Dialogue support system and dialogue support method

1 . A dialogue support system comprising:

a processor; and

a memory, wherein

the memory holds a plurality of question/answer pair candidates each including a question sentence vector indicating a sentence of question by a questioner, an emotion vector indicating a question emotion that is an emotion of the questioner when the question is asked, and an answer sentence vector indicating a sentence of an answer to the question by a respondent, and question/answer pair information in which a question/answer pair is stored, and

the processor

generates, from the plurality of question/answer pair candidates, a first group based on similarity between the answer sentence vectors and similarity between the question sentence vectors, and a second group based on similarity between the answer sentence vectors and similarity between the emotion vectors,

stores, in a case where it is determined that a variance of the emotion vectors of question/answer pair candidates included in the first group is large based on a predetermined condition, a question/answer pair based on the question/answer pair candidates included in the first group in the question/answer pair information as a sentence dependence question/answer pair in which an answer sentence to a question depends on a sentence of the question,

stores, in a case where it is determined that a variance of the question sentence vectors of question/answer pair candidates included in the second group is large based on a predetermined condition, a question/answer pair based on the question/answer pair candidates included in the second group in the question/answer pair information as an emotion dependence question/answer pair in which an answer sentence to a question depends on the question emotion corresponding to the question,

receives a search instruction including a question sentence vector and an emotion vector,

selects a question/answer pair as a search result for the search instruction from the question/answer pair information based on first similarity between a question sentence vector of the sentence dependence question/answer pair stored in the question/answer pair information and a question sentence vector included in the search instruction, second similarity between an emotion vector of the emotion dependence question/answer pair stored in the question/answer pair information and an emotion vector included in the search instruction, and third similarity between a question sentence vector of the sentence/emotion dependence question/answer pair stored in the question/answer pair information and a question sentence vector included in the search instruction, and

outputs an answer sentence indicated by an answer sentence vector included in the selected question/answer pair.

2 . The dialogue support system according to claim 1 , wherein

the processor

generates, from the plurality of question/answer pair candidates, a third group based on similarity between the answer sentence vectors, similarity between the question sentence vectors, and similarity between the emotion vectors, and

stores a question/answer pair based on question/answer pair candidates included in the third group in the question/answer pair information as a sentence/emotion dependence question/answer pair in which an answer sentence to a question depends on a sentence of the question and the question emotion corresponding to the question.

3 . The dialogue support system according to claim 1 , wherein

the memory holds a dialogue sentence vector indicating a sentence of a dialogue between the questioner and the respondent, and a questioner emotion indicating an emotion of the questioner when an utterance by the questioner is being made in the dialogue, and

the processor

extracts an utterance indicating a question by the questioner in the dialogue or an expression of an emotion of the questioner in the dialogue based on the dialogue sentence vector and the questioner emotion,

extracts an utterance indicating an answer by the respondent, the utterance corresponding to the extracted utterance or the extracted expression of the emotion, based on the dialogue sentence vector,

determines, as the question sentence vector in the question/answer pair candidates, a vector indicating a sentence of an utterance by the questioner indicated by the dialogue sentence vector when the extracted utterance or the extracted expression of the emotion is made,

determines, as the emotion vector in the question/answer pair candidates, a vector indicating the questioner emotion when the extracted utterance or the extracted expression of the emotion is made, and

determines a vector indicating a sentence of the extracted utterance indicating the answer by the respondent as the answer sentence vector in the question/answer pair candidates.

4 . The dialogue support system according to claim 1 , wherein

the processor

calculates a sentence dependence degree indicating a degree of dependence of an answer sentence to a question in the sentence dependence question/answer pair on a sentence of the question based on a reciprocal of a variance of question sentence vectors included in the first group and stores the calculated sentence dependence degree in the question/answer pair information in a case where the sentence dependence question/answer pair is stored in the question/answer pair information, and

calculates an emotion dependence degree indicating a degree of dependence of an answer sentence to a question in the emotion dependence question/answer pair on the question emotion corresponding to the question based on a reciprocal of a variance of emotion vectors included in the second group and stores the calculated emotion dependence degree in the question/answer pair information in a case where the emotion dependence question/answer pair is stored in the question/answer pair information.

5 . A dialogue support method performed by a dialogue support system, wherein

the dialogue support system includes a processor and a memory, and

the memory holds a plurality of question/answer pair candidates each including a question sentence vector indicating a sentence of question by a questioner, an emotion vector indicating a question emotion that is an emotion of the questioner when the question is asked, and an answer sentence vector indicating a sentence of an answer to the question by a respondent, and question/answer pair information in which a question/answer pair is stored, and

the dialogue support method comprising the steps of:

generating, from the plurality of question/answer pair candidates, a first group based on similarity between the answer sentence vectors and similarity between the question sentence vectors, and a second group based on similarity between the answer sentence vectors and similarity between the emotion vectors,

storing, in a case where it is determined that a variance of the emotion vectors of question/answer pair candidates included in the first group is large based on a predetermined condition, a question/answer pair based on the question/answer pair candidates included in the first group in the question/answer pair information as a sentence dependence question/answer pair in which an answer sentence to a question depends on a sentence of the question, and

storing, in a case where it is determined that a variance of the question sentence vectors of question/answer pair candidates included in the second group is large based on a predetermined condition, a question/answer pair based on the question/answer pair candidates included in the second group in the question/answer pair information as an emotion dependence question/answer pair in which an answer sentence to a question depends on the question emotion corresponding to the question,

receiving a search instruction including a question sentence vector and an emotion vector,

selecting a question/answer pair as a search result for the search instruction from the question/answer pair information based on first similarity between a question sentence vector of the sentence dependence question/answer pair stored in the question/answer pair information and a question sentence vector included in the search instruction, second similarity between an emotion vector of the emotion dependence question/answer pair stored in the question/answer pair information and an emotion vector included in the search instruction, and third similarity between a question sentence vector of the sentence/emotion dependence question/answer pair stored in the question/answer pair information and a question sentence vector included in the search instruction, and

outputting an answer sentence indicated by an answer sentence vector included in the selected question/answer pair.

6 . The dialogue support method according to claim 5 , further comprising the steps of:

generating, from the plurality of question/answer pair candidates, a third group based on similarity between the answer sentence vectors, similarity between the question sentence vectors, and similarity between the emotion vectors, and

storing a question/answer pair based on question/answer pair candidates included in the third group in the question/answer pair information as a sentence/emotion dependence question/answer pair in which an answer sentence to a question depends on a sentence of the question and the question emotion corresponding to the question.

7 . The dialogue support method according to claim 5 , wherein

the memory holds a dialogue sentence vector indicating a sentence of a dialogue between the questioner and the respondent, and a questioner emotion indicating an emotion of the questioner when an utterance by the questioner is being made in the dialogue, and

the dialogue support method further comprising the steps of:

extracting an utterance indicating a question by the questioner in the dialogue or an expression of an emotion of the questioner in the dialogue based on the dialogue sentence vector and the questioner emotion,

extracting an utterance indicating an answer by the respondent, the utterance corresponding to the extracted utterance or the extracted expression of the emotion, based on the dialogue sentence vector,

determining, as the question sentence vector in the question/answer pair candidates, a vector indicating a sentence of an utterance by the questioner indicated by the dialogue sentence vector when the extracted utterance or the extracted expression of the emotion is made,

determining, as the emotion vector in the question/answer pair candidates, a vector indicating the questioner emotion when the extracted utterance or the extracted expression of the emotion is made, and

determining a vector indicating a sentence of the extracted utterance indicating the answer by the respondent as the answer sentence vector in the question/answer pair candidates.

8 . The dialogue support method according to claim 5 , further comprising the steps of:

calculating a sentence dependence degree indicating a degree of dependence of an answer sentence to a question in the sentence dependence question/answer pair on a sentence of the question based on a reciprocal of a variance of question sentence vectors included in the first group and stores the calculated sentence dependence degree in the question/answer pair information in a case where the sentence dependence question/answer pair is stored in the question/answer pair information, and

calculating an emotion dependence degree indicating a degree of dependence of an answer sentence to a question in the emotion dependence question/answer pair on the question emotion corresponding to the question based on a reciprocal of a variance of emotion vectors included in the second group and stores the calculated emotion dependence degree in the question/answer pair information in a case where the emotion dependence question/answer pair is stored in the question/answer pair information.