JP7294924B2 - dialogue agent system - Google Patents

Description

The present disclosure relates to dialog agent systems.

Patent Literature 1 discloses a dialog agent system that interacts with a user to execute a task. Specifically, the dialogue agent system described in Patent Document 1 detects the user's voice and understands the content of the user's utterance. The dialogue agent system then determines response content based on the understood utterance content, and outputs a voice message based on the determined response content. Furthermore, this system displays an agent character, which is a character embodying a dialogue agent, on a user terminal or the like, and outputs a voice message as if the agent character were speaking.

JP 2015-194864 A

In the system described in Patent Literature 1, the same agent character repeatedly responds to user utterances without changing the appearance of the agent character displayed on the user terminal or the like. Therefore, it is feared that the user will become bored and the user's motivation to actively use the system will decrease.

An object of one aspect of the present disclosure is to provide a system in which a character embodying a dialogue agent is displayed and a response content is spoken by the character, in which the user continues to be interested.

One aspect of the present disclosure is a dialogue agent system that includes a user speech accepting unit, a response acquisition unit, a display processing unit, a speech processing unit, an input data accepting unit, and a character updating unit. The user utterance accepting unit accepts user utterances. The response content acquisition unit acquires the response content to the user's utterance. The display processing unit causes the display unit of the user's information device to display an agent character, which is a character embodying the dialogue agent. The speech processing section causes the agent character displayed on the display section to speak the content of the response. The input data receiving unit receives input data acquired by the information device. The character updating section selects a character from among a plurality of types of characters based on input data, and causes the selected character to update the agent character.

According to such a configuration, the agent character is updated to a character selected from a plurality of types of characters based on the input data. Therefore, the appearance of the agent character is not updated, and the user can continue to be interested in the system as compared with a dialog agent system in which the same agent character repeatedly responds. As a result, new value can be provided to users.

One aspect of the present disclosure may further include a property updater. The characteristic update unit updates the characteristic information based on the input data. The characteristic information is information representing characteristics of the user. Then, the character updating section may select a character from among a plurality of types of characters based on the property information updated by the property updating section, and cause the selected character to update the agent character.

According to such a configuration, the more the user continues to use the dialogue agent system, the more the characteristic information is updated in consideration of the user's characteristics. As a result, the more the user continues to use the system, the more the agent character is updated to a character that takes into consideration the characteristics of the user. Therefore, according to the above configuration, it is possible to further keep the user interested in the system.

One aspect of the present disclosure may further include an update determination unit. The update determination unit determines whether or not a predetermined condition for updating the agent character is established for at least one of the plurality of types of classification information based on the input data. The multiple types of classification information are information for classifying the input data. Multiple types of characters are associated with multiple types of classification information. When the update determining unit determines that the predetermined condition is satisfied, the character updating unit may update the agent character to the character associated with the classification information determined to satisfy the predetermined condition.

According to such a configuration, the agent character is updated to the character corresponding to the classification information corresponding to the input data among the plurality of types of characters. As a result, when the input data represents information, interests, etc., that the user wants to know, the agent character can be updated to a character corresponding to the information, interests, etc., that the user wants to know. Therefore, it is possible to make the user more interested.

An aspect of the present disclosure may further include a specifying unit and a contribution updating unit. The identifying unit identifies the classification information corresponding to the input data and the amount of contribution degree change when the input data receiving unit receives the input data. The amount of contribution change is the amount of change in the cumulative degree of contribution to the update of the agent character. The contribution level updating unit stores, in a storage unit that stores, in association with each of the plurality of types of classification information, the cumulative degree of contribution to the update of the agent character, the contribution level corresponding to the classification information specified by the specifying unit. The cumulative degree is updated based on the amount of contribution change specified by the specifying unit. The update determination unit may determine that a predetermined condition has been met for the classification information when the cumulative degree of contribution of certain classification information reaches a predetermined degree.

According to such a configuration, when input data corresponding to certain classification information is input a plurality of times, the degree of accumulation of the degree of contribution related to the classification information reaches a predetermined level, and the character corresponding to the classification information becomes the agent character. is updated. Therefore, compared to the case where the agent character is immediately updated to the corresponding character when the specific input data is accepted, the agent character can be updated to the character that takes into consideration the characteristics of the user. As a result, it is possible to keep the user interested.

In one aspect of the present disclosure, when the input data accepting unit accepts first input data, the specifying unit includes classification information corresponding to the first input data, a first contribution change amount, may be specified. Further, when the input data accepting unit accepts second input data different from the first input data, the specifying unit stores the classification information corresponding to the second input data and the first contribution degree A second contribution change amount different from the change amount may be specified.

According to such a configuration, the amount of contribution change differs between when certain input data is accepted and when different input data is accepted. That is, the agent character can be updated to the corresponding character more quickly when certain input data is accepted. Therefore, depending on what kind of input data is input, the character after the agent character is updated and the speed until the character is updated differ. Therefore, it is possible to further improve the game characteristics of the dialogue agent system, and furthermore to keep the user interested in the system.

One aspect of the present disclosure may further include a condition determination value determination unit. The condition determination value determination unit determines a condition determination value corresponding to the input data or the classification information corresponding to the input data using the determination processing model. A condition determination value is a value for determining whether or not a predetermined condition is satisfied. The judgment processing model is a model generated by machine learning, and outputs a condition determination value, which is a value for determining whether or not a predetermined condition is satisfied, based on input data or classification information corresponding to the input data.

According to such a configuration, even if input data or classification information that is not used as teacher data for the determination processing model is input, the corresponding condition determination value can be output.
In one aspect of the present disclosure, the input data may include user speech.

According to such a configuration, the updated character of the agent character is selected based on the user's utterance. Therefore, it is possible to update the agent character to a character that matches the user's interest expressed by the user's utterance.

In one aspect of the present disclosure, the input data may include position information of the information device.
According to such a configuration, the character after the update of the agent character is selected based on the position information of the information device and, by extension, the user. That is, it is determined that the update result of the agent character changes depending on where the user is located. Therefore, it is possible to further improve the game characteristics of the dialogue agent system.

In one aspect of the present disclosure, the input data may include numerical data indicating user characteristics.
According to such a configuration, the update result of the agent character is selected based on the numerical data indicating the characteristics of the user. Therefore, it is possible to update the agent character to a character that considers the user's characteristics.

FIG. 1 is a block diagram showing the configuration of a dialogue agent system. FIG. 2 is a diagram showing an agent character. 3A is a property information table regarding contents, FIG. 3B is a property information table regarding position information of information equipment, and FIG. 3C is a property information table regarding numerical data. FIG. 4 shows a promotion information table. FIG. 5 is a flow chart showing the response content transmission process. FIG. 6 is a flowchart showing characteristic update processing. FIG. 7 is a flowchart showing character update processing. FIG. 8 is a branch diagram showing changes in the agent character.

EMBODIMENT OF THE INVENTION Hereinafter, the form for implementing this disclosure is demonstrated, referring drawings.
[1. composition]
A dialogue agent system 1 shown in FIG.

The dialog agent system 1 is a system that interacts with a user to execute a task. First, the basic functions of the dialogue agent system 1 will be explained.
The dialogue agent system 1 receives a user's utterance from the information device 3 owned by the user, and transmits a voice signal of the user's utterance to the voice recognition server 4 . Then, the voice signal received by the voice recognition server 4 is converted into text data by the voice recognition function, and the text data is transmitted to the information device 3 . The information device 3 transmits the text data representing the user's utterance received from the speech recognition server 4 to the property server 2 . The property server 2 generates a response (hereinafter referred to as “response content”) corresponding to the received user utterance, and transmits the generated response content to the information device 3 . Then, the information device 3 outputs a sound corresponding to the content of the generated response and executes a task corresponding to the content of the response.

The dialogue agent system 1 of this embodiment has the following functions in addition to the basic functions described above. That is, in the dialog agent system 1, as shown in FIG. 2, the display unit 31 of the information device 3 displays an agent character 51, which is a character embodying a dialog agent that interacts with the user. Then, as if the displayed agent character 51 were speaking, the content of the response corresponding to the user's speech is output. This agent character 51 is also a character imitating a dialogue agent.

In other words, the user feels as if he or she is interacting with the agent character 51, and the dialogue agent is more than just "voice".
Furthermore, the agent character 51 evolves (that is, its appearance is updated) based on the input data acquired by the information device 3 . Specifically, the dialogue agent system 1 analyzes the characteristics of the user of the information device 3 based on the input data. Then, the dialogue agent system 1 selects a character from a plurality of types of characters according to user characteristics, and updates the agent character 51 with the selected character.

The dialogue agent system 1 can be used by activating an agent app, which is dedicated application software, in the information device 3 . The dialogue agent system 1 will be described in detail below.

The property server 2 shown in FIG. 1 is assumed to be owned by, for example, an agent application production company. The property server 2 includes a communication unit 21 , a dialogue response platform (hereafter referred to as “PF”) 22 , a storage unit 23 and a control unit 24 .

The communication unit 21 is a communication interface for performing wired or wireless communication with an external device. The property server 2 is connected to the Internet N via the communication unit 21 and is capable of data communication with an external device via the Internet N. FIG. In this embodiment, the property server 2 communicates with the information device 3 via the communication unit 21 .

When the dialogue response PF 22 receives the text data representing the user's speech from the information device 3, it generates a response content for the received text data.
Specifically, inside the dialogue response PF 22, there are correspondences associated with dialogue patterns with the user of the information device 3 (hereinafter also referred to as "terminal user") and requests for viewing contents such as moving images, music, and texts. A data table of control codes for operating the information device 3 is registered. The control code is, for example, a content reproduction instruction or display instruction, volume adjustment, activation of application software, or the like. When the dialogue response PF 22 receives text data representing the user's utterance from the information device 3 via the communication unit 21, it searches the data table with the text data and acquires the corresponding control code to obtain the response content. Generate.

The storage unit 23 stores various information. In this embodiment, the storage unit 23 stores a characteristic information table and a promotion information table.
<Characteristic information table>
The characteristic information table is a data table as shown in FIGS. 3A to 3C, and is information representing terminal user characteristics. This characteristic information table is also a table that arbitrarily defines terminal user characteristics. User characteristics in the present embodiment include user's internal characteristics such as preferences and interests, physical characteristics such as blood type, and location information such as the area where the user is located.

Specifically, the characteristic information table is information in which classification information and importance information are associated with each other.
The classification information is information for classifying input data acquired by the information device 3 . Here, the input data referred to in the present embodiment includes a voice signal representing a user's utterance, text data representing the voice signal, response content corresponding to the user's utterance, position information of the information device 3, and numerical data having attribute meanings. etc. are included.

Numerical data having the meaning of attributes is, for example, numerical data indicating characteristics of a user. Numerical data are values that indicate characteristics defined within the dialogue agent system 1 . Examples of such numerical data include numerical data indicating a user's blood type. That is, for example, when a user answers a questionnaire within an agent application, points are given according to the answer to the questionnaire, and the points are associated with the user's blood type, etc., the numerical data is the answer of the questionnaire. It may be points.

As classification information for classifying such input data, in the present embodiment, classification information of content related to user utterance (for example, content desired to be viewed by the user), classification information of position information of the information device 3, and numerical data are stored in the storage unit 23 .

The content classification information is represented, for example, by multiplying the type of content and the genre. The content classification information is added by multiplying the type of content and the genre, and the process of evolution of the agent character 51 is separated as described later. The content classification information is, for example, "music x pops", "movie x action", "TV program x family", and the like.

The classification information of the location information is, for example, place names such as Tokyo, Nagoya, Hong Kong, and Taiwan.
The classification information of numerical data is, for example, numerical values such as "0xFF" and "0x0F".
On the other hand, importance information is set in association with each piece of classification information. The importance information is information indicating how important the corresponding classification information is in quantitatively or qualitatively representing the characteristics of the terminal user. That is, the higher the degree of importance information (in this embodiment, the score of the importance information), the higher the interest of the terminal user, and the more representative the characteristics of the terminal user.

In this embodiment, when input data such as user utterances, position information, and numerical data are input by the information device 3, classification information corresponding to the input data is identified, and importance information corresponding to the identified classification information is identified. The score is increased (that is, added points). Then, based on the score of the importance information, a character is selected from a plurality of types of characters prepared in advance, and the agent character 51 is changed (that is, updated) to the selected character.
<Promotion information table>
The promotion information table, as shown in FIG. 4, contains keywords, positions, numerical values, etc. for which the point value of importance information is to be increased from the standard point value (that is, to be promoted), the point value of importance information, is associated information.

That is, in the present embodiment, when the input data input by the information device 3 is specific input data, in other words, when the input data includes keywords such as content titles to be promoted, The score of the importance information of the classification information is increased by a significant addition value (eg, 30 points) from the standard addition value (eg, 5 points).

Specifically, when the input data is a user utterance, if the user utterance contains a specific keyword (for example, the title of a movie currently being talked about), the score of the importance information of the corresponding classification information is added as standard. Updated more than the value. Further, when the input data is the position information of the information device 3, if the position represented by the position information is a specific position, the score of the importance level information of the corresponding classification information is updated significantly more than the standard added value. be. Further, when the input data is numerical data, if the numerical value represented by the numerical data is a specific numerical value, the score of the importance level information of the corresponding classification information is updated significantly more than the standard degree of change.

For example, in the promotion information table shown in FIG. 4, the keyword "action movie X" indicating the title of a specific movie and its added value "40 points" are associated and stored. Also, in the promotion information table, the specific position "Taiwan" and its added value "30 points" are stored in association with each other. Further, in the promotion information table, a specific numerical value "0xFF" and its added value "30 points" are associated and stored.

The keywords and the like registered in the promotion information table are updated (changed) at an appropriate timing by the service provider of the dialogue agent system 1, such as an agent application development company.

The above is the information stored in the storage unit 23 .
Next, the control unit 24 shown in FIG. 1 is configured mainly by a well-known microcomputer having a CPU 241 and a semiconductor memory such as RAM, ROM, flash memory (hereinafter referred to as memory 242). Various functions of the control unit 24 are realized by executing programs in the control unit 24 configured by the CPU 241 and the memory 242 . A method corresponding to the program is executed. Note that the number of microcomputers that constitute the control unit 24 may be one or more. By executing the program, the control unit 24 executes a response content transmission process shown in FIG. 5 and a characteristic update process shown in FIG. 6, which will be described later.

On the other hand, the information equipment 3 is, for example, a smart phone, a tablet terminal, a personal computer, or the like. The information device 3 includes a display section 31 , an operation section 32 , a microphone 33 , a speaker 34 , a storage section 35 , a communication section 36 , a position detection section 37 and a control section 38 .

The display unit 31 has a display device capable of displaying an image, such as a liquid crystal display or an organic EL display.
The operation unit 32 receives user operations. The operation unit 32 includes, for example, buttons, switches, a touch panel, and the like.

Microphone 33 is configured to input sound.
The speaker 34 is configured to output sound.
The storage unit 35 stores various information. In this embodiment, the storage unit 35 stores image data and voice data of characters corresponding to each of a plurality of types of classification information included in the property information table stored in the storage unit 23 of the property server 2. may Here, the character stored in the storage unit 35 is not particularly limited, but may be a real or virtual person, animal, robot, or the like. For example, a character corresponding to the classification information of "music x pops" may be a character imitating a famous singer in the field of pop music.

The storage unit 35 also stores condition information indicating conditions for evolving the agent character 51 (that is, updating its appearance). Condition information is set for each classification information. The condition information is, for example, that the score of the importance information corresponding to the classification information reaches a predetermined score. This predetermined score is hereinafter referred to as "threshold score". That is, when the score of importance level information corresponding to certain classification information (for example, "music x pops" classification information) reaches a threshold score (for example, 30 points), the agent character 51 is assigned to the character corresponding to the classification information. Updated. Note that the threshold score may differ for each piece of classification information, or may be the same for all pieces of classification information.

In this embodiment, the agent character 51 is set to a default (initial value) character when the agent application is installed. In the period from the start of use of the application, when the score of importance information for any category information is less than the threshold score, when the sum of the points of importance information of all category information reaches a certain point, Agent character 51 is updated from the default character to another character. Then, when the score of importance information of certain classification information reaches the threshold score, the agent character 51 is updated to a character corresponding to the classification information.

The above is the information stored in the storage unit 35 .
The communication unit 36 is a communication interface for performing wired or wireless communication with an external device. The information equipment 3 is connected to the Internet N via the communication unit 36, and can perform data communication with an external device via the Internet N. FIG. In this embodiment, the information device 3 communicates with the property server 2 and the voice recognition server 4 via the communication unit 36 .

The position detection unit 37 detects the current position of the information device 3 using a GPS (Global Positioning System) or the like.
The control unit 38 is mainly composed of a well-known microcomputer having a CPU 381 and a semiconductor memory such as RAM, ROM, flash memory (hereinafter referred to as memory 382). Various functions of the control unit 38 are realized by the control unit 38 executing programs. The number of microcomputers constituting the control unit 38 may be one or more. The control unit 38 executes a character update process shown in FIG. 7, which will be described later, as the CPU 381 executes a program.

On the other hand, the voice recognition server 4 is assumed to be owned by a company called Virtual Assistant, such as Amazon (registered trademark) of the United States and Google (registered trademark) of the United States, which provides a service for operating devices by voice. . The voice recognition server 4 receives a voice signal of user's speech from the information device 3 via the Internet N, and converts the received voice signal into text data by a voice recognition function. The speech recognition server 4 then transmits text data representing the user's utterance to the information device 3 .

[2. process]
[2-1. Response content transmission process]
Next, the response content transmission processing executed by the control unit 24 of the property server 2 will be described using the flowchart of FIG. The response content transmission process is executed each time the characteristic server 2 receives text data representing the user's utterance from the information device 3 via the communication unit 21 .

First, in S<b>101 , the property server 2 receives text data representing user speech from the information device 3 .
Subsequently, in S102, the property server 2 searches the received text data representing the user's utterance to obtain the content of the response.

Subsequently, in S103, after transmitting the response content to the information device 3, the characteristic server 2 terminates the response content transmission process of FIG.
[2-2. Characteristic update process]
Next, the property update processing executed by the control unit 24 of the property server 2 will be described using the flowchart of FIG. Note that the characteristic update process is started when the characteristic server 2 receives, from the information device 3 , received contents such as text data representing user utterances, position information of the information device 3 , or numerical data via the communication unit 21 .

First, in S<b>201 , the property server 2 receives received content from the information device 3 .
Subsequently, in S202, the property server 2 identifies classification information corresponding to the received content. Specifically, the property server 2 identifies the classification information as follows.

That is, when the received content is text data representing user utterances, the property server 2 extracts the content title (for example, "action movie X") from the text data (for example, "I want to see action movie X") that is the received content. do. Then, the property server 2 acquires meta information (content type×content genre in this embodiment) corresponding to the extracted content title (for example, “action movie X”) as classification information. Thereby, the classification information corresponding to the received content is specified. Note that in the present embodiment, a data table indicating correspondence between content titles and meta information is stored in the storage unit 23 in advance. The property server 2 identifies meta information (classification information) corresponding to the extracted content title by referring to the data table.

Also, when the received content is location information, the characteristic server 2 identifies the area indicated by the location information (for example, "Tokyo", "Nagoya", etc.) as classification information.
Further, when the received content is numerical data, the property server 2 identifies data indicated by the numerical data (for example, "0xFF") as classification information.

Subsequently, in S203, the property server 2 inputs the classification information (for example, content type x content genre) specified in S202, and acquires the added value of importance information as its output.

Note that, in the present embodiment, the added point value of the importance information corresponding to the classification information is set to a fixed point (for example, 5 points uniformly) regardless of the classification information.
Subsequently, in S204, if the keyword, location information, or numerical data corresponding to the received content is the keyword, location information, or numerical data set in the promotion information table, the property server 2 adds the point value acquired in S203 to To increase.

Specifically, the property server 2 refers to the promotion information table shown in FIG. 4 and determines whether or not the received content includes the keyword or the like set in the promotion information table. When the characteristic server 2 determines that the keyword or the like is included in the received content, it obtains the added point value of the importance information corresponding to the keyword or the like corresponding to the received content, and uses the added point value to calculate the score obtained in S203. replace. As a result, the added point value acquired in S203 is increased. For example, if the text data representing the user's utterance includes the keyword "action movie X", the score (5 points) obtained in S203 is increased to 40 points.

3A to 3C stored in the storage unit 23. Subsequently, in S205, the characteristic server 2 obtains the score of the importance information corresponding to the classification information specified in S202 in S203. The score is updated (increased) by the score or the score increased in S204. After that, the property server 2 terminates the property update processing of FIG.

[2-3. Character update process]
Next, character update processing executed by the control unit 38 of the information device 3 will be described with reference to the flowchart of FIG. Note that the character update process is started by activating the agent application in the information device 3 and ends by terminating the agent application.

First, in S301, the information device 3 acquires the property information table shown in FIGS. 3A to 3C from the property server 2 via the communication unit .
Subsequently, in S<b>302 , the information device 3 identifies the agent character 51 to be displayed on the display unit 31 based on the characteristic information table acquired in S<b>301 , and displays the identified agent character 51 on the display unit 31 . Specifically, the information device 3 identifies the agent character 51 displayed on the display unit 31 as follows.

That is, the information device 3 refers to the property information table and determines whether or not there is any classification information whose importance information has reached the threshold score. If it is determined that there is no classification information whose importance information score has reached the threshold score, that is, if it is determined that the importance information score for all classification information is less than the threshold score, the information device 3 is calculated, and it is determined whether or not the sum reaches a preset threshold value. For example, in the characteristic information table shown in FIG. 3A, the sum of the points of the importance information of the classification information of "music x pops", "movie x action", and "TV program x family" is 15 + 20 + 12 = 47 points. be. The information device 3 thus adds the points of importance information of each piece of classification information, and calculates the sum of the points of importance information of all the pieces of classification information.

When the information device 3 determines that the sum reaches the threshold value, the information device 3 specifies the character associated with the threshold value as the agent character to be displayed on the display unit 31 .

In this embodiment, a plurality of thresholds (a first threshold, a second threshold, etc.) are set for the sum, and the score of the importance information for any of the classification information is the threshold Before the score is reached, the agent character 51 is updated in multiple stages.

On the other hand, when the information device 3 determines that there is classification information whose importance information has reached the threshold score, the character corresponding to the classification information is displayed on the display unit 31 as an agent character. Identify. If there are a plurality of pieces of classification information whose scores have reached the threshold score, the character corresponding to the classification information with the highest score may be specified as the agent character to be displayed on the display section 31 .

For example, as shown in FIG. 8, before the points of the importance information for all the classification information reach the threshold points, the sum of the points of the importance information does not reach the minimum first threshold. In this case, the agent character 51 displayed on the display unit 31 is identified as the character A. Character A is the default (initial value) character when the agent application is installed. Then, when the sum of the points of the importance level information reaches the minimum first threshold value, the character B is identified as the agent character 51 displayed on the display unit 31 . Then, when the sum of the points of the importance level information reaches the second threshold value, which is the second largest after the first threshold value, the character C is identified as the agent character 51 displayed on the display unit 31. .

Subsequently, in S<b>303 , the information device 3 acquires the position information of the information device 3 from the position detection unit 37 and transmits the acquired position information to the property server 2 .
Subsequently, in S<b>304 , the information device 3 transmits the numerical data to the property server 2 . That is, in the present embodiment, the storage unit 35 stores, as numerical data, the scores of the results of questionnaires conducted in the past by the terminal user within the agent application. The information device 3 transmits these numerical data stored in the storage unit 35 to the property server 2 .

Subsequently, in S<b>305 , the information device 3 monitors whether or not an audio signal of user's speech is received from the terminal user via the microphone 33 .
When the information device 3 determines in S305 that it has received the voice signal of the user's utterance from the terminal user, the processing proceeds to S306.

In S<b>306 , the information device 3 transmits a voice signal representing the user's utterance to the voice recognition server 4 and receives text data representing the user's utterance from the voice recognition server 4 .
Subsequently, in S<b>307 , the information device 3 transmits the text data representing the user's utterance received in S<b>306 to the property server 2 and receives the corresponding response content from the property server 2 .

Subsequently, in S308, the information device 3 causes the agent character 51 to speak the content of the response. Here, the information equipment 3 causes the display unit 31 to display an animation showing the action of the agent character 51 stored in the storage unit 35, and responds with a voice suitable for the agent character 51 stored in the storage unit 35. The content is output from the speaker 34.

Subsequently, in S309, the information device 3 acquires the property information table shown in FIGS. 3A to 3C from the property server 2 via the communication unit . The characteristic information table is updated in the characteristic server 2 based on the position information transmitted in S303, the numerical data transmitted in S304, and the text data representing the user's utterance transmitted in S307.

Subsequently, at S310, the information device 3 determines whether or not a predetermined condition for updating the agent character 51 is satisfied based on the characteristic information table acquired at S309. Specifically, in the information device 3, the score of importance information in the most recent property information table acquired in S301 has not reached the threshold score, and the property information table acquired in S309 has no importance information. It is determined whether or not there is any classification information whose score has reached the threshold score. When the information device 3 determines that there is such classification information, it determines that the predetermined condition is met.

In addition, the information device 3 determines whether or not a predetermined condition is established as follows before the score of the importance level information reaches the threshold score for any of the classification information. That is, the information device 3 does not reach the threshold in the characteristic information table acquired most recently in S301, out of the thresholds relating to the sum of the points of importance information of all the classification information, In the characteristic information table acquired in S309, if there is a threshold such that the sum reaches the threshold, it is determined that the predetermined condition is satisfied.

When the information device 3 determines that the predetermined condition is not satisfied in S310, the processing returns to S305 described above.
On the other hand, when it is determined in S310 that the predetermined condition is satisfied, the information device 3 proceeds to S311 and updates the agent character 51 to the character corresponding to the classification information for which the predetermined condition is satisfied. In addition, the information equipment 3 updates the agent character 51 to a character corresponding to the threshold value that satisfies the predetermined condition before the score of the importance information reaches the threshold score for any of the classification information. . After that, the character update process of FIG. 7 is ended.

That is, in the dialogue agent system 1, as shown in FIG. 8, there are a plurality of update systems for the agent character 51 according to a plurality of types of classification information, such as a system for updating to character D and a system for updating to character E. exist. In other words, the process of updating the agent character 51 branches into tree rows.

[3. effect]
According to the embodiment detailed above, the following effects are obtained.
(1) In this embodiment, the control unit 38 of the information device 3 responds to input data acquired by the information device 3, including numerical data indicating user utterances, position information of the information device 3, and terminal user characteristics. A character is selected from among a plurality of types of characters based on this, and the agent character 51 is updated with the selected character. Therefore, the appearance of the agent character 51 is not updated, and the terminal user can continue to be interested, as compared with the interactive agent system in which the same agent character 51 repeatedly responds. As a result, new value can be provided to terminal users.

(2) In this embodiment, the control unit 24 of the property server 2 updates the property information stored in the storage unit 23 based on the input data acquired by the information device 3 . Then, the control unit 38 of the information device 3 selects a character from among a plurality of types of characters based on the updated property information, and updates the agent character 51 with the selected character.

Therefore, as the terminal user continues to use the dialogue agent system 1, the characteristic information stored in the storage unit 23 of the characteristic server 2 is updated in consideration of the terminal user's characteristic. As a result, the more the terminal user continues to use the dialogue agent system 1, the more the agent character 51 is updated to a character that takes into account the characteristics of the terminal user. Therefore, according to the configuration of the present embodiment, it is possible to further keep the terminal user interested in the system 1 .

(3) In this embodiment, the control unit 38 of the information device 3 determines whether or not a predetermined condition for updating the agent character 51 has been established for at least one of a plurality of types of classification information based on the input data. Determine (S310). Then, the control unit 38 updates the agent character 51 with the character associated with the classification information determined to satisfy the predetermined condition (S311).

Therefore, the agent character 51 is updated to the character corresponding to the classification information corresponding to the input data among the plurality of types of characters. Further, when the input data represents the type of information the terminal user wants to know, interests, etc., the agent character 51 can be updated to a character corresponding to the information, interests, etc. the terminal user wants to know. Therefore, it is possible to make the terminal user more interested.

(4) In this embodiment, when the cumulative score of the importance information of certain classification information reaches the threshold score, the control unit 38 of the information device 3 updates the agent character 51 with respect to the classification information. It is determined that a predetermined condition for is satisfied.

Therefore, when input data corresponding to certain classification information is input a plurality of times, the character corresponding to the classification information can be assigned to the character as an agent until the cumulative degree of points of the importance information related to the classification information reaches the threshold score. Character 51 is updated. Therefore, compared to the case where the agent character 51 is immediately updated to the character corresponding to the reception of specific input data, the agent character 51 can be updated to a character that takes into account the characteristics of the terminal user. can. As a result, it is possible to keep the terminal user interested.

(5) In this embodiment, the control unit 24 of the property server 2, when the input data includes a specific keyword or the like set in the promotion information table, Increase the score value. That is, in this embodiment, there are standard point addition values and special point addition values that are larger than the standard point values. In other words, when the first input data is accepted, the control unit 38 of the information device 3 stores the classification information corresponding to the first input data and the first addition value that is the standard addition value. (that is, the first amount of change) is specified. Further, when second input data different from the first input data is received, the control unit 38 sets the classification information corresponding to the second input data and the second point value different from the first addition value. A point addition value of 2 (that is, the second amount of change) is specified.

Therefore, when certain input data is accepted, the agent character 51 can be updated to the corresponding character more quickly. Therefore, depending on what kind of input data is input, the updated character of the agent character 51 and the speed until it is updated to the character differ. Therefore, it is possible to further improve the game characteristics of the dialogue agent system 1 , and eventually to keep the terminal user interested in the system 1 .

(6) In this embodiment, the input data includes user speech. Therefore, the updated character of the agent character 51 is selected based on the user's utterance. Therefore, the agent character 51 can be updated to a character that matches the terminal user's interest expressed by the user's utterance.

(7) In this embodiment, the input data includes position information of the information device 3 . Therefore, the updated character of the agent character 51 is selected based on the position information of the information device 3 and, by extension, the terminal user. That is, the update result of the agent character 51 is determined depending on where the terminal user is located. Therefore, the game performance of the dialogue agent system 1 can be further improved. Also, for example, when the terminal user visits a specific place, the agent character 51 can be easily updated to a specific character, and the frequency of visits to the place by the terminal user can be increased.

(8) In the present embodiment, the input data includes numerical data indicating terminal user characteristics. Therefore, the update result of the agent character 51 is selected based on the numerical data indicating the characteristics of the terminal user. That is, it is possible to update the agent character 51 to a character that takes into account the characteristics of the terminal user.

In this embodiment, the importance information corresponds to the degree of contribution, the score of the importance information corresponds to the cumulative degree of contribution, the added value of the importance information corresponds to the amount of change in contribution, and the characteristic information table corresponds to characteristic information. Further, S101 corresponds to processing as a user utterance receiving unit, S102 corresponds to processing as a response content acquisition unit, S202 and S203 correspond to processing as a specifying unit, and S203 corresponds to processing as a condition determination value determination unit. S205 corresponds to the processing of the property updating unit and the contribution updating unit, S302 corresponds to the processing of the display processing unit, S308 corresponds to the processing of the speech processing unit, S303, S304 and S305 corresponds to processing as an input data receiving unit, S310 corresponds to processing as an update determination unit, and S311 corresponds to processing as a character updating unit.

[4. Other embodiments]
Although the embodiments for implementing the present disclosure have been described above, the present disclosure is not limited to the above-described embodiments and can be implemented with various modifications.

(1) In the above-described embodiment, the predetermined condition for updating the agent character 51 is that the accumulated points of importance information reach the threshold score, but the predetermined condition is not limited to this. For example, the predetermined condition may be to receive user utterances containing a specific keyword from the user. Further, for example, the predetermined condition may be that the user charges in the agent application. In this case, the updated character of the agent character 51 may be selected depending on which of the plurality of types of classification information is to be billed or which of the plurality of types of characters is to be billed.

(2) In S203 of FIG. 6 of the above embodiment, the additional point value of the importance information and the condition determination value may be obtained as follows. Here, the condition determination value is a value for determining whether or not a predetermined condition for updating the agent character 51 is satisfied.

That is, a data table showing the correspondence relationship between the classification information and the scores of the importance information is created in advance, and the data table is stored in the storage unit 23 of the property server 2 . By referring to the data table stored in the storage unit 23, the control unit 24 may obtain the score of the importance information corresponding to the input classification information.

Further, for example, a condition decision value corresponding to specified classification information or input data may be determined using a judgment processing model generated by machine learning. Specifically, input data such as user utterances, position information of the information device 3, and numerical data indicating user characteristics, and condition determination values are used as training data to learn a determination processing model through machine learning. Alternatively, the classification information corresponding to the input data and the condition determination value are used as teaching data to learn the determination processing model by machine learning. Then, based on the input data or classification information, a judgment processing model is generated that outputs corresponding condition decision values. Then, using the learned determination processing model, the classification information identified in S202 of FIG. 6 and the condition determination value corresponding to the input data may be determined. The machine learning model to be used includes any machine learning model such as a model constructed using a neural network, a model constructed using a decision tree, a model constructed using a support vector machine, etc. be done.

Such a configuration provides the following effects. That is, when determining a condition determination value using a data table in which a correspondence relationship between input data or classification information corresponding to the input data and the condition determination value is set, input data not set in the data table Alternatively, if the classification information is input, it is assumed that the condition determination value cannot be output. On the other hand, by using a decision processing model generated by machine learning, even if input data or classification information not used in the teacher data of the decision processing model is input, the corresponding condition determination value can be output. can be done.

(3) Classification information is not limited to that of the above embodiment. For example, the content classification information may be only the content type. Further, for example, the classification information of the position information may be the coordinate values of the position of the information device 3 . Further, for example, the classification information of the numerical data may be numerical data having the meaning of the attributes of the merchandise such as barcodes of various merchandise and JAN codes.

(4) In the above embodiment, the classification information of the location information of the information device 3 is relatively large areas such as Tokyo and Nagoya, but the classification information of the location information is not limited to this. For example, the classification information of the location information may be a relatively narrow range (point, etc.) such as indoors of a specific facility.

(5) In the above embodiment, multiple threshold points may be set for one piece of classification information. In this case, for example, the higher the cumulative value of the points of the importance information with respect to the classification information, the more strongly the characteristic of the classification information may be updated to a character.

(6) In the above embodiment, all or part of the processing executed by the control unit 24 of the property server 2 may be executed by the control unit 38 of the information device 3 .
In particular, in the above embodiment, the characteristic information table is updated on the characteristic server 2 side, but the device that updates the characteristic information table is not limited to the characteristic server 2 . For example, the characteristic information table may be updated on the information device 3 side. Specifically, for example, the property server 2 stores a property information table that has not yet been updated according to the property of the terminal user, and the information device 3 acquires the property information table from the property server 2, 3 is stored in the storage unit 35. Then, the information device 3 may update the property information table stored in the storage unit 35 in consideration of the property of the terminal user.

Conversely, all or part of the processing executed by the control unit 38 of the information device 3 may be executed by the control unit 24 of the property server 2 .
(7) In the above-described embodiment, the functions realized by the speech recognition server 4 (such as the speech recognition function) may be realized in the information device 3 or the like.

(8) In the above embodiments, part or all of the functions executed by the control units 24 and 38 may be configured as hardware using one or a plurality of ICs or the like.
(9) In addition to the control units 24 and 38 described above, a system having the control units 24 and 38 as components, a program for causing a computer to function as the control units 24 and 38, a semiconductor memory storing this program, etc. The present disclosure can also be implemented in various forms, such as a non-transitional substantive storage medium, a method of updating an agent character to a character selected from a plurality of types of characters based on input data, and the like.

(10) A plurality of functions possessed by one component in the above embodiment may be realized by a plurality of components, or a function possessed by one component may be realized by a plurality of components. . Also, a plurality of functions possessed by a plurality of components may be realized by a single component, or a function realized by a plurality of components may be realized by a single component. Also, part of the configuration of the above embodiment may be omitted. Moreover, at least part of the configuration of the above embodiment may be added or replaced with respect to the configuration of the other above embodiment. It should be noted that all aspects included in the technical idea specified by the wording described in the claims are embodiments of the present disclosure.

DESCRIPTION OF SYMBOLS 1... Dialogue agent system, 2... Characteristic server, 3... Information equipment, 4... Voice recognition server,
51... Agent character.

Claims (8)

a user utterance accepting unit that accepts user utterances;
a response acquisition unit that acquires response content to the user's utterance;
a display processing unit for displaying an agent character, which is a character embodying a dialogue agent that interacts with a user, on a display unit of the user's information device;
an utterance processing unit that causes the agent character displayed on the display unit to utter the content of the response;
an input data reception unit that receives input data acquired by the information device;
a character updating unit that selects a character from a plurality of types of characters based on the input data and updates the agent character with the selected character;
an update determination unit for determining whether or not a predetermined condition for updating the agent character is established for at least one of the plurality of types of classification information based on the input data, wherein the plurality of types of classification information is an update determination unit that is information for classifying the input data, and that the plurality of types of classification information are associated with the plurality of types of characters;
with
The character update unit updates the agent character to a character associated with the classification information determined to satisfy the predetermined condition when the update determination unit determines that the predetermined condition is satisfied. is configured to
Content is provided from the agent character to the user,
The multiple types of characters include multiple types of characters according to the type of content provided to the user,
A dialogue agent system in which the plurality of types of classification information includes information representing the types of the content, and at least a part of the plurality of types of classification information is associated with a character corresponding to the type of the content. A dialogue agent system according to claim 1, comprising:
a property updating unit that updates property information based on the input data, wherein the property information is information representing the property of the user;
The character updating unit selects a character from among the plurality of types of characters based on the property information updated by the property updating unit, and causes the selected character to update the agent character. A dialogue agent system according to claim 1 or claim 2 ,
a specifying unit that specifies the classification information corresponding to the input data and a contribution degree change amount when the input data is received by the input data receiving unit, wherein the contribution degree change amount is the agent a specific part that is the amount of change in the cumulative degree of contribution to updating the character;
a storage unit that stores, in association with each of the plurality of types of classification information, the cumulative degree of contribution to updating of the agent character, wherein the contribution degree corresponding to the classification information specified by the specifying unit is stored; a contribution updating unit that updates the cumulative degree based on the contribution change amount specified by the specifying unit;
further comprising
The update determination unit determines that the predetermined condition has been established for the classification information when the cumulative degree of contribution of the classification information reaches a predetermined degree. A dialogue agent system according to claim 3 , wherein
The identifying unit identifies the classification information corresponding to the first input data and the first contribution degree change amount when the first input data is received by the input data receiving unit. , when second input data different from the first input data is received by the input data receiving unit, the classification information corresponding to the second input data and the first contribution change and a second said contribution variation amount different from the amount. A dialogue agent system according to any one of claims 1 to 4 ,
A judgment processing model generated by machine learning, which outputs a condition determination value that is a value for determining whether or not the predetermined condition is satisfied based on the input data or the classification information corresponding to the input data. further comprising a condition determination value determination unit that determines the condition determination value corresponding to the input data or the classification information corresponding to the input data using a determination processing model;
The dialogue agent system, wherein the update determination unit determines whether or not the predetermined condition is satisfied based on the condition determination value acquired by the condition determination value determination unit. A dialogue agent system according to any one of claims 1 to 5 ,
A dialog agent system, wherein the input data includes the user's speech. A dialogue agent system according to any one of claims 1 to 6 ,
A dialogue agent system, wherein the input data includes location information of the information device. A dialogue agent system according to any one of claims 1 to 7 ,
The dialogue agent system, wherein the input data includes numerical data indicating characteristics of the user.

Images