interactive apparatus according to claim 3, wherein the interactive apparatus determines that the information reception type or the information acquisition type is selected.   When it is determined that the response generation unit is the information providing type, the response generation unit uses the information selected from the plurality of pieces of information stored in the information database storing a plurality of pieces of information to be provided to the user. The interactive apparatus according to claim 1, 3, or 4.   When it is determined that the response generation means is the question response type, the question sentence stored in the question answer database in which the question sentence or the request sentence and the answer sentence to the question sentence or the request sentence are stored in association with each other Alternatively, from the request sentence, the question sentence or the request sentence corresponding to the request or question represented by at least one of the utterance and the input sentence is searched, and the response sentence is used by using the answer sentence to the searched question sentence or request sentence. The interactive device according to claim 1, wherein the interactive device is generated. An analysis unit that analyzes the structure of at least one of the utterance and the input sentence input by the input unit;
When it is determined that the response generation unit is the information acceptance type, the structure is stored in a response database in which the structure and a response sentence to the structure are associated with each other and analyzed by the analysis unit The dialogue apparatus according to claim 1, wherein the response sentence is generated using a response sentence corresponding to.   When it is determined that the response generation unit is the information acquisition type, the response generation unit uses a question sentence selected from the plurality of question sentences stored in a question request database storing a plurality of question sentences or request sentences to the user. 5. The dialogue apparatus according to any one of claims 2 to 4, wherein the response sentence is generated. The input means inputs an utterance by a user,
The request question determination unit extracts a feature amount of speech for the utterance input from the input unit, and the extracted feature amount and the feature amount of an utterance representing a question or request obtained in advance. The dialogue apparatus according to any one of claims 1 to 8, wherein at least one of the utterance and the input sentence indicates whether or not it represents a user request or a question. Computer
A no-input time measuring means for measuring the time during which there is no input to the input means for inputting at least one of the utterance and the input sentence by the user,
Request question determination means for determining whether at least one of the utterance and the input sentence input to the input means represents the user's request or question,
When the time measured by the non-input time measuring means is a predetermined time or more, it is determined that the dialogue with the user is an information providing type in which the device itself provides information voluntarily, and the request question When it is determined by the determination means that the user's request or question is represented, it is determined that the dialogue with the user is a question response type in which the own device answers the user's request or question, When it is determined that the request question determination unit does not represent the user's request or question, the user's device is an information reception type in which the user's own device accepts information that the user voluntarily provides. And an interactive type discriminating unit that discriminates between the utterance and the input sentence, and a response generating unit that generates a response sentence according to the discrimination result by the interactive type discriminating unit Because of the program. Computer
A no-input time measuring means for measuring the time during which there is no input to the input means for inputting at least one of the utterance and the input sentence by the user,
Request question determination means for determining whether at least one of the utterance and the input sentence input to the input means represents the user's request or question,
When the time measured by the no-input time measuring means is a predetermined time or more, it is determined that the dialogue with the user is an information acquisition type in which the own device asks a question spontaneously, and the request question determining means When it is determined that the user's request or question is represented by the user's request or question, it is determined that the user's request or question is a question answering type in which the device answers the request. When it is determined by the question determination means that the request or question of the user is not represented, it is determined that the interaction with the user is an information reception type in which the device itself receives information that the user voluntarily provides An interactive discriminating means for generating a response sentence corresponding to a discrimination result by the interactive discriminating means for at least one of the utterance and the input sentence. Program. Computer
A no-input time measuring means for measuring the time during which there is no input to the input means for inputting at least one of the utterance and the input sentence by the user,
Request question determination means for determining whether at least one of the utterance and the input sentence input to the input means represents the user's request or question,
In the dialogue between the user and the user device, the information providing type in which the user device voluntarily provides information, the information acquisition type in which the user device voluntarily asks a question, the question response in which the user device answers the user request or question A transition history storage means for storing a transition history of interactive classification consisting of a type and an information reception type in which the device accepts information voluntarily provided by the user,
When the time measured by the non-input time measuring means is a predetermined time or more, based on the past determination result and the transition history, the interaction with the user is either the information providing type or the information acquiring type. If it is determined by the request question determination means that it represents the user's request or question, it is determined that the interaction with the user is the question response type, and the request When it is determined by the question determination means that the user's request or question is not represented, the interactive determination means for determining that the dialogue with the user is the information acceptance type, and the utterance and the input sentence A program for causing at least one to function as a response generation unit that generates a response sentence according to a determination result by the interactive determination unit.

Images