JP2010262147A - Response creating device and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To achieve natural dialogues by creating an appropriate response text considering negative intentions. <P>SOLUTION: When a negative expression determination part 22 determines that no negative expression is included, a result using response creating part 26 creates the response text based on the analysis result. When the negative expression determination part determines that negative expression is included and the preceding response type is "response to inquire object words in parallel", a topic change response creating part 30 creates a response text for changing the topic. When the negative expression determination part determines that negative expression is included and the preceding response type is other than "response to inquire object words in parallel", an affirmative response creating part 32 creates a response text for affirmative response. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a response generation device and a program, and more particularly, to a response generation device and a program for performing a smooth dialogue with a user.

  Conventionally, from the system side, when requesting a reply of contents meaning affirmation or denial from the system side, switching to affirmative / negative identification mode and generating a response based on the result, in other cases, A speech recognition dialogue apparatus that generates a response based on a user's recognition result is known (Patent Document 1). In this speech recognition dialogue apparatus, for example, the system makes an utterance that clearly requests “Yes / No” from the user, such as “Let's go to see cherry blossoms” or “Have you tomorrow?” Only in the case where the user utterance is switched to the positive / negative identification mode, it is determined whether the next user utterance is “Yes” or “No”, and a system response is generated accordingly.

JP-A-9-6389

  In the technology described in Patent Document 1, the system switches to the positive / negative detection mode only when the system generates an utterance that clearly requests “Yes / No” to the user. Users such as 5W1H question responses such as “Did you go somewhere else?” Or “How much were you fun?” Or empathy responses using emotion estimation results such as “I was fun?” Even if it is not clearly requesting “Yes / No”, negative expressions such as “No, I am not going anywhere” or “No, I was sad” in some cases There is a possibility that a response containing "" will be returned from the user. There is a problem that an appropriate response cannot be generated by capturing a negative intention with respect to the utterance from such a user.

  The present invention has been made to solve the above-described problem, and provides a response generation apparatus and program capable of generating an appropriate response sentence in consideration of a negative intention and performing a natural dialogue. With the goal.

  In order to achieve the above object, the response generation apparatus according to the present invention determines whether or not a predetermined negative expression is included in the input sentence based on an analysis result obtained by analyzing a structure of the input sentence from the user. If it is determined that the negative expression is not included, and a negative expression determination means for determining whether or not the negative expression is included, at least one response type including a response asking for a parallel object is selected based on the analysis result A first response for generating a response sentence prepared in advance for the selected response type or a response sentence in which a word included in the input sentence is added to a response sentence format prepared in advance. When it is determined that the generation means and the negative expression are included, and the response type of the response sentence generated last time is a response for asking a parallel object, prepared in advance ,topic A second response generating means for generating a response sentence for changing, and an object in which the response type of the response sentence generated last time is determined to include the negative expression, If the response is not a response asking for a response, a third response generation means for generating a response sentence prepared in advance is included.

  The program according to the present invention is a negative expression determination means for determining whether or not a predetermined negative expression is included in the input sentence based on an analysis result obtained by analyzing a structure of the input sentence from a user. If it is determined that the negative expression is not included, at least one response type including a response asking for a parallel object is selected based on the analysis result, and the selected response type is selected. First response generation means for generating a response sentence prepared in advance for the type of response, or a response sentence in which a word included in the input sentence is added to a response sentence format prepared in advance, the negative expression included If the response type of the response sentence generated last time is a response asking for a parallel object, a response prepared for changing the topic is prepared in advance. Second response generation means for generating a sentence and a case where it is determined that the negative expression is included, and the response type of the response sentence generated last time is not a response asking for a parallel object In this case, it is a program for functioning as third response generation means for generating a response sentence of a prepared prepared answer.

  According to the present invention, the negative expression determination unit determines whether or not a predetermined negative expression is included in the input sentence based on the analysis result obtained by analyzing the structure of the input sentence from the user. If it is determined that a negative expression is not included, the first response generation means selects at least one response type including a response asking for a parallel object based on the analysis result, and the selected response is selected. For the response type, a response sentence prepared in advance for the response type or a response sentence in which a word included in the input sentence is added to a response sentence format prepared in advance is generated.

  Further, when it is determined that a negative expression is included, and the type of response of the response sentence generated last time is a response asking for a parallel object, the second response generation means A response sentence prepared in advance for changing the topic is generated. If it is determined that a negative expression is included, and the response type of the response sentence generated last time is not a response asking for a parallel object, it is prepared in advance by the third response generation unit. Generates a response message for each other.

  In this way, when a negative expression is included, if the previous response type is a response asking for an object in parallel, a response sentence is generated to change the topic, and the previous response type is parallel. If it is not a response asking for an object, an appropriate response sentence in consideration of the intention of negation can be generated by generating a response sentence that is compatible, and a natural dialogue can be performed.

  The first response generation means of the present invention asks for a parallel object based on the analysis result when the input sentence includes a clause representing a predetermined situation with respect to a response asking for the parallel object. Select a response and add a clause indicating the situation included in the input sentence to the response sentence prepared in advance for asking the parallel object for the clause indicating the situation included in the input sentence or a prepared response sentence format Can be generated.

  The negative expression judging means according to the present invention negates the input sentence based on the analysis result, depending on whether or not the input sentence includes a predetermined expression as a negative expression or a combination of a part of speech type and an expression. It can be determined whether an expression is included.

  The above input sentence can be acquired by voice recognition of the utterance by the user.

  As described above, according to the response generation device and the program of the present invention, when a negative expression is included, the type of the previous response is a response for asking an object to be paralleled. If a response sentence is generated and the previous response type is not a response asking for an object in parallel, an appropriate response sentence considering negative intention is generated by generating a companion response sentence, and natural dialogue The effect that it can be performed is acquired.

It is a block diagram which shows schematic structure of the response generation apparatus which concerns on this Embodiment. It is an image figure which shows an example of a negative expression dictionary. It is an image figure which shows an example of a situation noun dictionary. It is an image figure which shows an example of the response sentence example of a parallel object question, and a response sentence example format. It is an image figure which shows an example of the example of a response sentence of a topic change response. It is an image figure which shows an example of the example of a response sentence of a conflict response. It is a flowchart which shows the content of the response production | generation process routine in the response production | generation apparatus which concerns on this Embodiment. It is an image figure which shows an example of the result of a morphological analysis. It is an image figure which shows another example of the result of a morphological analysis. It is an image figure which shows another example of the result of a morphological analysis. It is an image figure which shows another example of the result of a morphological analysis.

  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 a response generation apparatus that executes a predetermined process and outputs a voice by using an utterance from a user as an input.

  As shown in FIG. 1, the response generation apparatus 10 according to the present embodiment is connected to a microphone 12 that collects user utterances to generate an audio signal, a speaker 14 that outputs audio, and the microphone 12 and the speaker 14. And a computer 16 for executing a predetermined process for generating an appropriate response sentence.

  The computer 16 includes a CPU that controls the entire response generation apparatus 10, a ROM as a storage medium that stores various programs such as a response generation program to be described later, a RAM that temporarily stores data as a work area, and a storage unit that stores various information. As an HDD, an I / O (input / output) port, and a bus connecting them. A microphone 12 and a speaker 14 are connected to the I / O port.

  The computer 16 will be described with functional blocks divided for each function realization means determined based on hardware and software. As shown in FIG. A language analysis unit 20 that performs morphological analysis on character string information indicating a user utterance that has been voice-recognized using a typical morphological analyzer, and a negative expression is included in the user utterance based on the analysis result of the language analysis unit 20 A negative expression determination unit 22 that determines whether or not a response is generated, a response type storage unit 24 that stores a response type of a response sentence generated last time, and a language analysis unit 20 when the user utterance does not include a negative expression The result use response generation unit 26 that generates a response sentence using the analysis result of the above and stores the response type of the generated response sentence in the response type storage unit 24, and the user utterance If a negative expression is included, the immediately preceding response determination unit 28 that determines whether the response type of the response sentence generated immediately before is “response for asking parallel objects”, and the response generated immediately before When the response type of the sentence is “response for asking for parallel objects”, a response sentence for changing the topic is generated, and the response type “topic change response” of the generated response sentence is set as the response type storage unit 24. When the response change type of the topic change response generation unit 30 to be stored and the response sentence generated immediately before is not “response for asking for parallel objects”, a response sentence of a companion is generated and a response of the generated response sentence It can be expressed by a configuration including an interaction generation unit 32 that stores the type “interaction response” in the response type storage unit 24, and an output unit 34 that converts the generated response sentence into an audio signal and outputs it from the speaker 14. .

  The negative expression determination unit 22 determines whether or not a predetermined negative expression is included in the analysis result of the language analysis unit 20 for the user utterance. A negative expression dictionary storing a plurality of combinations of expressions as negative expressions and types of parts of speech and expressions as shown in FIG. 2 is created in advance, and the determination is made with reference to the negative expression dictionary at the time of determination. . For example, if the analysis result includes (i) “part of speech“ verb ”+“ not ””, (ii) “part of speech“ verb ”+“ not ””, (iii) “not” (Iv) If “not” is included, it is determined that a negative expression is included.

  The result use response generation unit 26 selects a response type based on the analysis result of the language analysis unit 20 with respect to the user utterance as described below, and generates a response sentence candidate for each selected response type.

  First, it is determined whether or not a word representing the situation is included in the analysis result of the language analysis unit 20 for the user utterance. A situation is a person's behavior or an event or event that a person experiences. Here, when the analysis result includes (i) “verb” and (ii) “situation noun + (“ da ”or“ is ””), “situation” It is determined that it is included. “Situation noun” is a noun indicating an event such as “rain” or “athletic meet”. Situation nouns are determined in advance by creating a situation noun dictionary as shown in FIG. 3 and referring to this situation noun dictionary.

  Next, a clause representing the situation is extracted from the analysis result of the user utterance determined to contain the situation. In the case of (i) above, a clause including the corresponding “verb” is extracted, and in the case of (ii), a clause including the corresponding “situation noun” is extracted.

  It is determined whether or not the extracted clause is registered in a parallel object questionable determination DB, which will be described later. If it is registered, a response type “response for asking parallel objects” is selected.

  When the response type “response for asking parallel objects” is selected, a question sentence for asking the user for parallel objects is generated as a response sentence candidate. For example, a response sentence example and a response sentence example format as shown in FIG. 4 are prepared in advance, and the response sentence candidates are randomly selected from the response sentence examples or generated using the response sentence example format. The response sentence example format can be, for example, a format such as “What else?”, And a “verb” of a clause representing a situation extracted from the analysis result of the user utterance in the “˜” part. A response sentence candidate can be generated by inserting a certain word. The generated response sentence candidate is temporarily stored in a predetermined storage area.

  Further, it is determined whether or not a clause representing the situation extracted from the analysis result of the user utterance determined to contain the situation is registered in the frequency question impossible judgment DB described later. If it is not registered, the response type “response that asks the frequency of the situation” is selected. When the response type “response that asks the frequency of the situation” is selected, a question message that asks the user the frequency is generated as a response message candidate. For example, a response sentence candidate and a response sentence example format are prepared in advance, and the response sentence candidate is selected at random from the response sentence example or generated using the response sentence example format. The response sentence example format can be, for example, a format such as "Do you do well?", And insert a clause indicating the situation extracted from the analysis result of the user utterance in the "~" part. Sentence candidates can be generated. The generated response sentence candidate is temporarily stored in a predetermined storage area. In the frequency question impossibility determination DB, sections representing situations where it is inappropriate to ask the frequency are registered in advance.

  In addition, a word representing a situation is extracted from the analysis result of the user utterance determined to contain the situation. When (i) “verb” is included in the analysis result, the word that is “verb” is extracted, and (ii) “situation noun + (“ da ”or“ is ””) is included If so, a word that is a “situation noun” is extracted.

  Then, it is determined whether or not the extracted word is registered in the emotion questionability determination DB 28 described later. If registered, the response type “response to ask the user's emotion to the situation” is selected. When the response type “response that asks the user's emotion to the situation” is selected, a question message that asks the user about the emotion is generated as a response message candidate. Response sentence candidates are selected as basic question sentences, for example, “What do you think?” And “How do you feel?” In advance, and selected from these basic question sentences based on the analysis results of user utterances. The response sentence candidate is generated by correcting the tense and expression (such as polite expression or broken expression) of one sentence. Alternatively, a question sentence example may be prepared in advance and selected randomly from the above. The generated response sentence candidate is temporarily stored in a predetermined storage area. In addition, in the emotion questionable determination DB, words that are appropriate for the words to ask for emotions are registered in advance.

  Further, “verb” and “case particle” in the user utterance are extracted from the analysis result of the language analysis unit 20 for the user utterance. For example, considering the user utterance “I spoke with a friend on the phone”, the verb “speak” and the case particles “to” and “de” are extracted. Next, the indispensable case of the extracted verb is searched using a verb indispensable case dictionary database created in advance. The verb required case dictionary database is a dictionary in which “the case that the verb should have at least” is registered. In the above example, as the required case of the verb “speak”, “ga (who)”, “(what) "," And (who) "are registered. Next, if the deficient case of the extracted verb is detected by comparing the required case of the retrieved verb with the case particle extracted from the user's utterance, and the deficient case of the verb is detected The response type “response to ask for deficiency” is selected. When the response type “response asking for deficiency” is selected, a question sentence asking for deficiency is generated as a response sentence candidate. For example, a response sentence example format is prepared in advance, and the response sentence candidate is generated using the response sentence example format. The response sentence example format can be, for example, a format such as “What?” Corresponding to the deficiency “(What)”, and extracted from the analysis result of the user utterance in the “˜” part. A candidate for a response sentence can be generated by inserting a clause representing the selected verb. For example, in the above example (“I talked with a friend over the phone”), “What (what)” is insufficient, so the question sentence “What did you say?” Is generated as a response candidate. . The generated response sentence candidate is temporarily stored in a predetermined storage area.

  Regardless of the analysis result in the language analysis unit 20, the response type “repetitive response” is selected, and a question sentence for repeatedly confirming a word included in the user utterance is generated as a response sentence candidate. For example, a response sentence candidate format is prepared in advance, and the response sentence candidate is generated using the response sentence example format. The response sentence example format can be, for example, a format such as “~?”, And the word included in the analysis result of the language analysis unit 20 for the user utterance is inserted into the “˜” part, and the response sentence is inserted. Can be generated.

  The parallel object questionable determination DB may be generated as follows.

  For example, one sentence is extracted from a large-scale corpus such as text on the web, and morphological analysis is performed on the extracted one sentence. Judgment is made on whether or not the analysis result includes a situation, and if the situation is included, whether or not a verb with the object of noun parallel is included in the given language pattern Judgment based on whether or not. If included, a sentence containing verbs whose object is parallel nouns is divided into clauses whose object is each of multiple parallel nouns, and each clause represents a situation Extract as a clause.

  The section representing the extracted situation is registered in the situation DB in which the section representing the situation and the count value obtained by counting the appearance frequency of the section are defined. In the case of the same clause as that already registered, the count value is incremented by one.

  It is determined whether or not the above processing has been performed for all sentences on the corpus, and if there is an unprocessed sentence, the next sentence is processed.

  Referring to the situation DB, a clause representing a situation where the appearance frequency is equal to or higher than the threshold is registered in the parallel object questionable judgment DB. A high count value (occurrence frequency) in the situation DB often appears as a sentence whose object is parallel nouns originally, and it is assumed that there are other omitted objects. The question asking for the object to do is appropriate. When all the clauses whose count value is greater than or equal to the threshold value are registered in the parallel object questionable determination DB, this process is terminated.

  With the above processing, a parallel object questionable determination DB is generated in which a clause representing a predetermined situation is registered for a response asking for a parallel object.

  The result use response generation unit 26 generates a response sentence candidate as described above, and randomly selects one of the generated response sentence candidates as a response sentence. In addition, the result use response generation unit 26 selects the response types of the selected response sentence candidates “response to ask for parallel objects”, “response to ask the frequency of the situation”, “response to ask the user's feelings about the situation”, “ The response type storage unit 24 stores “repeat response” or “response to ask for deficiency”.

  Based on the response type stored in the response type storage unit 24, the immediately preceding response determination unit 28 determines whether the response type of the response sentence generated immediately before is “response for asking parallel objects”. To do.

  The topic conversion response generator 30 generates a response sentence for changing the topic. As the response text, for example, a response text example as shown in FIG. 5 is prepared in advance, and is selected at random from this response text example. Further, the topic conversion response generation unit 30 generates a response sentence for changing the topic, and stores the response type “topic conversion response” in the response type storage unit 24.

  The conflict generation unit 32 generates a response statement of the conflict. For example, a response sentence example as shown in FIG. 6 is prepared in advance, and the response sentence is selected at random from the response sentence example. In addition, the conflict generation unit 32 generates a response statement of the conflict and stores the response type “conflict response” in the response type storage unit 24.

  Next, a response generation processing routine in the response generation device 10 of the present exemplary embodiment will be described with reference to FIG. This routine is performed by the CPU executing a response generation program stored in the ROM.

  In step 100, it is determined whether or not a user utterance has been input from the microphone 12. If a user utterance has been input, the process proceeds to step 102. If not, the determination in this step is repeated until it is input. . Here, it is assumed that the user utterance “I went to an amusement park” is input.

  In step 102, the input voice signal indicating the user utterance is recognized as voice string information, and morphological analysis is performed on the string information. Here, as a result of analysis, as shown in FIG. 8, “swimming (noun)” “shi (verb)” “ni (particle)” “rai (verb)” “ta (auxiliary verb)” “yo (particle)” Is obtained.

  Next, in step 104, it is determined whether or not a negative expression is included in the user utterance based on the analysis result of the morphological analysis. If a negative expression is included, the process proceeds to step 112; otherwise, the process proceeds to step 106. Here, since a negative expression is not included, it is determined that a negative expression is not included, and the process proceeds to step 106.

  In step 106, a response type is selected based on the analysis result of the morphological analysis. First, based on the analysis result of the morphological analysis, a response type is selected by determining whether the user utterance includes a clause representing a situation registered in the parallel object questionable determination DB. If it is included, select “response to ask parallel objects”. Further, the response type is selected by determining whether the user utterance includes a clause representing a situation registered in the frequency question impossibility determination DB. If it is not included, select “response to ask the frequency of the situation”. In addition, the response type is selected by determining whether the user utterance includes a word representing a situation registered in the emotion questionable determination DB. If it is included, the “response for asking the user's emotion to the situation” is selected. Further, “repetitive response” is selected regardless of the analysis result of the morphological analysis.

  In step 107, response sentence candidates are generated for each of the response types selected in step 106. For example, “Claim (verb)” included in the analysis result is inserted into the response sentence format, and “Did you come” as a response sentence candidate for “Repetitive response”.

  In the next step 108, one response sentence candidate randomly generated in the above step 107 is selected as a response sentence, and the response type stored in the response type storage unit 24 is selected as the response selected in the above step 108. The response type of the sentence candidate is updated, and the process proceeds to step 110. Here, a response sentence candidate of “repeat response” is selected, and “Did you come” is the generated response sentence.

  In step 110, the generated response sentence is converted into an audio signal and output. Here, “Did you come?” Is output as a voice, and the process returns to step 100.

  Next, it is assumed that a user utterance “I did not come” but a system response “Did you come?” Is input. At this time, in step 102, as the analysis result, as shown in FIG. 9, “Kai (verb)” “Ta (auxiliary verb)” “N (noun)” “Ja (particle)” “None (auxiliary verb)” “ (Verb), “go (verb),“ ta (auxiliary verb) ”and“ no (noun) ”.

  In step 104 above, based on the analysis result of the morphological analysis, it is determined that the negative utterance “not not ~” is included in the user utterance. In step 112, the response type of the response sentence generated immediately before is It is determined whether or not “response for asking for parallel objects”. If it is “response for asking for parallel object”, the process proceeds to step 114, but if it is other than “response for asking for parallel object”, the process proceeds to step 116. Here, since “repeat response” is stored in the response type storage unit 24, it is determined that the response type storage unit 24 is not “response for asking parallel objects”, and the process proceeds to step 114.

  In step 114, a response sentence (for example, “Oh, what is it?”) Is generated by selecting from standard sentences prepared in advance for the response, and the response stored in the response type storage unit 24. The type is updated to “confusion response”, and the process proceeds to step 110.

  Further, it is assumed that the user utterance “I went to an amusement park” is input. At this time, in step 102, as a result of analysis, as shown in FIG. 10, “swimming (noun)” “shi (verb)” “ni (particle)” “go (verb)” “ta (auxiliary verb)” “ Yo (particle) ".

  In this case, in step 104, it is determined that the user utterance does not include a negative expression. In step 106, the user utterance is a clause ("Done" indicating a situation registered in the parallel object questionable determination DB. Suppose that it is determined that a clause that includes the verb “)” is included. A response type including “response for asking parallel objects” is selected as the response type, and in step 108, a response sentence candidate of “response asking for parallel objects” is generated. "Is selected, and the response type stored in the response type storage unit 24 is updated to" response for asking parallel objects ".

  Then, it is assumed that a user utterance “I have not gone anywhere” is input to the system response “Did you go elsewhere?” At this time, in step 102, as shown in FIG. 11, as the analysis result, “where (noun)” “ni (particle)” “go (verb)” “te (auxiliary verb)” “not (auxiliary verb)” “ Yo (particle) ".

  In step 104 described above, it is determined that the user utterance contains the negative expression “part of speech“ verb ”+“ no ”” based on the analysis result of the morphological analysis. In step 112, the response of the response sentence generated immediately before is determined. It is determined that the type is “response for asking parallel objects”. In step 116, a response sentence for changing the topic (for example, “Let's change the topic”) is generated by selecting from the standard sentences prepared in advance for the topic change response, and the response type is stored. The response type stored in the unit 24 is updated to “topic change response”, and the process proceeds to step 110.

  By executing the response generation processing routine described above, for example, when the recognition result is “I came to swim” for a user utterance “I went to an amusement park”, a response message “Return response” Tantan "is generated. At this time, when the user makes a correction utterance “No, I did.”, For example, “Oh, that ’s right” is generated for this correction utterance, so that a false recognition loop occurs. No, you can continue the conversation smoothly. For example, it is possible to avoid a misrecognition loop in which the “performed” part is erroneously recognized as “cut” and a response is generated again using the erroneous recognition result.

  Also, for example, when a user utters “No, I do n’t go anywhere” in response to a response asking for a parallel object like “Did you go elsewhere?” Since a response sentence that changes the topic such as “Let's change the topic” is generated, the conversation can be continued smoothly.

  As described above, according to the response generation device of the present embodiment, when a negative expression is included in the user utterance, if the immediately preceding response type is “response for asking parallel objects”, the topic If the response type immediately before the response type is something other than "a response that asks for a parallel object", an appropriate response sentence that takes into account the negative intention is generated by generating a response sentence Generate and interact naturally.

  Not only when a response (= Yes / No question response or solicitation response) that explicitly requests the user to say “affirmation / denial” is generated, but even when a 5W1H question response or repeated response is generated, the user Since it is determined whether or not a negative expression is included in an utterance and a response is generated based on the result, a smoother conversation is possible.

  In the above-described embodiment, the case where voice output by a speaker is performed has been described as an example. However, the present invention is not limited to this, and a response sentence may be displayed on a display. Moreover, although the case where audio | voice was input into the microphone from the user was demonstrated to the example, you may make it a user input the text as an input sentence using a keyboard etc.

  When generating a response sentence using the analysis result of morphological analysis, at least one response type is selected based on the analysis result of morphological analysis, and the response sentence generated for each selected response type The case where response sentences are randomly selected from candidates has been described as an example, but the present invention is not limited to this. For example, based on the analysis result of morphological analysis, only one response type is selected, a response sentence is generated for the selected response type, and the selected response type is stored in the response type storage unit 24. It may be.

  It is also possible to provide the program of the present invention by storing it in a storage medium.

DESCRIPTION OF SYMBOLS 10 Response production | generation apparatus 12 Microphone 14 Speaker 16 Computer 20 Language analysis part 22 Negative expression determination part 24 Response type memory | storage part 26 Result utilization response generation part 28 Immediate response determination part 30 Topic change response generation part 32 Affinity generation part 34 Output part

Claims (5)

Negative expression determination means for determining whether or not a predetermined negative expression is included in the input sentence based on an analysis result obtained by analyzing a structure of the input sentence from the user;
If it is determined that the negative expression is not included, based on the analysis result, select at least one response type including a response asking for a parallel object, and for the selected response type, First response generation means for generating a response sentence prepared in advance for the type of response, or a response sentence in which a word included in the input sentence is added to a response sentence format prepared in advance;
If it is determined that the negative expression is included, and the response type of the response sentence generated last time is a response for asking for a parallel object, the prepared topic is changed. Second response generation means for generating a response sentence for
If it is determined that the negative expression is included, and the response type of the response sentence generated last time is not a response for asking for a parallel object, a response sentence prepared in advance is used. Third response generating means for generating;
A response generation device including:   The first response generation means inquires the parallel object based on the analysis result when the input sentence includes a clause representing a predetermined situation with respect to a response asking for the parallel object. A response sentence prepared by selecting a response and inquiring a parallel object for a clause representing a situation included in the input sentence, or a clause representing a situation included in the input sentence in a prepared response sentence format The response generation apparatus according to claim 1, wherein a response sentence to which is added is generated.   The negative expression determination means determines whether the input sentence includes an expression predetermined as the negative expression, or a combination of a part of speech and a type of expression based on the analysis result. The response generation device according to claim 1, wherein it is determined whether or not the negative expression is included.   The response generation apparatus according to claim 1, wherein the input sentence is acquired by voice recognition of an utterance made by a user. Computer
Negative expression determination means for determining whether or not a predetermined negative expression is included in the input sentence based on an analysis result obtained by analyzing a structure of the input sentence from the user;
If it is determined that the negative expression is not included, based on the analysis result, select at least one response type including a response asking for a parallel object, and for the selected response type, A first response generation means for generating a response sentence prepared in advance for the type of response, or a response sentence in which a word included in the input sentence is added to a response sentence format prepared in advance;
If it is determined that the negative expression is included, and the response type of the response sentence generated last time is a response for asking for a parallel object, the prepared topic is changed. Second response generation means for generating a response sentence for the response, and when it is determined that the negative expression is included, and the response type of the response sentence generated last time is a parallel object If it is not a response to ask, a program for functioning as a third response generation means for generating a response sentence prepared in advance.

Images