Automatic counseling system

Abstract

To provide an automatic counseling system which can satisfy a cognitive aspect of a user to give a response with which the user has empathy.SOLUTION: An automatic counseling system 100 includes: a storage device 1 including an interaction scenario storage unit 11 storing therein an interaction scenario in which a plurality of question sentences and causal relations between the plurality of question sentences are registered; a question sentence generation unit 21 which outputs the question sentences registered in the interaction scenario to a user through an output unit 4; a response sentence recognition unit 22 which recognizes a response obtained from the user; and an emotional empathy response generation unit 26 which outputs an emotional empathy response corresponding to the polarity of an emotional expression detected from the response obtained from the user and satisfying the causal relations.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an automated counseling system. [Background technology]

[0002] There is an automatic interactive counseling system that analyzes the user's condition and generates and outputs appropriate responses.

[0003] Patent Document 1 discloses an invention that calculates a user's emotional state from an input sentence entered by the user, updates the emotional state history, and generates a response sentence for an empathetic response that satisfies the emotional aspect by reading and concatenating multiple sets of character strings from a response master based on the updated emotional state history. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2015-014834 A Summary of the Invention [Problem to be solved by the invention]

[0005] Providing advice from the perspective of others requires understanding the cause of the user's emotions. However, the invention described in Patent Document 1 has no means for recognizing the user's emotional state and the cause of the emotion, so it has a problem in that it is difficult to respond in a way that gains the user's sympathy.

[0006] In consideration of the above, an object of the present invention is to provide an automated counseling system that satisfies the cognitive aspects of a user and is capable of responding in a way that gains the user's sympathy. [Means for solving the problem]

[0007] In order to achieve the above-mentioned object, the automated counseling system described in claim 1 includes a storage device having a dialogue scenario memory unit that stores a dialogue scenario in which a plurality of questions and causal relationships between each of the plurality of questions are registered, an output unit that outputs the questions registered in the dialogue scenario to a user, an answer sentence recognition unit that recognizes answers obtained from the user, and an emotional empathy response generation unit that outputs an emotional empathy response according to the polarity of an emotional expression detected from the answer obtained from the user that satisfies the causal relationship.

[0008] According to the automated counseling system of claim 1, by generating and outputting an emotional empathy response according to the polarity of the emotional expression detected from the answer, it is possible to provide a response that satisfies the cognitive aspects of the user and gains the user's empathy. The generation of the emotional empathy response is realized, for example, by using a trained model constructed by machine learning.

[0009] The automated counseling system described in claim 2 further includes a cognitive empathy response generation unit that, when a pair of answers corresponding to a question sentence that satisfies the causal relationship is detected, transforms the pair of answers to generate and output a cognitive empathy response based on the causal relationship, and the emotional empathy response generation unit outputs the emotional empathy response after outputting the cognitive empathy response.

[0010] According to the automated counseling system of claim 2, a pair of answers corresponding to a question sentence that satisfies a causal relationship can be transformed to output a cognitive empathic response that can gain the empathy of a user. Note that the generation of the cognitive empathic response is realized, for example, by using a trained model constructed by machine learning.

[0011] The automated counseling system described in claim 3 further includes an emotional state recognition unit that searches an emotional expression dictionary that corresponds the emotional expressions to degrees of polarity, and extracts the degree of polarity of the emotional expression detected from the answer, and the emotional empathy response generation unit extracts an empathy expression corresponding to the degree of polarity extracted by the emotional state recognition unit from the emotional empathy response dictionary that records empathy expressions corresponding to the degrees of polarity, and outputs it as the emotional empathy response.

[0012] According to the automated counseling system described in claim 3, an emotional empathy response corresponding to the degree of polarity indicated by the user's answer can be extracted and output from an emotional empathy response dictionary which records empathy expressions corresponding to the degree of polarity indicating the positive or negative degree of emotional expression.

[0013] The automated counseling system described in claim 4 further includes a paraphrase expression memory unit in which synonyms of everyday words and definition sentences of words listed in the emotional expression dictionary are respectively stored, and a paraphrase expression generation unit that replaces replaceable expressions included in each of the cognitive empathy response generated by the cognitive empathy response generation unit and the emotional empathy response generated by the emotional empathy response generation unit with at least one of the synonyms and the definition sentences stored in the paraphrase expression memory unit.

[0014] According to the automated counseling system described in claim 4, each of the cognitive empathic responses and the emotional empathic responses can be rephrased using synonyms and definition sentences to understand the user and make the user feel that their feelings are accepted, thereby enhancing the relationship of trust between the automated counseling system and the user and promoting self-disclosure by the user. Effect of the Invention

[0015] As described above, according to the automated counseling system of the present invention, it is possible to obtain an automated counseling system that satisfies the cognitive aspects of the user and is capable of responding in a way that gains the user's sympathy. [Brief description of the drawings]

[0016] [Figure 1] 1 is a block diagram showing an example of a specific configuration of an automated counseling system according to a first embodiment of the present invention. [Diagram 2] 2 is a flowchart showing an example of processing of the automated counseling system according to the first embodiment of the present invention. [Diagram 3] 1 is an example of a dialogue scenario in which a plurality of questions and the causal relationships between each of the questions are registered. [Figure 4] 13A is an example of an emotional expression dictionary in which emotional expressions are associated with polarity scores, and FIG. 13B is an example of an emotional empathy response dictionary in which empathy expressions according to polarity scores are recorded. [Diagram 5] FIG. 11 is a block diagram showing an example of a specific configuration of an automated counseling system according to a second embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017] [First embodiment] Hereinafter, the automated counseling system 100 according to this embodiment will be described with reference to Fig. 1. Fig. 1 is a block diagram showing an example of a specific configuration of the automated counseling system 100 according to this embodiment. The automated counseling system 100 is a computer including a storage device 1 such as an HDD (Hard Disk Drive), a processing device 2 having an arithmetic device such as a CPU (Central Processing Unit), an input unit 3 such as a microphone or a keyboard, and an output unit 4 such as a speaker or a display.

[0018] The storage device 1 includes a dialogue scenario storage unit 11 that stores dialogue scenarios for mental care such as cognitive behavioral therapy. As shown in Fig. 3, the dialogue scenarios stored in the dialogue scenario storage unit 11 are registered with a plurality of questions and causal relationships between each of the questions.

[0019] The processing device 2 includes a question generation unit 21 that presents questions stored in the dialogue scenario storage unit 11 to the user through the output unit 4 in the number (sequence) shown in FIG. 3, an answer recognition unit 22 that recognizes answers from the user input through the input unit 3 and converts them into text data, and a cognitive empathy response generation unit 24 that temporarily stores the user's answers converted into text data by the answer recognition unit 22 and generates the following answer 1 and answer 2 corresponding to the question 1 and question 2 that satisfy the causal relationship registered in the dialogue scenario storage unit 11. the causal relationship recognition unit 23 that outputs a pair of answer 1 and answer 2 to the user, a cognitive empathy response generation unit 24 that transforms the pair of answer 1 and answer 2 output from the causal relationship recognition unit 23 to generate a cognitive empathy response based on the causal relationship, an emotional state recognition unit 25 that determines whether or not an expression indicating emotion is present in the user's answer that has been converted into text data by the answer sentence recognition unit 22, and the polarity of the emotion, whether positive or negative, and an emotional empathy response generation unit 26 that generates an emotional empathy response in accordance with the emotional expression determined by the emotional state recognition unit 25 and the polarity of the emotional expression.

[0020] With the above-described configuration, the processing device 2 generates a cognitive empathic response that takes into consideration the cognitive aspect and an emotional empathic response that takes into consideration the emotional aspect, as described below. (1) The question generator 21 selects the following question 1 from the dialogue scenario storage unit 11 and outputs it: Question 1: What do you find difficult? (2) User response to question 1 Answer 1: "I can't teach my juniors how to do their jobs well." (3) The question generator 21 selects question 2, which is causally related to question 1, from the dialogue scenario memory 11 and outputs it. Question 2: How did you feel at that time? (4) User response to question 2 Answer 2: "It makes me hate myself." (5) The cognitive empathy response generation unit 24 transforms Answer 1 and Answer 2 to output a cognitive empathy response. Cognitive empathy response: "I see. So you're feeling self-loathing because you can't teach your juniors how to do their jobs." (6) The emotional empathy response generation unit 26 generates and outputs an emotional empathy response according to the polarity determined by the emotional state recognition unit 25. Emotional Empathy Response: "That must be hard for you."

[0021] The generation of the cognitive empathic response by the cognitive empathic response generation unit 24 through the transformation of each of the answers 1 and 2, the determination of the polarity of the emotional expression of the answer by the emotional state recognition unit 25, and the generation of the emotional empathic response by the emotional empathic response generation unit 26 are each realized using a trained model. This trained model is constructed by machine learning using an architecture such as LSTM (Long Short-Term Memory) with the question sentence, the answer sentence, the polarity of the emotional expression contained in the answer sentence, the cognitive empathic response, and the emotional empathic response prepared in advance as samples as training data.

[0022] 2 is a flowchart showing an example of the processing of the automated counseling system 100 according to this embodiment. In step S100, the question generation unit 21 speaks (outputs) the question shown in FIG. 3 stored in the dialogue scenario storage unit 11 from number 1 via the output unit 4. If the question has the structure shown in FIG. 3, "What are you feeling distressed about?" corresponding to number 1 is selected and output.

[0023] In step S102, an answer is obtained from the user via the input unit 3. An example of the answer is, "I can't teach my junior how to do the job well." The obtained answer is converted into text data by the answer sentence recognition unit 22.

[0024] In step S103, the causal relationship recognition unit 23 judges whether an answer corresponding to the question sentence having a causal relationship has been acquired. Specifically, the causal relationship recognition unit 23 temporarily stores the answer from the user converted into text data, and judges whether a pair of answers to the question sentence satisfying the causal relationship registered in the dialogue scenario storage unit 11 has been acquired. For example, at the stage when the question 1 is uttered, a pair of answers satisfying the causal relationship is not acquired, so the procedure proceeds to step S105. If a pair of answers satisfying the causal relationship is acquired in step S103, the procedure proceeds to step S104. Details of the process in step S104 will be described later.

[0025] In step S105, the emotional state recognition unit 25 judges whether the answer sentence of the user converted into text data contains an expression expressing an emotion, and the polarity of the emotion (positive or negative). In step S105, if the answer sentence contains an emotional expression, the procedure proceeds to step S106. In step S105, if the answer to question 1 does not contain any particular emotional expression, such as "I can't teach my juniors how to do their jobs well," the procedure proceeds to step S107. The process in step S106 will be described in detail later.

[0026] In step S107, the question generation unit 21 determines whether or not there are any unuttered questions remaining in the dialogue scenario storage unit 11. When only question 1 has been uttered, questions 2 and onward have not been uttered, so the procedure proceeds to step S101.

[0027] When the procedure is shifted from step S107, in step S101, the question generator 21 utters, for example, question 2 stored in the dialogue scenario storage unit 11, "How did you feel at that time?".

[0028] In step S102, an answer is obtained from the user via the input unit 3. An example of the answer is "It makes me hate myself." The obtained answer is converted into text data by the answer sentence recognition unit 22.

[0029] In step S103, the causal relation recognition unit 23 judges whether or not an answer having a causal relation has been acquired. For example, at the stage where the question 2 is uttered, the pair of the question 1 and the question 2 that satisfy the causal relation shown in Fig. 3 is completed, so the causal relation recognition unit 23 makes a positive judgment, outputs the answer pair of the question 1 and the question 2, Answer 1 "I can't teach my juniors how to do their jobs well" to the cognitive empathy response generation unit 24, and shifts the procedure to step S104.

[0030] In step S104, the cognitive empathy response generating unit 24 converts the pair of answers output from the causal relationship recognizing unit 23, answer 1 "I can't teach my juniors how to do their jobs well" to answer 2 "It makes me hate myself", into an expression expressing a causal relationship. Specifically, answer 1 and answer 2 are transformed to generate a cognitive empathy response. For example, answer 1 "I can't teach my juniors how to do their jobs well" is transformed to "I see. You can't 'teach your juniors how to do their jobs well'", and answer 2 "It makes you hate yourself" is transformed to "You're hating yourself, aren't you?". In the transformation of answer 1, "You can't teach your juniors how to do their jobs well" is quoted from answer 1, and in the transformation of answer 2, "It makes you hate yourself" is quoted from answer 2. Then, the cognitive empathy response "I see. You can't teach your juniors how to do their jobs well, so you're hating yourself" is generated and spoken by combining the transformed answers 1 and 2.

[0031] In step S105, the emotional state recognition unit 25 judges whether or not there is an expression expressing an emotion in the answer sentence of the user that has been converted into text data, and the polarity of the emotion. Specifically, the emotional state recognition unit 25 searches an emotional expression dictionary (FIG. 4(A)) that associates an emotional expression with a polarity score, which is stored in the emotional state recognition unit 25, for the wording included in the answer sentence, and makes a positive judgment if it can extract a polarity score associated with an emotional expression hit in the search. For example, if answer 1 is "I can't teach my junior how to do the job well" and answer 2 is "It makes me hate myself," answer 2 contains "self-hatred" stored in the emotional expression dictionary, so in step S105, the emotional state recognition unit 25 extracts a polarity score of "-0.3" and moves the procedure to step S106.

[0032] In step S106, the emotional empathy response generation unit 26 generates and speaks an emotional empathy response according to the emotional expression output from the emotional state recognition unit 25 and the polarity score corresponding to the emotional expression. Specifically, the emotional empathy response generation unit 26 searches an emotional empathy response dictionary previously stored in the emotional empathy response generation unit 26 as shown in FIG. 4(B), extracts an empathy expression corresponding to the polarity score output from the emotional state recognition unit 25, and speaks it. For example, when the polarity score output from the emotional state recognition unit 25 is "-0.3", the emotional empathy response generation unit 26 speaks "That must be hard." As shown in FIG. 4(B), the emotional empathy response dictionary records empathy expressions according to polarity scores indicating the degree of polarity.

[0033] In step S107, the question generation unit 21 determines whether or not there are any unuttered questions remaining in the dialogue scenario storage unit 11. In this embodiment, the procedures of steps S101 to S106 are repeated until all questions stored in the dialogue scenario storage unit 11 are uttered. For example, in the dialogue scenario shown in Fig. 3, if up to question 8 is uttered in step S101, a cognitive empathy response is uttered in step S104, and an emotional empathy response is uttered in step S106, a negative determination is made in step S107 and the process ends.

[0034] As described above, according to this embodiment, pairs of answer sentences having a causal relationship are extracted from multiple answer sentences of a user obtained in chronological order, and an emotional empathy response according to the polarity of the emotional expression detected from the answer sentences is generated and output, thereby making it possible to provide a response that satisfies the cognitive aspects of the user and gains the user's empathy.

[0035] In this embodiment, the emotional expression and the polarity of the emotional expression are extracted by searching a preset emotional expression dictionary with the wording included in the user's reply. Then, the empathy expression corresponding to the extracted polarity score is extracted from a preset emotional empathy response dictionary and output. Through this process, this embodiment can satisfy the cognitive aspect of the user, who wants to think about things from the perspective of others.

[0036] [Second embodiment] Next, a second embodiment of the present invention will be described with reference to Fig. 5. Fig. 5 is a block diagram showing an example of a specific configuration of an automated counseling system 200 according to this embodiment. This embodiment differs from the first embodiment in that the storage device 1 includes a paraphrase expression storage unit 12 and the processing device 2 includes a paraphrase expression generation unit 27. However, other configurations are the same as those of the first embodiment, and therefore the same components as those of the first embodiment are denoted by the same reference numerals as those of the first embodiment and detailed description thereof will be omitted.

[0037] The paraphrase expression storage unit 12 stores synonyms of everyday words and definitions of phrases listed in the emotion expression dictionary. For example, a synonym of "teaching work" is "guidance," and the definition of the phrase "self-loathing" listed in the emotion expression dictionary is "hate yourself."

[0038] The paraphrase expression generation unit 27 replaces expressions that can be replaced with synonyms or definition sentences of emotional expressions written in the emotional expression dictionary, among the phrases generated by the cognitive empathy response generation unit 24 and the emotional empathy response generation unit 26, with the synonyms or definition sentences. Specifically, when the cognitive empathy response generated in step S104 of FIG. 2 is "I see. You're feeling self-loathing because you can't teach your juniors how to do their jobs," the following paraphrase is generated. "I see, so you're not able to properly guide your juniors and you're starting to hate yourself."

[0039] In this embodiment, by restating the information as described above, the user feels that they are better understood and that their feelings are accepted, thereby enhancing the relationship of trust between the automated counseling system 200 and the user and encouraging self-disclosure by the user. [Explanation of symbols]

[0040] 1 Storage device 2 Processing equipment 3. Input section 4 Output section 8. Questions 11 Dialogue scenario storage unit 12 Paraphrase expression memory section 21 Question generation section 22 Answer sentence recognition unit 23 Causal Recognition Department 24 Cognitive Empathy Response Generation Unit 25 Emotional state recognition unit 26 Emotional Empathy Response Generation Unit 27 Paraphrase Generation Unit 100 Automated Counseling System 200 Automated Counseling System

Claims

1. a storage device having a dialogue scenario storage unit that stores a dialogue scenario in which a plurality of questions and causal relationships between each of the plurality of questions are registered; an output unit that outputs a question sentence registered in the dialogue scenario to a user; an answer sentence recognition unit that recognizes an answer obtained from the user; an emotional empathy response generation unit that outputs an emotional empathy response according to a polarity of an emotional expression detected from the answer that satisfies the causal relationship obtained from the user; An automated counseling system including:

2. a cognitive empathy response generation unit that, when a pair of answers corresponding to a question sentence that satisfies the causal relationship is detected, transforms the pair of answers to generate and output a cognitive empathy response based on the causal relationship; The automated counseling system according to claim 1 , wherein the emotional empathy response generation unit outputs the emotional empathy response after outputting the cognitive empathy response.

3. The system further includes an emotional state recognition unit that searches an emotional expression dictionary in which the emotional expressions are associated with degrees of polarity, and extracts a degree of polarity of the emotional expression detected from the answer, The automated counseling system of claim 2, wherein the emotional empathy response generation unit extracts an empathy expression corresponding to the degree of polarity extracted by the emotional state recognition unit from an emotional empathy response dictionary that records empathy expressions corresponding to the degree of polarity, and outputs the extracted empathy expression as the emotional empathy response.

4. a paraphrase expression storage unit storing synonyms of everyday words and definitions of words written in the emotion expression dictionary; a paraphrase expression generation unit that replaces replaceable expressions included in the cognitive empathy response generated by the cognitive empathy response generation unit and the emotional empathy response generated by the emotional empathy response generation unit with at least one of the synonyms and the definition sentences stored in the paraphrase expression storage unit; The automated counseling system of claim 3 further comprising:

Images