JP2022146699A - Program, device, and method for agent performing dialog using a plurality of characters - Google Patents

Abstract

To provide a program, device, and method for an agent that performs a dialog properly, by using a plurality of characters in accordance with an utterance content of a user.SOLUTION: A program is configured to make a computer function as: dialog control means that inputs an utterance sentence of a user, and outputs a response sentence of a character; character personality word registration means that registers a personality word for each emotional polarity as to each character; main character setting means that selects a main character performing the dialog with the user, and sets the main character to the dialog control means; utterance emotion estimation means that extracts the emotional polarity and personality word from the utterance sentence of the user; sympathy determination means that determines whether the emotional polarity and personality word of the user are included in the emotional polarity and personality word of the main character; sub character selection means that, when it is determined that the emotional polarity and personality word of the user are not included, selects a sub character in which the emotional polarity and personality word of the user are included, and sets the sub character to the dialog control means.SELECTED DRAWING: Figure 2

Description

The present invention relates to technology of an agent interaction system that interacts with a user.

Agents that interact naturally with users using terminals such as smartphones, tablets, and AI (Artificial Intelligence) speakers are widely used. An agent is a function in which a character such as computer graphics or artificial voice interacts with a user through text or voice. During the conversation, some smartphone displays show the agent's appearance (chatbot type). The agent progresses the dialogue according to the current situation, tastes and preferences, and feelings as a dialogue partner from the user's point of view.
Agents have personalities (age, gender, etc.) in their characters as well as user profiles. A user can have a strong sense of familiarity by interacting with a character who has a personality with which he or she can sympathize.

Conventionally, there is also a technology of text-based dialogue with a user through a posting system of SNS (Social Networking Service) (see, for example, Non-Patent Document 1). According to this technology, different posting systems are prepared for each emotional polarity (positive, negative, neutral), and a posting system having the same emotional polarity as the user's utterance content responds. For example, when the user utters, "I don't like it snowing today," the content of the utterance is determined to have an emotional polarity of "negative." In this case, a posting system with a negative emotional polarity responds, for example, "It sure is snowing and I'm tired." In this way, by outputting a response sentence based on the same feeling as the user, it is possible to empathize with the user and improve the degree of interaction satisfaction.

Shunichi Tahara et al., "Proposal and evaluation of an empathy dialogue system focusing on the polarity of web news-related tweets," Proceedings of the 80th National Conference 2018.1 (2018): 29-30, [online], [3rd year of Reiwa Searched on the 14th of the month], Internet <URL: https://ipsj.ixsq.nii.ac.jp/ej/?action=repository_action_common_download&item_id=187867&item_no=1&attribute_id=1&file_no=1> AI and Machine Learning Products, “Sentiment Analysis”, [online], [searched on March 14, 2021], Internet <URL: https://cloud.google.com/natural-language/docs/analyzing-sentiment ?hl=en＞ Shinya Ishida et al., "Generation of Listener Responses to Encourage Utterance for Active Listening Dialogue Systems," SIG-SLUD, 5(01), 1-6.(2016), [online], [March 14, 2021 Search], Internet <URL: https://jsai.ixsq.nii.ac.jp/ej/index.php?action=pages_view_main&active_action=repository_action_common_download&item_id=1147&item_no=1&attribute_id=1&file_no=1&page_id=13&block_id=23>

However, when the user is interacting with a character having a specific personality, it may happen that the character's personality does not match. If the user receives a response from the character that does not sympathize with the user, it is considered that the user feels uncomfortable with the dialogue.
For example, assume that the user and the "game-loving" character have the following dialogue.
(Character): Do you like games?
(User) : I guess I don't really like games because they're bad for my eyes.
(charactor):·····
Since the character is composed of a dialogue scenario based on the individuality of liking games, it is difficult to output a response sentence that can sympathize with the user's utterance. A character with a game-loving personality cannot output a response sentence that can empathize with a user who does not like games.

Of course, the character can also respond as follows, even though he likes the game.
(Character): It's true that games are bad for your eyes, so I don't like them either.
However, the dialogue is inconsistent with the character's preset personality. Ultimately, in the dialogue scenario of the character, the dialogue breaks down, and the user feels uncomfortable with the character.

In response to this, the inventors of the present application considered whether it would be possible to respond to the user with a plurality of characters. In other words, it was thought that if an appropriate character could respond according to the user's utterance content, the dialogue would not collapse from the user's point of view.

SUMMARY OF THE INVENTION Accordingly, it is an object of the present invention to provide a program, apparatus and method for an agent that interacts with a plurality of appropriate characters according to the contents of a user's utterance.

According to the present invention, a program for causing a computer to function as an agent having interaction control means for inputting a user's utterance sentence and outputting a character's response sentence,
Character individual word registration means for registering individual words for each character for each emotional polarity;
main character selection means for selecting a main character to interact with the user and setting the main character as interaction control means;
an utterance emotion estimating means for extracting emotion polarity and individual words from a user's utterance;
empathy determination means for determining whether or not the user's emotional polarity and individuality words are included in the main character's emotional polarity and individuality words;
When the sympathy determination means determines false, the computer functions as a secondary character selection means for selecting a secondary character containing the user's emotional polarity and individual words and setting the secondary character as the dialogue control means. and

According to another embodiment of the program of the present invention,
Emotional polarity is classified into positive and negative,
It is also preferred to have the computer function so that each individual word is associated with either a positive or a negative.

According to another embodiment of the program of the present invention,
The speech emotion estimation means extracts negative and individual words from the speech sentences of the user who interacts with the main character,
The sympathy determination means determines whether or not the negative and individual words of the user are included in the negative and individual words of the main character,
It is also preferred that the secondary character selection means causes the computer to select secondary characters that include the user's negative and personal words.

According to another embodiment of the program of the present invention,
The speech emotion estimation means extracts positive and individual words from the speech sentences of the user who interacts with the main character,
The sympathy determination means determines whether or not the positive and individual words of the user are included in the positive and individual words of the main character,
It is also preferred that the secondary character selection means cause the computer to select secondary characters in which the user's positive and personal words are included.

According to another embodiment of the program of the present invention,
For each user, it functions as a user's individual word registration means for registering the emotional polarity and individual words extracted from the user's uttered sentence by the uttered emotion estimation means,
The sub-character selection means selects a sub-character including an emotion polarity and a personality word identical or similar to the user's emotion polarity and personality word registered in the user's personality word registration means when the sympathy determination means determines false. It is also preferable to have the computer function

According to another embodiment of the program of the present invention,
For each user, it functions as a user's individual word registration means for registering the emotional polarity and individual words extracted from the user's uttered sentence by the uttered emotion estimation means,
It is also preferable that the main character selection means causes the computer to select a main character including the same or similar emotional polarity and personality word as the user's emotional polarity and personality word registered in the user's personality word registration means.

According to another embodiment of the program of the present invention,
Posted message collection means for collecting a group of posted messages posted in the past from a user account of a user from an SNS (Social Networking Service) server;
It is also preferable to make the computer function as posted emotion estimation means for extracting characteristic emotional polarities and individual words from a group of user's posted sentences and registering them in the user's individual word registration means.

According to another embodiment of the program of the present invention,
It is also preferred that the dialogue control means cause the main character and the sub-character to function simultaneously, and to function the computer to establish dialogue between the user and the main character and the sub-character.

According to another embodiment of the program of the present invention,
It is also preferable that the dialogue control means causes the computer to output to the user a spoken sentence introducing the sub-character when the sub-character participates from the main character.

According to the present invention, a dialogue apparatus having dialogue control means for inputting a user's utterance sentence and outputting a character's response sentence,
Character individual word registration means for registering individual words for each character for each emotional polarity;
main character selection means for selecting a main character to interact with the user and setting the main character as interaction control means;
an utterance emotion estimating means for extracting emotion polarity and individual words from a user's utterance;
empathy determination means for determining whether or not the user's emotional polarity and individuality words are included in the main character's emotional polarity and individuality words;
a sub-character selection means for selecting a sub-character containing the user's emotional polarity and individual words and setting the sub-character to the dialog control means when the sympathy determination means determines false. .

According to the present invention, a dialogue method for a device having dialogue control means for inputting a user's utterance sentence and outputting a character's response sentence, comprising:
The device
having a character individual word registration unit for registering a character individual word for each emotional polarity for each character,
a first step of selecting a main character that interacts with the user and setting the main character as interaction control means;
a second step of extracting emotional polarity and individual words from the user's utterance;
a third step of determining whether or not the user's emotional polarities and individuality words are included in the main character's emotional polarities and individuality words;
and executing a fourth step of selecting a sub-character containing the user's emotional polarity and individual words and setting the sub-character to the dialog control means if the third step is determined to be false. and

According to the agent program, device, and method of the present invention, it is possible to interact using a plurality of appropriate characters according to the content of the user's utterance.

FIG. 4 is a first explanatory diagram showing interaction between a user and a character; 1 is a functional configuration diagram of an interactive device according to the present invention; FIG. FIG. 4 is an explanatory diagram showing registered contents in a user individual word registration unit and a character individual word registration unit; FIG. 10 is an explanatory diagram showing processing of an utterance emotion estimation unit, an empathy determination unit, and a sub-character; FIG. 4 is an explanatory diagram showing a dialog scenario of a dialog control unit; FIG. 11 is a second explanatory diagram showing interaction between a user and a character;

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a first explanatory diagram showing interaction between a user and a character.

According to FIG. 1, an interactive device 1 communicates with a terminal 2 based on user operations and functions as an agent to the user.
A user program is installed in the terminal 2 as an interactive interface with the user. The terminal 2 is, for example, a smart phone or a tablet, and has a display for displaying an agent character. The character received from the interactive device 1 is displayed on the terminal 2, and the response sentence of the character is clearly shown to the user. On the other hand, the character input of the user's utterance sentence is received.
The terminal 2 may also be, for example, an AI speaker (such as Alexa (registered trademark)) and have a microphone and speaker for the user. The terminal 2 can utter the response sentence of the character to the user as voice. In addition, speech recognition allows the user to input an utterance sentence by speaking his/her own voice.

First, a character that interacts with the user presents the user with topics that match their personality. Depending on the user's utterance sentence to it, it can progress according to a dialogue scenario.
However, the dialogue scenario of each character corresponds to the individuality of each character, and it is not possible to write a response sentence that can always sympathize with the user's utterance sentence.

According to FIG. 1(a), the user @a who operates the terminal 2 and the character B who is an "OL who likes make-up and buds" are interacting as follows.
Character B: "Do you play badminton?"
User@a: "Yeah, I play badminton a lot."
Character B: "Yes, I love you too!"
User@a: "I also like cleaning."
Character B: "..."
Here, if character B has a negative personality for "cleaning", no positive response sentence for "cleaning" is described in the dialogue scenario.

According to FIG. 1(b), the user @b who operates the terminal 2 and the character C, who is a "salaryman who likes to clean", have a conversation as follows.
Character C: "Do you like cleaning?"
User@b: "Cleaning is troublesome."
Character C: "..."
Here, when the character C has a positive personality for "cleaning", no negative response sentence for "cleaning" is described in the dialogue scenario.

FIG. 2 is a functional configuration diagram of the interactive device in the present invention.

According to FIG. 2, a dialogue device 1 executing an agent is connected via a network to a terminal 2 executing a user program.
The dialogue device 1 inputs a user's utterance sentence from the terminal 2 and outputs a response sentence corresponding to the character. The dialogue device 1 allows characters to interact with a user as naturally as possible.

According to FIG. 2, the dialogue device 1 includes a dialogue control unit 10, a user individual word registration unit 101, a character individual word registration unit 102, a main character selection unit 11, an utterance emotion estimation unit 12, and an empathy determination unit. 13 , a secondary character selection unit 14 , a posted message collection unit 15 , a posted emotion estimation unit 16 , and a dialogue interface 17 . These functional components are implemented by executing a program that causes a computer installed in the device to function. In addition, the processing flow of these functional components can also be understood as an interaction method of the device.

FIG. 3 is an explanatory diagram showing registered contents in the user individual word registration unit and the character individual word registration unit.

[User Personality Word Registration Unit 101]
The user's personality word registration unit 101 registers a personality word for each emotional polarity of the user.
Emotional polarities are classified into, for example, positive and negative, and each individual word is associated with either positive or negative.
According to FIG. 3, in the user personality word registration unit 101, the following personality words are registered according to the personality of the user @a.
User@a
Positive: badminton, mont blanc, cleaning
Negative: commuting, studying, walking

[Character individual word registration unit 102]
The character individual word registration unit 102 registers individual words for each emotional polarity for each character.
According to FIG. 3, in the character individual word registration unit 102, the following individual words are registered according to the character's individuality.
Character A (20s who likes games)
Positive: games, videos, walks
Negative: work, badminton, cleaning Character B (OL who likes make-up and bud)
Positive: sweets, badminton, makeup
Negative: commuting, studying, cleaning Character C (salaryman who likes cleaning)
Positive: work, computer, cleaning
Negative: videos, games, walks

[Main character selection unit 11]
The main character selection unit 11 selects a “main character” that interacts with the user from among a plurality of characters, and sets the main character in the interaction control unit 10 .
At least when a user launches an interactive application, the main character is initially selected.

The main character selection unit 11 selects a “main character” that includes emotional polarities and individual words that are identical to and similar to the user's emotional polarities and individual words registered in the user's individual word registration unit 101 .
For example, for only the emotional polarity "positive", a character including a personality word identical or similar to the user's personality word may be selected. Of course, for both of the emotional polarities "positive" and "negative," the character that contains the greatest number of individual words identical or similar to the user's individual word may be selected.

According to FIG. 3, the rate of matching between individual word groups in the user individual word registration unit 101 and the character individual word registration unit 102 is shown. Among them, the character with the highest average match rate is selected as the main character.
Comparison between user @a and character A
Positive match rate: 0.0%
Negative match rate: 33.3% (matched with "walking")
Average match rate: 16.7%
Comparison between user@a and character B
Positive match rate: 33.3% (matched on "Badminton")
Negative match rate: 66.6% (matched for "commuting" and "study")
Average match rate: 50.0%
Comparison of user@a and character C
Positive match rate: 33.3% (matched on "Badminton")
Negative match rate: 33.3% (matched with "walking")
Average match rate: 33.3%
In this case, character B with the highest average match rate is selected for user @a.

As another embodiment, the similarity between individual word groups can be used for comparison instead of the match rate between individual words. For example, individual words are vectorized by Word2vec (registered trademark), and determination is made based on whether the vectors are similar within a predetermined range. "Word2vec" is a technology that converts words into vectors and compresses the dimensions in order to capture the meaning and grammar of words. Of course, it is not limited to Word2vec as long as it can be converted into a feature vector obtained by analyzing the part of speech or meaning of each word.

According to FIG. 3, for example, the positive individual word "Mont Blanc" of the user @a and the positive individual word "sweets" of the character B are similar as vectors within a predetermined range and are determined to match. can also

[Utterance emotion estimation unit 12]
The uttered emotion estimation unit 12 extracts the emotion polarity and individual words from the user's utterance text input from the dialogue control unit 10 . Then, the speech emotion estimation unit 12 outputs the emotion polarity and the individual words to the empathy determination unit 13 .

The speech emotion estimation unit 12 estimates emotion polarity and individual words using a learning engine. The "emotional polarity" may be classified into, for example, "positive", "negative", and "neutral", or may be classified into four or more.
Specifically, the emotional polarity can be determined using a determiner such as Support Vector Machine or Cloud Natural Language API provided by Google (registered trademark) (see, for example, Non-Patent Document 2).
Specifically, CRFs (Conditional Random Fields) can be used for individual words (see, for example, Non-Patent Document 1). According to this technique, it is possible to extract individual words that are important words from a user's utterance using a model that has learned a plurality of feature quantities such as parts of speech and the order in which the parts of speech appear.

FIG. 4 is an explanatory diagram showing the processing of the speech emotion estimating section, the sympathy determining section, and the sub-character.

As shown in FIG. 1(a) described above, the emotional polarity "positive" and the individual word "cleaning" are estimated for the user @a's utterance sentence "I also like cleaning."
Also, as shown in FIG. 1B, the emotion polarity "negative" and the individual word "cleaning" are estimated for the uttered sentence "Cleaning is troublesome" of the user @b.

As another embodiment, it is also preferable that the speech emotion estimation unit 12 registers the estimated emotional polarity and individual words in the user individual word registration unit 101 . Emotional polarities and individual words are cumulatively registered in the user's individual word registration unit 101 . For example, the user's personality word registration unit 101 registers the emotional polarity "positive" and the personality word "cleaning" for the user @a, and registers the emotional polarity "negative" and the personality word "cleaning" for the user @b. be.
As a result, the main character selected by the main character selection unit 11 is also updated after the predetermined period of time has elapsed or after the application is restarted.

[Empathy determination unit 13]
The sympathy determination unit 13 determines whether or not the user's emotional polarities and individual words output from the uttered emotion estimating unit 12 are included in the main character's emotional polarities and individual words. If it is determined to be “false”, that effect is output to the sub-character selection unit 14 .

The sympathy determination unit 13 makes a determination according to the emotional polarity “positive” or “negative”. Specifically, when the spoken emotion estimation unit 12 estimates the negative of the user, it is determined whether or not the individual word is included in the negative of the main character. Conversely, when the speech emotion estimation unit 12 estimates that the user is positive, it is determined whether or not the individual word is included in the positive of the main character.

For example, according to FIGS. 1(a), 3, and 4, when the emotional polarity "positive" and the personality word "cleaning" are estimated from the utterance sentence of the user @a, the personality of the main character B is "positive." Determine if the word "cleaning" is included.
Further, for example, according to FIG. 1B, when the emotional polarity "negative" and the individual word "cleaning" are estimated from the utterance sentence of the user @b, the individual word "cleaning" is added to the main character C's "negative". Determine whether it contains
Of course, when comparing individual words, if the vectors of the individual words are within a predetermined distance, it may be determined that they are similar.

[Secondary character selection unit 14]
When the sympathy determination unit 13 determines "false", the sub-character selection unit 14 selects a sub-character whose personality word is included in the user's emotional polarity (positive or negative).

For example, according to FIG. 1(a) described above, the secondary character C including the emotional polarity "positive" and the individual word "cleaning" estimated from the utterance sentence of the user @a is selected.
For example, according to FIG. 1(b) described above, the sub-character A including the emotional polarity "negative" and the individual word "cleaning" estimated from the utterance sentence of the user @b is selected.

In another embodiment, the sub-character selection unit 14 selects a sub-character including an emotional polarity and a personality word identical or similar to the user's emotional polarity and personality word registered in the user's personality word registration unit 101. good too. That is, instead of simply selecting a sub-character containing a personality word, all personality words included in the emotion polarity may be vectorized and compared based on the distance between the vectors.

For example, according to FIG. 1A described above, for each character, the vector product of the individual word group included in the emotional polarity "positive" is compared with the vector of the individual word "cleaning" of the user @a, and the cosine similarity is obtained. Character C with the highest degree is selected.
For example, according to FIG. 1B described above, for each character, the vector product of the individual word group included in the emotional polarity "negative" is compared with the vector of the individual word "cleaning" of the user @b, and the cosine similarity is obtained. Character A with the highest degree is selected.
S(a,b)=cos θ=(Va·Vb)/(|Va||Vb|)
S(a,b): cosine similarity
Va: vector of individual words in the user's utterance
Vb: Character's individual word vector The cosine similarity S(a,b) is a value between 0 and 1, and approaches 1 as the similarity increases.

[Dialogue control unit 10]
The dialog control unit 10 inputs a user's utterance sentence and outputs a character's response sentence.
In addition to the "main character", the "secondary character" can also be set in the dialogue control section 10. FIG. That is, the dialogue control section 10 can cause the "main character" and the "sub-character" to function at the same time, and establish a dialogue between the user and the main character and the sub-character. The main character is always present during the dialogue, and the sub-character interacts with the user in place of the main character when the main character cannot output a response sentence that resonates with the user.

FIG. 5 is an explanatory diagram showing a dialogue scenario of the dialogue control unit.

According to FIG. 5, a unique word (tag) is assigned to each response sentence output from the dialogue control unit 10 to the user.
(Response sentence) (Individual word)
do you like games? -> Game Do you like badminton? -> Badminton Do you like cleaning? -> Cleaning It is important what kind of emotional polarity the user utters in response sentences containing individual words. For this purpose, a response sentence output from the dialogue control unit 10 is registered as a dialogue scenario according to the emotional polarity of the user's uttered sentence.

For example, according to FIG. 5A, in the case of the main character B, when the emotional polarity "positive" and the personality word "badminton" are estimated from the utterance sentence of the user @a, the personality of the main character B becomes "positive". It contains the word "badminton". Therefore, the response is "Yes, I love you too!".

For example, according to FIG. 5B, in the case of the main character B, when the emotional polarity "positive" and the personality word "cleaning" are estimated from the utterance sentence of the user @a, the personality of the main character B is "positive". The word "cleaning" is not included, and the response would normally be "That's right." At this time, if the secondary character C is selected, it is possible to respond with "It feels good to clean, isn't it?".

For example, according to FIG. 5(c), in the case of the main character C, when the emotional polarity "negative" and the personality word "cleaning" are estimated from the utterance sentence of the user @b, the personality of the main character B "negative" Since the word "cleaning" is not included, the response would normally be "Yes". At this time, if the sub-character A is selected, it is possible to respond with "same! I'm not good at it."

It should be noted that a response sentence that has been output once will not be output again unless the user's utterance sentence in response to it is forgotten. Therefore, when most of the response sentences of the main and sub-characters have been output to the user, it is also preferable to call up other characters according to the user's emotional polarity and personality words. It is hoped that this will expand the scope of the dialogue.

[Posted text collection unit 15]
The posted text collection unit 15 collects, from the SNS server 3, a group of posted texts posted in the past from user accounts of users.
The SNS server publishes messages posted by an unspecified number of users, and the interactive device 1 can access the SNS server 3 via the Internet. For example, suppose that the following posted sentences are collected.
User@a
"I'm going to play badminton tomorrow, but I'm looking forward to it."
"I'm not good at cleaning"
"If you have time to take a walk, do something else."

[Posted emotion estimation unit 16]
The posted emotion estimation unit 16 extracts characteristic emotional polarities and individual words from the user's posted text group, and registers them in the user individual word registration unit 101 .
As a result, the user's individual word is automatically registered from the user's daily posted sentences without setting by the user himself/herself. Also, a main character who has the same feelings as the user about things is selected from the user's everyday emotional polarities and individual words.
User@a
"I'm going to play badminton tomorrow, but I'm looking forward to it."
->Emotional Polarity "Positive" Personality Word "Badminton"
"I'm not good at cleaning"-> emotional polarity "negative", personality word "cleaning"
"If you have time to take a walk, do something else."
-> Emotional polarity "negative", personality word "walk"
The posted emotion estimation unit 16 may estimate the emotion polarity and the individual word by the same algorithm as the speech emotion estimation unit 12 described above.

As another embodiment, the posted emotion estimating unit 16 may vectorize by Bag of Words in order to extract important individual words from posted comments. Specifically, a user's posted text is broken down into a plurality of words by morphological analysis, and then converted into feature vectors using Bag of Words. "Bag-of-Words" refers to a vector that expresses only the appearance frequency of each word contained in a sentence. The word order is ignored here.

[Dialogue interface 17]
The dialogue interface 17 allows dialogue between the user and the character through a dialogue form. It is preferable to also display an image (character icon) depicting the figure of the character for each character.
The dialog interface 17 displays the character's response sentence as a character balloon to the user. When a user inputs an utterance in text form, an input form is displayed for the user to input.
Further, the dialog interface 17 may output the response sentence of the character as a voice converted by a voice conversion process. Also, when the user inputs a response sentence by voice, a microphone input button is displayed on the display of the terminal 2 . A text obtained by converting the voice uttered by the user by voice recognition processing is acquired as the uttered sentence of the user.

The interactive interface 17 communicates with an application pre-installed in the terminal 2, such as a smartphone or tablet. The interactive form displayed on terminal 2 functions by executing an application.

FIG. 6 is a second explanatory diagram showing interaction between the user and the character.

According to FIG. 6(a), which corresponds to FIG. 1(a), the user @a who operates the terminal 2 and the character B who is an "OL who likes make-up and buds" interact as follows. is proceeding.
Character B: "Do you play badminton?"
User@a: "Yeah, I play badminton a lot."
Character B: "Yes, I love you too!"
User@a: "I also like cleaning."
Character B: "I'll introduce you to someone who suits you."
Character C: "It feels good to clean."
User@a: "That's right."
In this manner, the character C responds to a unique word in which the emotional polarity of the user and the emotional polarity of the character B are different, so that the user can have a dialogue with a response sentence that sympathizes with the user.

According to FIG. 6(b), which corresponds to FIG. 1(b), the user @b who operates the terminal 2 and the character C, who is a "salaryman who likes to clean," have a dialogue as follows. I am letting
Character C: "Do you like cleaning?"
User@b: "Cleaning is troublesome."
Character C: "I'll introduce you to my friend."
Character A: "Same! I'm not good at it."
User@b: "Yeah yeah"
In this way, character A responds to individual words in which the emotional polarity of the user and the emotional polarity of character B are different.

As shown in FIG. 6, the dialogue control section 10 preferably outputs to the user an utterance sentence introducing the sub-character when the sub-character participates from the main character. The sudden switching of characters from the user's point of view is uncomfortable. For example, by replying, "I'll introduce you to someone who suits you," the sudden appearance of the secondary character is prevented.

Also, even if a sub-character participates between the user and the main character, the main character takes the lead in advancing the dialogue with the user. The sub character outputs a response sentence only when the main character cannot output a response sentence that can empathize with the user.

Furthermore, a response sentence (topic) based on a personality word that is as close as possible to the main character and sub-character may be output with respect to the emotional polarity of the user. It is thought that the more the user, the main character, and the sub-character all have the same emotional polarity, the more the user will sympathize with the topic, and the more satisfied the dialogue will be.

Furthermore, if neither the main character nor the sub-character can output an appropriate response sentence to the user's emotional polarity and personality word, the current sub-character is excluded, and another sub-character is selected by the sub-character selection unit 14 again. may be selected. However, if the content of the user's utterance simply does not match the individuality of the character and the characters are switched frequently, the dialogue will no longer be consistent from the user's point of view and will only cause discomfort. For this reason, after the sub-character is selected, re-selection of the sub-character is not executed until dialogue is exchanged a certain number of times or more.
In addition, when the secondary character is switched, the main character will say something like "I see. That's right. ○○ (the current secondary character) seems to be going home on an urgent business. Instead, he will call someone else who is thinking the same thing as you. It is preferable to output an intermediary response sentence such as, for example,

As described in detail above, according to the agent program, device and method of the present invention, it is possible to interact appropriately using a plurality of characters according to the content of the user's utterance.

According to the present invention, when the main character cannot output a response sentence with which the main character can sympathize with the user's emotional polarity and individual words, a secondary character who can sympathize with the emotional polarity can participate in the dialogue. In this way, a dialogue that the user can sympathize with can be realized by having a character who has the same emotion as the user's uttered sentence participate.

For the various embodiments of the present invention described above, various changes, modifications and omissions within the spirit and scope of the present invention can be easily made by those skilled in the art. The foregoing description is exemplary only and is not intended to be limiting. The invention is to be limited only as limited by the claims and the equivalents thereof.

1 Dialogue Device 10 Dialogue Control Unit 101 User Personality Word Registration Unit 102 Character Personality Word Registration Unit 11 Main Character Selection Unit 12 Spoken Emotion Estimation Unit 13 Empathy Determination Unit 14 Secondary Character Selection Unit 15 Posted Sentence Collection Unit 16 Posted Emotion Estimation Unit 17 Dialogue Interface 2 Terminal 3 SNS server

Claims (11)

A program that causes a computer to function as an agent having dialogue control means for inputting a user's utterance sentence and outputting a character's response sentence,
Character individual word registration means for registering individual words for each character for each emotional polarity;
main character selection means for selecting a main character to interact with the user and setting the main character as interaction control means;
an utterance emotion estimating means for extracting emotion polarity and individual words from a user's utterance;
empathy determination means for determining whether or not the user's emotional polarity and individuality words are included in the main character's emotional polarity and individuality words;
When the sympathy determination means determines false, the computer functions as a secondary character selection means for selecting a secondary character containing the user's emotional polarity and individual words and setting the secondary character as the dialogue control means. A program to Emotional polarity is classified into positive and negative,
2. The program according to claim 1, causing a computer to function so that each individual word is associated with either positive or negative. The speech emotion estimation means extracts negative and individual words from the speech sentences of the user who interacts with the main character,
The sympathy determination means determines whether or not the negative and individual words of the user are included in the negative and individual words of the main character,
3. The program according to claim 2, wherein the sub-character selection means causes the computer to select sub-characters containing negative and personal words of the user. The speech emotion estimation means extracts positive and individual words from the speech sentences of the user who interacts with the main character,
The sympathy determination means determines whether or not the positive and individual words of the user are included in the positive and individual words of the main character,
4. The program according to claim 2 or 3, wherein the sub-character selection means causes the computer to select sub-characters containing positive and personal words of the user. For each user, it functions as a user's individual word registration means for registering the emotional polarity and individual words extracted from the user's uttered sentence by the uttered emotion estimation means,
The sub-character selection means selects a sub-character including an emotion polarity and a personality word identical or similar to the user's emotion polarity and personality word registered in the user's personality word registration means when the sympathy determination means determines false. 5. The program according to any one of claims 1 to 4, which causes a computer to function. For each user, it functions as a user's individual word registration means for registering the emotional polarity and individual words extracted from the user's uttered sentence by the uttered emotion estimation means,
The main character selecting means causes the computer to select a main character including the same or similar emotional polarity and personality word as the user's emotional polarity and personality word registered in the user's personality word registration means. Item 5. The program according to any one of items 1 to 4. Posted message collection means for collecting a group of posted messages posted in the past from a user account of a user from an SNS (Social Networking Service) server;
7. The program according to claim 6, causing a computer to function as posted emotion estimation means for extracting characteristic emotional polarities and individual words from a group of user's posted sentences and registering them in user's individual word registration means. 8. The dialogue control means causes the computer to function so as to cause the main character and the sub-character to function simultaneously, and establish dialogue between the user and the main character and the sub-character. The program described in Section. 8. The dialogue control means according to any one of claims 1 to 7, wherein when the main character makes the sub-character participate, the dialogue control means causes the computer to output to the user an utterance sentence introducing the sub-character. program as described. A dialogue apparatus having dialogue control means for inputting a user's utterance sentence and outputting a character's response sentence,
Character individual word registration means for registering individual words for each character for each emotional polarity;
main character selection means for selecting a main character to interact with the user and setting the main character as interaction control means;
an utterance emotion estimating means for extracting emotion polarity and individual words from a user's utterance;
empathy determination means for determining whether or not the user's emotional polarity and individuality words are included in the main character's emotional polarity and individuality words;
a sub-character selection means for selecting a sub-character containing the user's emotional polarity and individual words and setting the sub-character to the dialog control means when the sympathy determination means determines false. interactive device. A dialogue method for a device having dialogue control means for inputting a user's utterance sentence and outputting a character's response sentence,
The device
having a character individual word registration unit for registering a character individual word for each emotional polarity for each character,
a first step of selecting a main character that interacts with the user and setting the main character as interaction control means;
a second step of extracting emotional polarity and individual words from the user's utterance;
a third step of determining whether or not the user's emotional polarities and individuality words are included in the main character's emotional polarities and individuality words;
and executing a fourth step of selecting a sub-character containing the user's emotional polarity and individual words and setting the sub-character to the dialog control means if the third step is determined to be false. How to interact with.

Images