JP2005339368A - Emotion grasping system and emotion grasping method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an emotion grasping system capable of properly grasping user emotions and taking into account the transition of user emotions.
SOLUTION: Service usage observation means (user task investigation unit) 1 for determining and observing whether or not the user is executing another task, and the user using the service usage observation means 1 If it is determined that it is not executed, it is determined that the user is executing another task in the interaction generating means 2 for generating a predetermined action for interacting with the user and the service use observation means 1 In such a case, the user's profile (profile) and context (through the interaction with the user by the interaction generating unit 2 and the autonomous behavior generating unit 3 that generates a predetermined behavior unrelated to the user's input itself) User model generation means 7 for generating a user model formed by context) The
[Selection] Figure 2

Description

  The present invention relates to an emotion grasping system and an emotion grasping method for interacting with a user and acquiring user information therein.

  The current Web service introduces technology that changes the display order and display contents by registering the user's gender, age, and interest, and prevents the user from doing unnecessary work. Some are. For these services, the service provider independently obtains registration information for each service, and displays information on the corresponding field from them (news, horoscope, etc.). Furthermore, using Cookie, it is possible to automatically recognize the fact that it has been registered once, and to use the previous registration information without user input. By storing information that has been previously purchased, grouping people who are interested in them, and performing collaborative filtering, information and products that are presumed to be of interest to the user without registering the information in particular. Services such as presenting (http://www.amazon.co.jp) are provided.

In addition, there are many services that acquire a context for use by a user and filter data provided from a lot of information. For example, when searching for information using a mobile phone, a service is provided in which the current position is acquired using GPS (Global Positioning System) and the like, and peripheral information is presented using that information.
A service that recommends, searches, etc. according to the emotion of the user is also available as a direct focus on emotion (see Non-Patent Document 1 and Non-Patent Document 2).

In Non-Patent Document 1, it is determined by a psychological test how the user is currently feeling, and movies belonging to two types of categories, those that promote the mood and those that are eliminated, are recommended for the mood. . These movies are classified in advance for each category.
In Non-Patent Document 2, eight emotional states (worried, tired, laid-back, careful, hostility, vitality, dear, surprised) are expressed with four levels of strength, and each person who reads the book expresses its degree. The average value is designated as the evaluation. When searching for books by mood, there are 16 types of moods (I want to be excited, I want to be happy, I want to be funny, I want to feel sad, I want to crush, I want to be thrilled, I want to feel relaxed, I want to feel excited, I want to feel thrill, I want to feel thrill, , Touched, want to feel horror, want to feel happy, want to get angry, want to be sad), and recommend those that are close to that mood.
http: // www. tutaya. co. jp / tos / servlet / diagb Sachiko Komatsu, Minoru Wang, Shinichi Nakayama, Ken Shiina: Basic Research for Retrieval of Books by Emotional Expression Words: Trial of Book Classification Based on Emotional Evaluation after Reading, Library Information University Research Report 17 (1) pp. 63-75.

In the above conventional example, it is necessary to register user information for each service, and there is a problem in that the user has to input many times.
In addition, there is a service in which the display content is changed according to the user's profile. However, in the case of a context, there is a problem that it is difficult to obtain and it is difficult to make a clear inference.

  If the true purpose of the task is a service tailored to emotions, the user communicates emotions, but otherwise, the user does not input. Therefore, during execution of the target task, other information is hardly input. So, interact with the user before performing the task, get the emotion, and if there is a change in the user emotion that happened while using the service, also acquire the change at that time, and when it is updated, Need to be notified.

  In addition, in order to acquire user emotions, it is necessary to have a means to convey less burden on the user to the agent. At that time, there is a slight difference from the current user emotions, and it is not possible to follow the user emotions completely. Very difficult and close to impossible. Therefore, it is necessary to present the reliability of the emotion in consideration of the transition of the emotion. For this reason, in the case of the purpose of the task, it is necessary to use + α such as changing the way of providing the service, not just reducing the user's emotion input task.

In addition, since sensuous expressions have different unique expressions, meanings, and degrees depending on users, it is necessary to understand words and actions expressed by the users themselves.
The present invention has been made in view of the above-described circumstances, and an object thereof is to provide an emotion grasping system and an emotion grasping method capable of properly grasping user emotions.
It is another object of the present invention to provide an emotion grasping system and an emotion grasping method that take account of changes in user emotion.

In order to solve the above problem, an emotion grasping system according to claim 1 is:
An emotion grasping system for grasping user emotions,
Service usage observation means for determining whether or not the user is performing another task;
If the service usage observation means determines that the user is not performing another task, an interaction generation means for generating a predetermined action to interact with the user;
When it is determined that the user is performing another task in the service usage observation unit, an autonomous behavior generation unit that generates a predetermined behavior independent of the input of the user,
It is provided with.

  According to this configuration, when the service use observation unit determines whether the user is executing another task, and determines that the user is not executing another task, If a predetermined action is generated in order to perform a dialogue and it is determined that the user is performing another task, a predetermined action unrelated to the user's input is generated by itself. When the application is used, the user feeling acquired in the interaction between the user and the agent can be notified to the application, and an effect can be obtained that the service can be provided in consideration of the user's feeling.

The emotion grasping system according to claim 2 is characterized in that in claim 1,
It is characterized by comprising user model generating means for generating a user model formed by the profile and context of the user through the interaction with the user by the interaction generating means.
According to this configuration, a user model can be formed based on the user's profile and context through the interaction with the user. Therefore, it is easy to notify the user emotion to the application based on the user model. The effect that it is possible to provide a service that takes into account the feelings of people.

The emotion grasping system according to claim 3 is characterized in that in claim 2,
The interaction generation means includes a user profile acquisition means for acquiring the user profile, a user profile storage means for storing the profile, a context acquisition means for acquiring the user context, and an emotion of the user from the context. It is characterized by comprising user emotion acquisition means for acquiring, and user emotion expression means for expressing the acquired emotion in a general manner.

  According to this configuration, the interaction generation unit includes the user profile acquisition unit that acquires the user profile and the user profile storage unit that stores the profile, and the user profile acquisition unit that stores the profile. Since the user emotion acquisition means to acquire and the user emotion expression means to express the acquired emotion in a generic manner, both the user's profile and emotion can be considered, and the user's emotion is appropriately It is possible to provide a service that can take into account

The emotion grasping system according to claim 4 is characterized in that in claim 3,
The user emotion acquisition means holds an emotion adjustment means for updating and adjusting the user's emotion and a correspondence table corresponding to a predetermined emotion vector in order to grasp the strength or meaning of the user's emotion. Emotion word expression / emotion vector correspondence table holding means and input action / emotion vector correspondence table holding means for holding a correspondence table corresponding to a predetermined emotion vector in order to grasp the input action of the user It is characterized by.

  According to this configuration, in the user emotion acquisition means, the words that express the user's own feelings are the emotion word expression / emotion vector correspondence table holding means for knowing the strength and meaning of the emotion, and the user is the agent An emotion adjustment means for updating the user emotion expressed by the user emotion expression means with reference to the input action / emotion vector correspondence table holding means for knowing the degree of emotion indicated by the action to be performed Therefore, the user's emotion can be known from the action performed by the user and the emotion word to be input, and the user's own expression and its meaning can be reflected.

The emotion grasping system according to claim 5 is characterized in that in claim 4,
The user emotion acquisition unit refers to the emotion word expression / emotion vector correspondence table holding unit or the input action / emotion vector correspondence table holding unit, and adjusts the user's emotion expressed by the user emotion expression unit. Emotion adjustment means, new emotion expression registration means for newly storing an emotion word expression or input action adjusted by the emotion adjustment means, and emotion expression adjustment means for adjusting the expression method of the registered new emotion It is characterized by comprising.

  According to this configuration, an emotion word expression indicating emotion and an input action are newly added, a new emotion expression registration means for adjusting the degree of emotion, and an emotion word expression or an input action adjusted by the emotion adjustment means are newly added. By having the emotion expression adjustment means to additionally store, it is possible to know the user's emotion from the user's action and the emotion word to be input, and to reflect the user's own expression and its meaning An effect is obtained.

The emotion grasping system according to claim 6 is characterized in that in claim 4 or 5,
When the service usage observation means determines that the user is performing another task, the service usage observation means includes a user emotion reference means for querying the user emotion expression means for the user's current emotion, and the user emotion reference The means determines the certainty of the emotion of the user based on the final update confirmation unit for confirming the final update of the emotion of the user by the emotion adjustment unit and the last update time confirmed by the final update confirmation unit. It is characterized by comprising an emotion certainty calculating means.

  According to this configuration, when the service usage observation unit determines that the user is performing another task, the service usage observation unit includes the user emotion reference unit that inquires the user emotion expression unit about the current emotion of the user. The user emotion reference means is convinced of the user's emotion by the final update confirmation means for confirming the final update of the user's emotion by the emotion adjustment means and the last update time confirmed by the final update confirmation means. Since an emotion certainty degree calculating means for determining the degree is provided, an effect that the emotion can be grasped in consideration of the transition of the changing user emotion is obtained.

The emotion grasping method according to claim 7 is:
An emotion grasping method for grasping a user's emotion,
A service usage observation step of determining whether the user is performing another task;
An interaction generation step for generating a predetermined action to interact with the user when the service usage observation means determines that the user is not performing another task;
If it is determined that the user is performing another task in the service usage observation means, an autonomous behavior generation step for generating a predetermined behavior independent of the input of the user,
It is characterized by including.

  According to this configuration, in the service use observation step, it is determined whether or not the user is executing another task, and when the user determines that the other task is not being executed, If the user generates a predetermined action to interact with the user and determines that the user is performing another task, the user generates a predetermined action unrelated to the user's input. When other applications are used, the user feeling acquired in the interaction between the user and the agent can be notified to the application, and the application can provide a service that takes into account the user's feeling.

As described above, the present invention does not represent a predetermined emotional state with respect to input information in a system in which the user interacts with the user and estimates the emotion of the user from the text and operation exchanged in the user. Rather, it incorporates a means to personally adapt the user's unique input information and the state that it means.
Moreover, since it is a burden to make the user input emotions every time the service is used, a certainty factor for introducing user emotion state information that can be trusted to some extent while reducing the burden is introduced.

  Specifically, in an agent system in which an electronic agent and a user can interact via an interface, the user's status such as whether other tasks of the user are being executed is observed by a service usage observation means. However, when other tasks are not being executed, actions are generated to interact with the user by the interaction generation means, and when the user is executing other tasks, autonomous action generation is performed. By the means, the agent itself performs an action not related to the user input. In the user model generation means, a user profile acquisition means and a user context acquisition means acquire a user profile and context through user interaction, and the obtained user profile and context are stored as a user profile storage means. Save in the user context saving means. Also, as one of the user texts, there is an emotion acquisition means for acquiring the user's emotion, and the user emotion acquired by the emotion expression means is expressed generically, and when using other services, the user emotion reference means By having a user profile / context notification means for inquiring the user's emotions to the emotion expression means and sending the emotions to the service provider as part of the user context, the user interacts with the agent when using other applications. The user emotion acquired in the application can be notified to the application, and the application can provide a service in consideration of the user's emotion.

  With the above configuration, the user obtains emotions by interacting with the agent other than when using the service, and the service provider makes the service suitable for that emotion without asking the user about the user's emotions. Response can be changed. As a result, the user does not perform any task processing that requires emotion input especially in the task, and provides a service that matches the user's emotions without providing emotion input. Can be expected to be more comfortable and increase the frequency of using the service.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a schematic diagram showing the overall configuration of an embodiment of the emotion grasping system of the present invention.
This system provides a comfortable service to the user by conveying the emotion as the user's context to the service provider 50.

  This system is composed of a user 30, an agent 40, and a service provider 50, and the service provider 50 sends the emotion as a context of the user 30 via the agent 40. It is a mechanism that provides a comfortable service in consideration of the emotion of the user 30 by transmitting the message.

  The agent 40 includes a user model 41 and a multidimensional emotion model 131 in the system. The user model 41 performs interaction communication with the user 30 and repeats questions and answers. The user model 41 also holds a user profile (User Profile) 42 and a user context (User Context) 43. The user profile 42 stores data such as name, hobby, and preference, and the user context 43 stores data such as the position (Place) and emotion (Emotion) of the user 30. In the present embodiment, the user's emotion is considered as a part of the user context 43.

  When the user 30 is not particularly performing a task, the agent 40 interacts with the user 30 (Interaction Communication), and acquires the user profile 42 and the user context 43 therein. Furthermore, when there is an input from the user 30 to the agent 40 while using other services, the emotion is acquired from the input.

When the user 30 uses a service other than the agent 40, the service provider 50 considers the emotional state of the user 30 by notifying the service provider 50 of the user emotional state acquired by the agent 40. The service can be changed.
Correspondence of appropriate services according to each emotion is left to the service provider 50, and this system notifies the emotion of the user 30 by an expression method that can be understood by other systems.

  In the present embodiment, the service provider 50 includes a service provider A and a service provider B, and the service provider A includes an information database 51 that searches by keyword and a customer service manual 52 as information related to emotions. The service provider B includes a music database 53 to be searched by keywords and a recommendation book 54 as information related to emotion.

  The information database 51 stores related information, and the customer service manual 52 is a manual that recommends the attitude according to the emotion. For example, if the emotion is “anger”, the attitude at that time is “polite” Yes, I recommend "How is it?" Also, if the emotion is “Easy”, the attitude at that time is “Friendly”, and the language to be spoken is recommended as “How?” Also, if the emotion is “anxiety”, the attitude at that time is “friendly”, and the recommended language is “I think this is good”.

  The music database 53 stores information related to music, and the recommended book 54 is a book that recommends the music to be heard at that time according to the emotion. For example, if the emotion is “angry”, the music to be heard at that time Recommends “Ballard A”, “Jazz B”, etc. If the emotion is “happy”, the music to be listened to at that time is recommended as “J Pop C”, “Rock D”, etc. In the case of “anxiety”, the music to be heard at that time is recommended as “Classic E”, “Healing F” or the like.

  In this way, in the case of information retrieval, related information is extracted from the information database 51 and the display method is extracted from the customer service manual 52. In the case of purchasing music or CD, the music recommended by the recommendation book 54 is extracted from the music database 53, and when searching for a specific music, the music database 53 is searched and extracted as it is.

An example of the application will be shown below.
The service provider A is a service that provides information, and the user inputs a keyword for information to be searched. The service provider 50 side extracts information related to the keyword from the information DB 51 and presents it to the user. At the same time, the customer service manual 52 that lists the emotional state and response attitude of the user is referred to, and the display method is changed according to the emotional state of the user. Therefore, the information obtained by the user has not changed. For example, the service provider A takes a “careful” attitude to the feeling of “anger” as the customer service manual 52, and as a result, the word “How is it?” Is issued. In addition, it takes a “friendly” attitude toward the feeling of “easy” and as a result, the word “how?” Is issued. In addition, it takes a “kind” attitude towards the feeling of “anxiety” and as a result, the word “I think this is good” is issued.

  Service provider B is an example in which the information content changes depending on the emotional state of the user. It has a music DB 53 and is a service for searching for and introducing songs and CDs to the user. For example, the music database 53 holds classical music, jazz, J-hop, music box, etc., and the recommendation book 54 is a specific song or CD, and a keyword indicated by the music is input and extracted from the music database 53. However, in the search for searching for something that is particularly good because the song has not been decided, even if the keyword entered by the user is only “JAZZ”, it is obtained from the agent 40 by referring to the recommendation list by emotion. If the emotion of the user 30 is sad, a calm Smooth JAZZ or a slightly faster tempo Swing JAZZ is preferentially presented.

FIG. 2 is a block diagram showing the configuration of the embodiment of the emotion grasping system of the present invention.
This system 100 includes a user task investigation unit (service usage observation unit) 1, an interaction generation unit 2, an autonomous action generation unit 3, an input determination unit 4, a user emotion reference unit 5, a user profile (Profile). / Context (Context) notification unit 6 and user model generation unit 7. The user model generation unit 7 includes a profile acquisition unit 8, a context acquisition unit 9, a user profile storage unit 10, a user profile update unit 11, a user emotion acquisition unit 12, and a user emotion display unit 13. Yes.

The user task investigation unit 1 grasps what action the user 30 is currently performing. At that time, in particular, information on which application is active is acquired. In other words, it is also referred to as a service usage observation unit in the sense of observing user service usage.
The interaction generation unit 2 generates an action for interacting with the user 30 when there is no input from the user 30.

When the autonomous action generation unit 3 determines that the user 30 is executing another task, the autonomous action generation unit 3 generates an action that is not related to the input of the user 30.
The input determination unit 4 determines whether the content input by the user is a profile or a context. If the content is a profile, the input determination unit 4 passes the content to the profile acquisition unit 8 of the user model generation unit 7. If the content is a context, the input determination unit 4 7 to the context acquisition unit 9.

The user emotion reference unit 5 inquires when the last user emotion update is performed, determines the certainty of the user's emotion based on the last update time, and uses the emotion expression expressed in the user emotion expression unit 13 and its value to determine Calculate the current emotional state.
The user profile / context notification unit 6 gives the emotion acquired by the user emotion acquisition unit 12 to the service provider via the user emotion expression unit 13 and the user emotion reference unit 5 as part of the user context.

The user model generation unit 7 includes a profile acquisition unit 8, a user profile storage unit 10, a user profile update unit 11, a context acquisition unit 9, a user emotion acquisition unit 12, and a user emotion expression unit 13. A user model such as the profile and context of the user 30 is generated.
The profile acquisition unit 8 acquires a profile of the user 30 such as name, hobby, preference, and age.
The user profile storage unit 10 stores and stores the profile of the user 30 acquired by the profile acquisition unit 8.

The user profile update unit 11 changes and updates the profile of the user 30 acquired by the profile acquisition unit 8 according to circumstances.
The context acquisition unit 9 acquires the context of the user 30.
The user emotion acquisition unit 12 acquires information on the emotion of the user 30 as a part of the context of the user 30 acquired by the context acquisition unit 9.
The user emotion expression unit 13 includes means for expressing the user's 30 emotion acquired by the user emotion acquisition unit 12 by various methods.

FIG. 3 is a block diagram showing the configuration of the user emotion acquisition unit and the user emotion expression unit in the embodiment of the emotion grasping system of the present invention. (A) is a user emotion acquisition part, (b) has shown the user emotion expression part.
As shown in FIG. 3A, the user emotion acquisition unit 12 includes an emotion adjustment unit 121, a new emotion expression registration unit 122, an emotion expression adjustment unit 123, and an emotion word expression / emotion vector correspondence table holding unit 124. , And an input action / emotion vector correspondence table holding unit 125.
The emotion adjustment unit 121 standardizes information about emotion to some extent in order to express the emotion of the user 30 in a predetermined format.

The new emotion expression registration unit 122 stores information on the emotion of the user 30 that has been standardized to some extent by the emotion adjustment unit 121 in a predetermined place.
The emotion expression adjustment unit 123 further adjusts the information related to the emotion that is registered to some extent by the new emotion expression registration unit 122.
The emotion word expression / emotion vector correspondence table holding unit 124 is connected to the emotion adjustment unit 121, the new emotion expression registration unit 122, and the emotion expression adjustment unit 123, and displays a correspondence table indicating the correlation between the emotion word expression and the emotion vector. Hold it and send it to each unit as necessary.

The input action / emotion vector correspondence table holding unit 125 is also connected to the emotion adjustment unit 121, the new emotion expression registration unit 122, and the emotion expression adjustment unit 123, and indicates the correlation between the input action by the user 30 and the emotion vector. The table is held and sent to each of the above parts as necessary.
As shown in FIG. 3B, the user emotion expression unit 13 stores a multidimensional emotion model 131 and others, and expresses user emotions by various methods.
The multi-dimensional emotion model 131 can express a user's emotion on a plurality of axes, can express a complicated user's emotion in more detail, and can convey even a fine emotional difference felt by the user.

FIG. 4 is a block diagram showing the configuration of the user emotion reference unit in the embodiment of the emotion grasping system of the present invention.
The user emotion reference unit 5 includes a final update confirmation unit 55, a feeling certainty calculation unit 56, and a user task survey unit 57 after the last update.
The final update confirmation unit 55 inquires when the final user emotion update is performed.
In the user task survey unit 57 after the last update, the execution status of the user's task after the last user emotion update, that is, whether or not the same task of the user is continued, whether or not another task has been executed, Inquires whether or not an input has been made.
The emotion certainty calculation unit 56 calculates the certainty of the emotion from the elapsed time from the last update obtained from the last update confirmation unit 55 and the updated user action history obtained by the user task survey unit 57 after the last update. decide.

FIG. 5 is a block diagram showing the configuration of the emotion adjustment unit in the embodiment of the emotion grasping system of the present invention.
The emotion adjustment unit 121 includes an emotion word input unit 1211, an action input unit 1212, an emotion word selection unit 1214, an action explanation sentence selection unit 1215, a user ID acquisition unit 1213, a user ID identification unit 1216, and a personal A database 1217, a general database 1218, a chart data storage unit 1219, a chart display generation unit 1220, a chart display unit 1221, a user input unit 1222, an overall intensity data holding unit 1223, and an overall intensity data generation unit 1224; The total intensity data display unit 1225 and a user input unit 1226 are included.

The emotion word input unit 1211 takes in an emotion word related to the emotion of the user 30 as a part of the context acquired by the user emotion acquisition unit 12, and the emotion word selection unit 1214 generates a useful emotion word from the emotion word. select.
The behavior input unit 1212 takes in information related to the behavior as a part of the context acquired by the user emotion acquisition unit 12, and the behavior explanation for explaining the behavior from the information related to the behavior in the behavior explanation sentence selection unit 1215. Select a sentence.

The user ID acquisition unit 1213 acquires a user ID as a part of the context acquired by the user emotion acquisition unit 12, and identifies the user 30 by identifying the user ID by the user ID identification unit 1216.
The personal database (DB) 1217 includes the emotion word selected by the emotion word selection unit 1214, the action description selected by the behavior description selection unit 1215, and the user identified by the user ID identification unit 1216. Store the ID.

The general database (DB) 1218 stores the emotional word selected by the emotional word selection unit 1214 and the behavioral explanation text selected by the behavioral explanation text selection unit 1215.
The chart data holding unit 1219 stores the information stored in the personal database 1217 and the general database 1218 as data for displaying a chart.
The chart display generation unit 1220 generates the data stored in the chart data storage unit 1219 in place of the data for chart display.

The chart display unit 1221 displays the data generated by the chart display generation unit 1220 as a chart.
The overall intensity data holding unit 1223 is used to determine the intensity of emotion by referring to the input information input from the user input unit 1226 for information on emotions stored in the personal database 1217 and the general database 1218. Retain as data.
The overall intensity generation unit 1224 generates the data held in the overall intensity data holding unit 1223 by changing the emotion intensity so that the emotion intensity can be determined.
The overall intensity display unit 1225 displays so that the intensity of the data generated by the overall intensity generation unit 1224 can be understood.

FIG. 6 is a flowchart showing the user emotion acquisition operation of this system.
In order to acquire user emotion, first, it is determined whether or not there is user input (determining whether or not there is user input) (step S61). If there is user input, the input action and emotion word are investigated. This is performed (step S62). Therefore, first, it is determined whether or not the input is an emotion word (step S63). If the input is an emotion word, the emotion state indicated by the emotion word is referred to (step S64), and the emotion vector is determined. A value is acquired (step S65), the user emotion model is updated according to the acquired emotion vector value (step S66), and the process ends. In step S61, if there is no user input, the action is generated (step 67), and the action is generated to interact with the user (step S68). Furthermore, a question is generated in order to acquire user emotion (step S69). Furthermore, in step S63, when the input is not an emotion word, after referring to the emotion state indicated by the input action (step S70), an emotion vector value is acquired (step S65), and the user emotion model is updated. (Step S66), and the process ends.

FIG. 7 is a flowchart showing the notification operation of this system.
The notification operation of the system first determines whether or not the user emotion is held (step S71). If the user emotion is held, after referring to the user model / user emotion (step S72), After confirming the final update of the user's emotion, the user emotion expression vector is acquired (step S73), the user emotion expression vector value and the certainty degree are both notified (step S74), and the process is terminated. Here, in step S71, when the user emotion is not held, it is notified that there is no user emotion (step S75), and the process ends.

FIG. 8 is a flowchart showing the overall monitoring operation of this system.
In the overall monitoring of the system, first, it is determined whether or not the user is executing another task (step S81). If the user is executing another task, the other application / service provider (SP) is informed. An emotion is notified (step S82) and it complete | finishes. In step S81, if the user is not executing another task, user emotion is acquired (step S83), and the process ends.

The agent uses an anthropomorphic character so that the user can easily recognize it as an input target when inputting emotions. It is assumed that the environment has a display such as a PC, PDA, mobile phone, etc., and can input text with a keyboard, buttons, etc., and can interact with graphics on the screen with buttons, mouse buttons, etc.
In addition, the user task survey unit 1 that grasps what action the user is currently performing acquires information on which application is active.

FIG. 9 is a diagram showing an emotion space. This figure represents each emotional state of the agent and the user as a vector in the four-dimensional emotional space.
In this embodiment, an emotion space representing the emotional state (degree / meaning) of the user is represented by a four-dimensional emotion space. The emotion space is based on the Plutchik emotion model. Plutchik's emotion model has eight basic emotions (delight, grief, caress, disgust, fear, rage, wonder, alert). Basic emotions are the smallest unit of emotions that cannot be further reduced, and various other emotions can be created by combining basic emotions.

As shown in the figure, each emotional state is expressed as a position in the four axes (joy / sorrow, favor / dislike, fear / anger, surprise / warning (expectation)) where the two basic emotions are opposites. The more you go to the center, the weaker your basic feelings are. Here, the elements of each of the four axes representing the emotion space are represented as Ei (1 ≦ i ≦ 4), the variables indicating the respective values on the axes are emotion variables, and Ea _i (1 ≦ i ≦ 4). .

  There are two types of user input: input of emotion words to convey the user's emotions to the agent, and actions to be performed on the agent. The emotion words and actions input by the user are expressed as a four-dimensional vector in the emotion space and stored in the emotion expression map. Since meanings of emotion words and actions are used differently depending on the individual, it is necessary to express the difference between them, and it is also necessary to add different emotion expressions for each individual.

  Since the user recognizes the difference in the expression of each word / action and uses it properly, it is suitable for the user himself to express or adjust the difference. However, the emotion space is four-dimensional, and it is difficult for a human to indicate the meaning and degree of each emotion word in the four-dimensional space. Therefore, a mechanism for the user to express the difference intuitively is necessary. Furthermore, in order to indicate the difference from other words, it is also necessary to express the relationship with other vocabularies and acts. Therefore, an adjustment interface on a plane is provided for the user to make adjustments.

  In this example, as shown in the figure, the joy / sad axis that is the main axis of the emotion space is the vector E1, the anger / dislike axis is the vector E2, the expectation / surprise axis is the vector E3, and the other emotion vectors are E4. Yes. The joy category vector W = (a, b, c, d) is completely contained in the joy above the joy / sadness axis. As described above, in the emotion space, the two aspects of adjusting the meaning of the emotion and adjusting the emotion strength are provided. A surface for adjusting the meaning of emotion is referred to as adjustment surface A, and a surface for adjusting the strength of emotion is referred to as adjustment surface B.

FIG. 10 is a diagram illustrating a method for adjusting the difference in meaning of emotion on the adjustment surface A.
When each emotional state is represented by a vector, the four axes of emotion are represented by vectors E1 to E4, the vector E1 is a happy / sad axis, the vector E2 is a good / dislike axis, the vector E3 is a fear / anger axis, and the vector E4 is surprised / As a warning, the emotional words "Happy", "Happy", "I'm happy", and "Excellent" indicate the corresponding position.

As shown in the figure, each of the four axes indicating basic emotions has an angle of 45 degrees and is arranged around the origin. When the emotion category vector value is (A, B, C, D), the vector W1 (= a ₁ , b ₁ , c ₁ , d ₁ ) representing a certain emotion expression represents the value of each axis as (a ₁ − A, b ₁ -B, c ₁ -C, d ₁ -D). A plane connecting adjacent coordinate values is displayed in different colors for each emotional expression. By this display, it is possible to visually grasp the difference in the elements of basic emotions of each emotional expression, and it is possible to change the value of basic emotions constituting the emotional expression desired to be adjusted on the chart. By simultaneously displaying emotion expressions within the same emotion vector, it is possible to make adjustments while comparing with other emotion expressions. In addition, related expressions in different categories are displayed as necessary.

For example, when it is desired to adjust the meaning of the word “happy”, the word belongs to the “joy” category. The emotion category vector value of the “joy” category is approximately (+7, +5, +2, 0), and “happy”, “great”, “yatter” in the same category as the category vector and the “happy” vector, The meaning is adjusted based on the difference from “Lunrun”. When the word “happy” is used for something more preferable than “happy”, the value of the good / dislike axis E ₂ is made larger than the value of “happy”.

As a result, the emotional word “Happy” has a strong meaning of “good”, the emotional word “happy” has a strong meaning of “joy”, “good” and “fear”, and the emotional word “happy” The meanings of “joy” and “surprise” are strong, and the meanings of “anger” and “surprise” are moderately strong. The emotion word “wonderful” has a stronger meaning of “politics”, “anger” and “dislike”.
As described above, the adjustment plane A adjusts the subtle difference in meaning of emotion expressed by each emotion expression. Here, centering on the emotion category vector to which the word to be adjusted belongs, the difference in the value of each axis from that vector is presented.

FIG. 11 is a diagram illustrating a method for adjusting a difference in emotional intensity on the adjustment surface B.
This figure is a diagram showing a method for adjusting the emotion strength of the emotion expression on the adjustment plane B. FIG.
As shown in the figure, the adjustment of the emotion strength of the emotion category vector is expressed by the length, and the comparison is made with the length to express the strength. This length is expressed by the magnitude of the vector. When the emotion category vector value is (A, B, C, D), L = (A ² + B ² + C ² + D ² ) ^1/2 . Similarly, the length of each emotional expression is calculated, and the length is displayed by changing the color. In a certain emotional expression W1 (= (a ₁ , b ₁ , c ₁ , d ₁ )), the length is
LWi = (a ₁ ² + b ₁ ² + c ₁ ² + d ₁ ² ) ^1/2

The adjustment made on each adjustment surface is converted back to four dimensions, and the emotion vector of the emotion expression is determined. In the adjustment performed on the adjustment surface A, each emotion axis value is changed. As a result of the adjustment, if there is a change of (−1, +3, +1, 0) than before, the vector W1 = a ₁ −1, b ₁ +3, c ₁ +1, d ₁ ). In the strength adjustment performed on the adjustment surface B, if n is weaker than the emotion category vector L, the word is expressed by the vector W1 = a−nA / L, b−nB / L, c−nC / L, d−nD / L).

  The adjustment side of this emotion vocabulary map is used when it is better to give instructions directly, such as when adding words, setting initial values of actions, or when the meaning or strength changes greatly. Given the initial value of emotional expression, further adjustments are made during the interaction between the agent 40 and the user 30. When the behavior expressed by the agent 40 is not as intended by the user 30 with respect to the emotional expression input by the user 30, the user 30 inputs good / bad and performs adjustment again.

  The notification that notifies the emotion state acquired from the user 30 to the service used by the user 30 is not the emotion expression input by the user 30, but the emotion vector value that is a commonly understood expression. To do. Here, it has a vector value expressed in 0.1 units. The service provider normalizes the emotion vector according to the accuracy of the available service and selects each response.

FIG. 12 is a diagram showing a method for creating a customer service manual centered on a service provider.
As shown in the figure, the service provider not only has the information DB 51 used in the main task of the service, but also has a list of texts to be displayed corresponding to the notified user emotion. In the example shown in FIG. 12, the user's emotions are divided into four stages of emotions, emotions, and emotions, each of which has a table in which response text displays are associated with each other. And determine the response. FIG. 12 is an example of the correspondence.

  Specifically, as the customer service manual 52, emotion vector values and response attitudes are discriminated for the feelings of "anger", "cheerful, fun", "depressed", and "anxiety", respectively, and display text is displayed. ing. First, for the emotion “anger”, the emotion vector value is E3 ≧ 3.0, the response attitude is polite, and the display text is “How is it?”. Similarly, for the emotion “cheerful, fun”, the emotion vector value is E1 ≧ 3.0 or E2 ≧ 1.0, the response attitude is friendly, and the display text is “how?” Or, “Is this not good?” Is displayed. Similarly, for the emotion “depression”, the emotion vector value is E1 ≦ −2.5, the response attitude is bald, and the display text is “That is good, but how about this?” Is displayed. For the emotion “anxiety”, the emotion vector values are E4 ≦ −1.0 and E3 ≧ 1.0, the response attitude is relative, and the display text is “This is good” And so on.

FIG. 13 is a diagram showing emotion categories.
There are many types of sensitivity words representing emotions, but according to a psychological experiment, Japanese is classified into 35 categories. FIG. 13 shows 35 categories of 7 columns and 5 columns. In this study, these 35 categories are called emotion categories and are used as major classifications of emotion words.
As shown in the figure, the emotion category is, from left to right in the upper row, along with “dislike, disappointment, enjoyment, strong dislike, humiliation”, and in the second row, “sluts, anxiety, doubt, warmth, pain The third row is lined with “Intolerance, Fear, Fear, Discomfort, Insult”, and the fourth row is lined with “Sadness, Pride, Loneliness, Regret, Curiosity” Along with "Threat, Vague, Happiness, Respect, Love", and the seventh row is followed by "Joy, hesitation, shame, anger, pity".

  Each emotion category provides an emotion category vector EC representing each emotion state included in the category. For example, the emotion category vector of the “dislike” category is (−3, −5, 0, 0). Each emotion word has information on the emotion category represented by the word. It is assumed that the emotion words belong to the above category with respect to the user's actions as well as being included in the category. In this study, adjectives and adjective verbs, impression verbs, pseudowords, and pictograms are targeted as emotional words.

FIG. 14 is a diagram showing emotion category vectors and emotion expressions (partial).
This figure shows each emotion category vector and a part of emotion words belonging to each category.
There are five categories: “joy”, “fun”, “sad”, “anxiety”, and “unpleasant”, and the vector values of “joy, good, good, strange” and their corresponding emotional expressions (vocabulary / Act). Emotional expression indicates a typical vocabulary or action that expresses emotion. For example, the vector value of the category “joy” is “+5, +4, 0, −2”, the emotional expression is “happy, happy, wonderful, Yatter, Lunrun”, and the vector value of the category “fun” is “+5” , +4, 0, +2 ”, the emotional expression is“ fun, interesting, exciting, exciting ”, the vector value of the category“ sad ”is“ −5, −3, −1, −2 ”, and the emotional expression is “Fair, sorrowful, rustic, sad, squeaky”, the vector value of the category “anxiety” is “−2, −2, −5, +2”, No ”, the vector value of the category“ unpleasant ”may be“ −3, −5, +3, −3 ”, and the emotional expression may be“ feeling uncomfortable, disgusting, unpleasant, strange ”or the like.

In addition, the actions are actions such as “striking”, “stroking”, “pushing”, etc. expressed by operating the mouse.
When indicating the difference in the degree and meaning of each emotional expression, we think that it is most meaningful to express the difference from other emotional expressions included in the same category. Make adjustments by comparing the relative positions of emotional expressions. Moreover, since there are many expressions unique to individuals, it is required not only to make adjustments, but also to add vocabulary and actions easily. Therefore, by introducing this emotion category, an initial value at the time of adjustment can be obtained when a new vocabulary is added, and the labor for adjustment from the beginning can be saved.

FIGS. 15 and 16 are diagrams showing a user emotion certainty factor and an emotion update elapsed time factor, respectively.
It is assumed that the service provider side sets what kind of emotion is classified into the emotion vector and what kind of response is to be made accordingly.
If the user 30 interacts with the agent 40 immediately before using the service and the emotion is input, it is considered that it is not so far from the emotion of the user 30 at the present time. If not, that feeling is likely different. Therefore, a factor of certainty of user emotion can be used.

  Therefore, the user emotion acquisition unit 12 not only refers to the emotional state of the user 30 but also obtains a time t at which the emotional state is updated. The user task investigation unit 1 observes the value and whether the user 30 has continued to interact with the agent 40, uses another application, or has not been input. The certainty level c of the current emotional state of the user 30 is set according to the given state. The service provider side decides whether or not to use the emotion based on the certainty factor. As the certainty factor, an exponential function is used. Here, the certainty factor is c = k ′ (0 ≦ c ≦ 1, k <1) because the certainty decreases with time. The value of k is determined by the task state of the user after the user emotion input.

As shown in FIG. 15, the coefficient k when the user state is the interaction continuation is 0.9, the coefficient k when there is no input is 0.8, and the coefficient k when using other applications is 0.7. And
Further, as shown in FIG. 16, the coefficient t for the elapsed time 0 to 10 is 0, the coefficient t for 10 to 20 is 0.5, and the coefficient t for 20 to 30 is 1.0. , 30 to “60, the coefficient t is 1.5, and the coefficient t in the case of 60 to 180 is 2, 0. These certainty factors are notified together with the emotion vector value.

  Thus, according to the configuration of the present invention, the user obtains emotions by interacting with the agent other than when using the service, and the service provider does not ask the user's emotions to the user, The service can be changed to a response suitable for the emotion. As a result, the user does not perform any task processing that requires emotion input especially in the task, and provides a service that matches the user's emotions without providing emotion input. Become more comfortable and the frequency of using the service increases.

  In addition, expressing emotions using a multi-dimensional emotion model, not selecting the expression prepared by the system, but judging the user's emotion from the emotion expression expressed by the user himself, Since the difference in particular nuances can be transmitted to the system, the user can feel that his / her feelings have been transmitted, and the use of the service is further promoted.

  In other words, the service provider can change the service to an appropriate response to the emotion without always asking the user about the user's emotion. As a result, it is possible to provide a service tailored to the user's emotion without input of emotion. Therefore, the probability of harming the user's mood is reduced, the emotion is expressed using the service, and it is subtle to judge the user's emotion from the emotion expression expressed by the user rather than selecting the expression prepared by the system side. It is possible to feel the difference of emotion and the difference of nuance that the user is particular about. Thereby, the use of the service is promoted.

  In the embodiment described above, in executing the processing shown in the flowcharts of FIGS. 6 to 8, the case where the control program stored in advance in the ROM is executed has been described. However, the present invention is not limited to this. Instead, the program may be incorporated into a RAM and executed from a storage medium on which a program showing these procedures is recorded. Alternatively, the program may be acquired from a network.

Here, the storage medium is a semiconductor storage medium such as RAM or ROM, a magnetic storage type storage medium such as FD or HD, an optical reading type storage medium such as CD, CDV, LD, or DVD, or a magnetic storage type such as MO. / Optical reading type storage medium, including any storage medium that can be read by a computer regardless of electronic, magnetic, optical, or other reading methods.
As mentioned above, although the emotion grasping | ascertainment system of this invention has been demonstrated centering on embodiment, this invention is not limited to this, A various change is possible without deviating from the meaning of this invention. is there.

It is a general view which shows the structure of the emotion expression consideration service using the emotion grasping | ascertainment system of this invention. It is a block diagram which shows the structure of the emotion recognition system of this invention. It is a block diagram which shows the structure of the user emotion acquisition part and user emotion expression part in the emotion grasping | ascertainment system of this invention. (A) shows a user emotion acquisition part, (b) shows a user emotion expression part. It is a block diagram which shows the structure of the user emotion reference part in the emotion grasping | ascertainment system of this invention. It is a block diagram which shows the structure of the emotion adjustment part in the emotion grasping | ascertainment system of this invention. It is a flowchart which shows the operation | movement of user emotion acquisition in the emotion grasping | ascertainment system of this invention. It is a flowchart which shows the operation | movement of the system notification in the emotion grasping | ascertainment system of this invention. It is a flowchart which shows the whole monitoring operation | movement in the emotion grasping | ascertainment system of this invention. It is a figure which shows an emotion space. It is a figure which shows the method of adjusting the difference in the meaning of an emotion. It is a figure which shows the method of adjusting the difference in the strength of emotion. It is a figure which shows the specific example of a customer service manual. It is a figure which shows the specific example of an emotion category. It is a figure which shows the specific example of an emotion category vector and emotion expression (part). It is a figure which shows the specific example of a user feeling certainty factor. It is a figure which shows the specific example of emotion state update elapsed time.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... User task investigation part, 2 ... Interaction production | generation part, 3 ... Autonomous action production | generation part, 4 ... Input judgment part, 5 ... User emotion reference part, 6 ... User profile / context notification part, 7 ... User model production | generation part, DESCRIPTION OF SYMBOLS 8 ... Profile acquisition part, 9 ... Context acquisition part, 10 ... User profile preservation | save part, 11 ... User profile update part, 12 ... User emotion acquisition part, 13 ... User emotion expression part, 30 ... User, 40 ... Agent, 41 ... User model, 42 ... profile, 43 ... context, 50 ... service provider, 51 ... information database, 52 ... customer service manual, 53 ... music database, 55 ... last update confirmation unit, 56 ... emotion confidence calculation unit, 57 ... last update Post-user task investigation unit, 100 ... emotion grasping system, 121 ... emotion adjustment unit, 122 ... Reference emotion expression registration unit, 123 ... emotion expression adjustment unit, 124 ... emotion word expression / emotion vector correspondence table holding unit, 125 ... input action / emotion vector correspondence table holding unit, 131 ... multidimensional emotion model, 1211 ... emotion word input 1212 ... Action input part, 1213 ... User ID acquisition part, 1214 ... Emotion word selection part, 1215 ... Action description sentence selection part, 1216 ... User ID identification part, 1217 ... Personal database, 1218 ... General database, 1219 ... Chart data storage unit, 1220 ... Chart display generation unit, 1221 ... Chart display unit, 1222 ... User input unit, 1223 ... Overall intensity data holding unit, 1224 ... Overall intensity generation unit, 1225 ... Overall intensity display unit, 1226 ... User input Part

Claims (7)

An emotion grasping system for grasping user emotions,
Service usage observation means for determining whether or not the user is performing another task;
If the service usage observation means determines that the user is not performing another task, an interaction generation means for generating a predetermined action to interact with the user;
When it is determined that the user is performing another task in the service usage observation unit, an autonomous behavior generation unit that generates a predetermined behavior independent of the input of the user,
Emotion grasping system characterized by having In claim 1,
An emotion grasping system comprising user model generation means for generating a user model formed by a profile and a context of the user through the interaction with the user by the interaction generation means. In claim 2,
The interaction generation means includes a user profile acquisition means for acquiring the user profile, a user profile storage means for storing the profile, a context acquisition means for acquiring the user context, and an emotion of the user from the context. An emotion grasping system comprising: a user emotion acquisition means for acquiring; and a user emotion expression means for generally expressing the acquired emotion. In claim 3,
The user emotion acquisition means holds an emotion adjustment means for updating and adjusting the user's emotion and a correspondence table corresponding to a predetermined emotion vector in order to grasp the strength or meaning of the user's emotion. Emotion word expression / emotion vector correspondence table holding means and input action / emotion vector correspondence table holding means for holding a correspondence table corresponding to a predetermined emotion vector in order to grasp the input action of the user Emotion grasping system characterized by In claim 4,
The user emotion acquisition unit refers to the emotion word expression / emotion vector correspondence table holding unit or the input action / emotion vector correspondence table holding unit, and adjusts the user's emotion expressed by the user emotion expression unit. Emotion adjustment means, new emotion expression registration means for newly storing an emotion word expression or input action adjusted by the emotion adjustment means, and emotion expression adjustment means for adjusting the expression method of the registered new emotion Emotion grasping system characterized by having In claim 4 or 5,
When the service usage observation means determines that the user is performing another task, the service usage observation means includes a user emotion reference means for querying the user emotion expression means for the user's current emotion, and the user emotion reference The means determines the certainty of the emotion of the user based on the final update confirmation unit for confirming the final update of the emotion of the user by the emotion adjustment unit and the last update time confirmed by the final update confirmation unit. An emotion grasping system characterized by comprising an emotion certainty calculating means. An emotion grasping method for grasping a user's emotion,
A service usage observation step of determining whether the user is performing another task;
An interaction generation step for generating a predetermined action to interact with the user when the service usage observation means determines that the user is not performing another task;
If it is determined that the user is performing another task in the service usage observation means, an autonomous behavior generation step for generating a predetermined behavior independent of the input of the user,
Emotion grasping method characterized by including.

Images