JP2017079004A - Agent dialogue system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an agent dialogue system that gives a person a feeling as if speech is produced with a stretched interpretation thereof.SOLUTION: A robot dialogue system 10 makes robots R1 and R2 do a dialogue therebetween, all based on a script. When a sensor manager 18 recognizes the intention of a person H, a group manager 20, with a person intention processing module 23, makes the robots R1 and R2 do such a dialogue that stretches an interpretation of the recognized human intention to approve it.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an agent interaction system, and more particularly to a novel agent interaction system in which a plurality of agents and at least one person participate, for example.

  Along with the spread of the robot field, robot research is shifting to research on robots that work in everyday situations, and the development of robots that can interact with people in the environment where people live is attracting attention. .

  With the recent development of speech recognition technology, robots that have the ability to interact with humans in spoken language have been developed, but no matter how good the speech recognition technology is, It is not easy to continue to give people who interact with robots the same feeling as “the feeling of dialogue (sense of participating in dialogue)” that people have when participating in dialogue with people. There wasn't.

  Patent Document 1 which is an example of background art discloses that in a dialogue between a robot and a person, the robot estimates a person's emotion and determines an utterance of a response sentence and an action associated with the utterance.

  In Patent Document 2, in a dialogue system between a human and a robot, a continuous and natural interaction is attempted by synchronizing the two.

JP 2004-90109 A [B25J 13/00 ...] JP 2012-181697 A [G06F 3/16 ...]

  In both the technology of Patent Literature 1 and the technology of Patent Literature 2, there is a limit to processing based on voice recognition, and it is not easy for a person to maintain the above-mentioned “interactive feeling”.

  Therefore, the main object of the present invention is to provide a novel agent interaction system.

  Another object of the present invention is to provide an agent dialogue system capable of sustaining a sense of participation in dialogue with a person.

  The present invention employs the following configuration in order to solve the above problems. The reference numerals in parentheses, supplementary explanations, and the like indicate the corresponding relationship with the embodiments described in order to help understanding of the present invention, and do not limit the present invention.

  A first invention is an agent dialogue system in which a plurality of agents each having an utterance function interact in accordance with a script at a dialogue place, and a person sits at the dialogue place, an intention recognition means for certifying a person's intention, and an intention An agent interaction system comprising execution means for causing at least one agent to execute at least one of an utterance and a non-verbal action for approving an intention of a person authorized by an authorization means.

  In the first invention, the agent dialogue system (10: reference numerals exemplifying corresponding parts in the embodiment; the same applies hereinafter) includes a plurality of agents each having a speech function (46) at the dialogue place (12). (R1, R2) are arranged, and a person (H) sits at the dialogue place (12). Agents interact according to a script. For example, an intention recognition means (18) such as a sensor manager recognizes the intention of the person H regardless of the actual intention of the person H. The execution means (20a, 23, S77, S83) causes at least one agent to execute an utterance and / or a non-language operation for approving the intention of the person. For example, when the intention of the person H is recognized, in step S83 of FIG. 15 which is an example of the execution means, not only the intention of the agent (R1, R2) is approved but also at least one agent (R1 and R1). The intention of the person H may be approved by the utterance of at least one of R2. Furthermore, in order to approve the intention of the authorized person H, the non-language operation in step S77 in FIG. 15 may be executed. In this case, step S77 also constitutes execution means.

  According to the first invention, since at least one agent executes an utterance that approves the intention of the authorized person, it is only necessary to cause a person to feel that he / she participates in a dialogue between agents (“sense of dialogue”). In addition, it is possible to give a person a feeling as if he / she uttered the premise of the approval.

  A second invention is an agent dialogue system according to the first invention, wherein the execution means includes a dialogue execution means for causing a plurality of agents to execute a dialogue in which a human intention is expanded and interpreted.

  According to the second aspect of the invention, it is possible to give a person a feeling as if he / she spoke an utterance with an expanded interpretation that is a precondition for approval.

  A third invention is dependent on the second invention, and the dialog execution means causes one agent to execute an utterance in which a person's intention is extended and executes an utterance in which at least one other agent approves the extended interpretation. It is an agent dialogue system that executes

  According to the third invention, one agent is caused to execute an utterance in which a person's intention is expanded and another at least one agent is allowed to execute an utterance that approves the expanded interpretation. It is possible to give a person a sense of utterance with an expanded interpretation.

  The fourth invention is an agent dialogue system (10) in which a plurality of agents (R1, R2) each having a speech function (46) interact according to a script at a dialogue place (12), and a person (H) sits in the dialogue place. The program is executed by a computer, and the program includes an intention recognition means (18) for certifying a person's intention to the computer, and an utterance and a non-language for certifying a person's intention authorized by the intention qualification means. It is a program of an agent dialogue system that functions as execution means (20a, 23, S77, S83) for causing at least one agent to execute at least one of operations.

  A fifth invention is dependent on the fourth invention, and the execution means is a program for an agent dialogue system including dialogue execution means for causing a plurality of agents to execute a dialogue in which human intentions are expanded and interpreted.

  In the fifth invention, as in the second invention, it is possible to give a person a feeling as if he / she has made an utterance with an expanded interpretation that is the premise for its approval.

  The sixth invention is dependent on the fifth invention, and the dialog execution means causes one agent to execute an utterance in which a person's intention is expanded and an utterance in which at least one other agent approves the expanded interpretation. This is an agent dialogue system program that executes

  In the sixth invention, as in the third invention, it is possible to give a person a stronger feeling as if he had made an utterance with an expanded interpretation that is the premise for approval.

  According to the present invention, even in the dialogue between agents, the person's action during the dialogue is accepted as a person's intention display regardless of the person's actual intention, thereby participating in the conversation to the person. A sense (“dialogue”) is generated, and a sense as if the utterance was accompanied by the approval can be given to the person.

  The above object, other objects, features, and advantages of the present invention will become more apparent from the following detailed description of embodiments with reference to the drawings.

FIG. 1 is a schematic diagram showing an outline of a robot dialogue system according to an embodiment of the present invention. FIG. 2 is a block diagram showing an example of the configuration of the sensor manager in the embodiment of FIG. FIG. 3 is a block diagram showing an example of the configuration of the group manager in FIG. 1 embodiment. FIG. 4 is a schematic diagram showing an example of the robot in the embodiment of FIG. FIG. 5 is a block diagram showing an example of the configuration of a robot controller that controls the robot of FIG. FIG. 6 is a schematic diagram showing an example of a script read from the dialog database shown in FIG. 1 to the group manager. FIG. 7 is a flowchart showing an example of the operation of the group manager shown in FIG. FIG. 8 is a flowchart showing an example of the operation of the robot controller shown in FIG. FIG. 9 is a schematic diagram showing an example of a dialogue sequence in the embodiment of FIG. FIG. 10 is a schematic diagram showing an example of a specific dialogue scene in the embodiment of FIG. FIG. 11 is a schematic diagram showing another example of a specific dialogue scene in the embodiment of FIG. FIG. 12 is a schematic diagram showing another example of a specific dialogue scene in the embodiment of FIG. FIG. 13 is a schematic diagram showing another example of a specific dialogue scene in the embodiment of FIG. FIG. 14 is a schematic diagram showing another example of a specific dialog scene in the embodiment of FIG. FIG. 15 is a flowchart showing an example of the operation of the human intention processing module shown in FIG. FIG. 16 is a schematic diagram showing an example of a specific dialog scene for processing a person's intention display in the FIG. 1 embodiment. FIG. 17 is a schematic diagram showing another example of a specific dialog scene for processing a person's intention display in the embodiment of FIG. FIG. 18 is a schematic diagram showing another example of a specific dialog scene for processing a person's intention display in the embodiment of FIG. FIG. 19 is a schematic view showing another example of a specific dialog scene for processing a person's intention display in the FIG. 1 embodiment. FIG. 20 is a schematic diagram showing another example of a specific dialog scene for processing a person's intention display in the FIG. 1 embodiment. FIG. 21 is a schematic view showing another example of a specific dialog scene for processing a person's intention display in the FIG. 1 embodiment. FIG. 22 is a schematic diagram showing another example of a specific dialog scene for processing a person's intention display in the FIG. 1 embodiment. FIG. 23 is a schematic diagram showing another example of a specific dialog scene for processing a person's intention display in the FIG. 1 embodiment. FIG. 24 is a schematic diagram showing another example of a specific dialog scene for processing a person's intention display in the FIG. 1 embodiment. FIG. 25 is a schematic view showing another example of a specific dialog scene for processing a person's intention display in the FIG. 1 embodiment.

  Referring to FIG. 1, a first robot R1 and a second robot R2 and one person H exist in a dialogue place 12 of the robot dialogue system 10 of this embodiment. However, the number of robots may be 3 or more, and the number of people may be 2 or more. Hereinafter, the first robot R1 and the second robot R2 may be simply referred to as the robot R when it is not necessary to distinguish between them.

  When a person speaks, if the other person's utterance is not generated in response to the utterance, it is difficult to continue the conversation itself, so a robot whose speech recognition and natural language processing capabilities are not perfect is human. It is not easy to continue to speak appropriately for the utterances, and the dialogue is likely to break down.

  On the other hand, humans can continue to listen to a dialogue while partially understanding or agreeing with it, even if the content is not completely understood or agreed when observing the dialogue between others. In addition, parts that could not be understood or agreed may be understood or agreed later. However, it is not easy to keep concentrating on the dialogue if there is no sense of participation (dialogue).

  Therefore, in this embodiment, when a robot speaks, it is not necessarily a form to make a human reply, but another robot is placed in the place of the dialogue, and an appropriate response to the robot's utterance is sent. Let the saddle robot speak, that is, let the robots communicate with each other to avoid failure, while letting people talk about the dialogue between the robots, and sometimes indicate the actions of humans, for example, "Agree", "Disagree" (“Yes”, “No”), etc., and accepting and confirming the intention of the person by continuing dialogue between robots or displaying non-verbal attention by multiple robots, Make humans feel a sense of dialogue.

  The robot interaction system as in this embodiment can be used as, for example, a communication service robot, an information providing service robot, an advertisement robot, etc. for a single person living alone.

  In this embodiment, a microphone 14 as an auditory sensor and a camera 16 as a visual sensor are provided in order to sense the situation of the dialogue place 12 of the robot interaction system 10. The microphones 14 are used to acquire environmental sounds including voices generated by the utterances of the robot R and the person H. If necessary, two or more microphones 14 may be installed. The camera 16 is also a camera that captures the situation of the dialogue place 12, particularly the facial expression and motion of the person H, and captures a moving image or a still image. If necessary, two or more cameras 16 may be installed.

  Further, in addition to the microphone 14 and the camera 16 described above, although not shown as a sensor, a living body such as a wearable posture sensor, an acceleration sensor, a heartbeat state, a respiratory state, a body motion (body movement) state, etc. A biological sensor that detects a signal, a motion capture system, or the like may be provided.

  Furthermore, in the dialogue system 10 of this embodiment, an intention display device 17 is provided as a means for expressing the intention of the person H. For example, the intention display device 17 includes an affirmation button 17a for inputting affirmation (Affirmation) or agreement (Agree) and a denial button 17n for inputting a negation (Negation) or disagreement (Disagree). However, the positive button 17a and the negative button 17n may be configured by hardware switches such as push button switches, or may be configured by software switches (virtual buttons) on the touch screen. The positive button 17a and the negative button 17n may be formed as individual switches or buttons.

  A voice signal acquired by the microphone 14, an image signal captured by the camera 16, an input signal detected by the intention display device 17 (affirmation, negative), a biological sensor that detects the above-described biological signal, and a person detected by the motion capture system The sensor signal such as the act or state change is input to the sensor manager 18. The sensor manager 18 acquires these sensor signals, determines the status of the dialogue place 12, and outputs the determination result as sensing data to the group manager 20 and the human intention processing module 23 described later.

  Referring to FIG. 2, the sensor manager 18 includes a CPU (Central Processing Unit) 18a, and a communication device 18c is connected to the CPU 18a via an internal bus 18b. The communication device 18c includes a network interface (NIC), for example, and the CPU 18a can communicate with the group manager 20 and the like via the communication device 18c, and can exchange data between them.

  A memory 18d is further connected to the CPU 18a via an internal bus 18b. The memory 18d includes a ROM and a RAM. For example, a sensor signal including an audio signal from the microphone 14 (FIG. 1) and an image signal from the camera 16 (FIG. 1) is input through a sensor I / F (interface) 18e configured by a DSP (Digital Signal Processor). . The memory 18d temporarily stores the sensor signal.

  The sensor manager 18 is a kind of determination device, and the CPU 18a determines the state of the dialogue place 12 and the intention of the person H based on the sensor data stored in the memory 18d. For example, an audio signal acquired by the microphone 14, an image signal captured by the camera 16, an input signal detected by the intention display device 17 (positive or negative), and a biological sensor or a motion capture system that detects the above-described biological signal are detected. It is possible to recognize and process sensor signals such as a person's actions and state changes as input of the intention expressed by the person H, regardless of the actual intention of the person H. Then, the sensor manager 18 sends data indicating the determined state to the group manager 20 and the human intention processing module 23.

  Further, programs (such as OS and sensor signal acquisition program) necessary for the sensor manager 18 are stored in the memory 18d. The sensor manager 18 operates according to a program stored in the memory 18d.

  In addition, a keyboard and a display (not shown) may be attached to the CPU 18a.

  The group manager 20 shown in FIG. 3 controls the speech operation (language operation: verbal operation) and behavior (non-verbal operation: nonverbal operation) of each of the two robots R, for example, according to the flowchart shown in FIG.

  The group manager 20 includes a CPU 20a, and a communication device 20c is connected to the CPU 20a via an internal bus 20b. The communication device 20c includes, for example, a network interface (NIC), and the CPU 20a can communicate with the sensor manager 18, the robot R, and the like via the communication device 20c, and can exchange data between them.

  Further, a memory 20d and an input device 20e are connected to the CPU 20a via an internal bus 20b. The memory 20d includes a ROM and a RAM. A script (dialog) is read from the dialog database 22 through the memory I / F 20f, and is temporarily stored in the memory 20d. However, “dialog” means a sequence of utterances and non-verbal movement commands to be performed during a dialogue, and the dialogue database 22 is a collection of dialogues (for example, what is important about dialogues like robots and robot care. Including a series of commands for dialogue on each topic, such as dialogue. “Script” is a character string representing a series of commands, and script data is a character string representing the commands. Therefore, the script data series is a script. Such a script is transmitted from the dialog queue 23c to the robots R1 and R2 by the group manager 20. However, the script data includes not only language data but also non-language data. The language data is script data for instructing the utterances of the robots R1 and R2, and the non-language data is, for example, the operation of the robots R1 and R2, seeing the person H, whispering, swaying the head, squeezing the neck, Script data for instructing non-language operations. In such script data, a description indicating that it is a phase (intention confirmation phase) for confirming the intention (explicit or implicit) of the person H is further added as necessary.

  Further, programs (such as OS and sensor signal acquisition program) necessary for the group manager 20 are stored in the memory 20d. The group manager 20 operates according to a program stored in the memory 20d.

  As described above, the CPU 20a of the group manager 20 controls the operation or behavior of each robot. The history of the behavior is stored in the memory 20d and provided to the sensor manager 18 as necessary. To do.

  Further, in the robot interaction system 10 shown in FIG. 1, a human intention processing module 23 is provided. The human intention processing module 23 may be an independent computer, but is configured as one program of the group manager 20 described above in this embodiment. Therefore, the human intention processing module 23 is controlled by the CPU 20a.

  As will be described in detail later with reference to FIG. 15, the human intention processing module 23 does not necessarily depend on the actual intention of the person H, but the intention of the person H who is recognized by the sensor manager 18, for example, positive or negative. Accordingly, script data (dialog) for the robots R1 and R2 to speak in order to deal with the intention of the authorized person H is read from the dialog database 22 and stored in the pool 23a, which is an area of the memory 20d. Let (load).

  The out-of-context response dialog pool 23b is loaded into a future group of dialogs or dialog queues 23c loaded into the context or next dialog candidate pool 23a and is exceptional, ie out-of-context, that was not scheduled for a dialog in progress. This is a storage area for storing a group of scripts that should execute a speech or non-verbal operation in order to cope with the intention of the person H. The intention display outside the context is, for example, the intention button 171a and / or the negative button 17n of the above-mentioned intention display device 17 repeatedly, regardless of the utterance waiting for the person's intention display and ignoring the intention forever. It is conceivable that no display is made.

  The dialog queue 23c is also an area in the memory 20d, for example. In this dialog queue 23c, script data loaded in the next dialog candidate pool 23a and script data loaded in the out-of-context response dialog pool 23b. Can then be loaded in a queue so that robots R1 and / or R2 can immediately execute.

  In the script, for example, the execution time (tnext) of the script data included therein is written as a header. In the dialog queue 23c, the script data is always sorted according to the execution time (tnext). The script data is sent to each robot controller 24 so that the script data having the same time is simultaneously executed. Therefore, for example, the robots R1 and R2 can simultaneously perform the same operation, for example, seeing the person H, and the same robot R1 or R2 can see the other robot or the person H at the same time as speaking, for example.

  Referring to FIG. 4, FIG. 4 shows the appearance of the robot R according to the embodiment, and the robot R is provided on the table 30 so as to be able to rotate forward, backward, left and right with respect to the table 30. In other words, the body 32 has two degrees of freedom.

  From the left and right positions corresponding to the shoulders of the body 32 of the human body, the right arm 34R and the left arm 34L are provided so as to be rotatable forward and backward and left and right by shoulder joints (not shown). That is, two degrees of freedom are set for each of the right arm 34R and the left arm 34L.

  A neck 36 is provided at the center of the upper end of the body 32, and a head 38 is provided thereon. The neck 36, that is, the head 38 is attached to the body 32 so as to be able to rotate back and forth and right and left. In other words, the neck 36, that is, the head 38 is set with three degrees of freedom: a roll angle (left-right tilt), a pitch angle (front-back tilt), and a yaw angle (left-right rotation).

  A right eye 40R and a left eye 40L are provided on the front surface of the head 38, that is, a surface corresponding to a human face, and eyeballs 42R and 42L are provided on the right eye 40R and the left eye 40L. The right eye 40R and the left eye 40L can close and open the eyelids, and the eyeballs 42R and 42L can rotate up, down, left and right, respectively. That is, one degree of freedom is set for the right eye 40R and the left eye 40L, that is, the eyelid, and two degrees of freedom are set for the eyeballs 42R and 42L.

  The face is further provided with a mouth 44, which can be closed or opened. That is, one degree of freedom is set for the mouth 44.

  At the position of the chest of the human body 32, a speaker 46 that makes a speech for the person H to hear in the dialogue system 10 and a microphone 48 that listens to the environment, particularly the voice of the person H, are provided.

  A camera 50 capable of shooting a moving image or a still image is built in a portion corresponding to the forehead of the head 38. The camera 50 can photograph the person H who is facing, and the camera signal (video signal) from the camera 50 is sent to the CPU 18a via the sensor I / F of the sensor manager 18 in the same manner as the environmental camera 16 (FIG. 1). , May be entered.

  FIG. 5 is a block diagram showing a robot controller 24 which is built in the robot R and controls the operation of the robot R (such as speech and gesture). Referring to FIG. 5, the robot controller 24 includes a CPU 24a, and a communication device 24c is connected to the CPU 24a via an internal bus 24b. The communication device 24c includes, for example, a network interface (NIC), and the CPU 24a communicates with the sensor manager 18, the group manager 20, and an external computer or other robot (both not shown) via the communication device 24c. And data can be exchanged between them.

  Further, a memory 24d is connected to the CPU 24a via an internal bus 24b. The memory 24d includes a ROM and a RAM. Control data and script data sent from the group manager 20 are temporarily stored in the memory 24d.

  Further, programs (such as OS and sensor signal acquisition program) necessary for robot control are stored in the memory 24d. The robot controller 24 controls the operation of the robot R according to the program stored in the memory 24d.

  That is, the CPU 24a of the robot controller 24 is further connected to an actuator control board 24e configured by, for example, a DSP. The actuator control board 24e is an actuator provided in each of the above-described parts of the robot R as described below. To control the operation.

  Two-degree-of-freedom movement of the body 32, that is, forward / backward / left / right rotation, is controlled by the CPU 24a controlling the body actuator 52 through the actuator control board 24e.

  The two-degree-of-freedom movement of the right arm 34R and the left arm 34L, that is, the front / rear / left / right rotation, is controlled by the CPU 24a controlling the arm actuator 54 through the actuator control board 24e.

  The movement of the neck 36, that is, the head 38, with three degrees of freedom, that is, the rotation in the front, rear, left, and right directions is controlled by the CPU 24a by the head actuator 56 through the actuator control board 24e.

  The opening / closing operation of the right eye 40R and the left eye 40L, that is, the eyelid is controlled by the CPU 24a controlling the eyelid actuator 58 through the actuator control board 24e. The two-degree-of-freedom movement of the eyeball 42R and the eyeball 42L, that is, the front / rear / left / right rotation, is controlled by the CPU 24a controlling the eyeball actuator 60 through the actuator control board 24e. The opening / closing operation of the mouth 44 is controlled by the CPU 24a controlling the mouth actuator 62 through the actuator control board 24e.

  A speaker 46 of the robot R shown in FIG. 4 is connected to the CPU 24a of the robot controller 24. The CPU 24a utters (speaks) from the speaker 46 according to the script data given from the group manager 20 and stored in the memory 24d as necessary.

  The robot controller 24 moves the head and arms of the robot R when necessary in the dialogue system 10, for example, when a non-language operation is requested by a script. In the following description, Specific control of the actuator and the like is not necessarily described because it can be easily estimated from the above description.

  As shown in FIG. 1, each robot R <b> 1 and R <b> 2 is provided with a built-in robot sensor 26 similarly to the robot controller 24. The robot sensor 26 includes an attitude sensor, an acceleration sensor, and the like for detecting the states of the movable components of the robots R1 and R2. Sensor signals from these sensors are input to the sensor manager 18. Therefore, the sensor manager 18 can sense the states of the robots R1 and R2 based on the sensor signal from the robot sensor 26.

  Note that the microphone 48 and the camera 50 of the robot R shown in FIG. 4 are input to the sensor manager 18 via the robot sensor 26. The sensor manager 18 stores the voice data captured from the microphone 48 in the memory 18d (FIG. 2), and executes voice recognition processing as necessary. The sensor manager 18 also processes the camera signal from the camera 50 to sense the status of the interaction location 12.

  Although only one sensor manager 18 is shown in the embodiment of FIG. 1, two or more arbitrary sensor managers may be provided. In this case, each sensor manager is a sensing item. Can be shared.

  Similarly, if necessary, two or more group managers 20 may be used. Conversely, the sensor manager 18 and the group manager 20 may be realized by a single computer.

  The robot R used in the dialogue system 10 of FIG. 1 embodiment is not limited to the robot described above with reference to FIG. 4 and may have at least a function capable of speaking according to a script.

  FIG. 6 is a schematic diagram showing an example of a script with which the robot R interacts accordingly. In this script, first, a portion (time zone) where the robots R interact with each other and a portion (time zone) where the robot R asks the person H (question or request for consent) are clearly distinguished. In the embodiment, the flag for that is “question flag Fah (FLAG_ASK_HUMAN)”.

  When the question flag Fah is reset, that is, when “Fah = 1” is not satisfied, the robots R output utterances predetermined by the script data from the speaker 46 (FIGS. 4 and 5), Tell the participant H, the participant, about the content of the dialogue. When the question flag Fah is set, that is, when “Fah = 1”, the robot R outputs an utterance sentence predetermined by the script data from the speaker 46 (FIGS. 4 and 5) and participates. Speak to person H who is a person.

  In the example of FIG. 6, the upper part (A) and the lower part (C) are script data for dialogue between the robots, and the middle part (B) asks the person H from the robot R. Script data for

  In the first line (a1), second line (a2), and third line (a3) of the script of part (A), “tnext = XX” is defined. Here, “XX1”, “XX2”, “XX3” defines the time or time at which each script data is to be executed. “RR” describes that the robot speaks to the robot, and “SPEAKER = R1 (R2)” and “LISTNERR = R2 (R1)” that follow it indicate that the speaker at that time is the first robot R1 ( Or the second robot R2) and the listener is defined as the second robot R2 (or the first robot R1).

  The description following “TEXT” in the script data is an utterance sentence that the robot should utter at that time. Then, the description following “NONVERBAL” defines non-linguistic operations to be performed by the first robot R1 and / or the second robot R2, for example, seeing the opponent robot or seeing the person H.

  In part (B), since the question flag Fah is set, that is, “Fah = 1”, one of the two robots R1 and R2, in this example, from the first robot R1 to the person H , The utterance of the question is output. Therefore, in the first line (b1), “RH: SPEAKER = R1, LISTNER = H” indicating that the first robot R1 speaks to a person follows the description “tnext = XX4” that defines time. Described. “TEXT” and “NONVERBAL” are as described above.

  Below the first line (b1) of part (B), there are script data of condition 1, condition 2, and condition 3, but these are the people who answered the question from robot R in the first line (b1). It is defined that when H speaks, the content (response) that the robot speaks next changes according to the condition.

  The condition 1 is script data in which the robots talk to, for example, “interesting” about the utterance from the person H. Therefore, following the description “tnext = XX5” defining the time, the second robot R2 is the second one. “RR: SPEAKER = R2, LISTNER = R1” indicating that one robot R1 speaks is described.

  Conditions 2 and 3 are both set so that the robot reacts to the utterance of the person H, and both follow the description “T = XX5” that defines the time, and then the second robot R2 or the first robot. “RH: SPEAKER = R2, LISTNER = H” or “RH: SPEAKER = R1, LISTNER = H” indicating that R1 speaks to a person is described.

  The utterance sentence of the condition 2 or the condition 3 is determined by the group manager 20 according to the result of the voice recognition by the sensor manager 18 recognizing the voice from the microphone 14. It is predefined as data. For example, in response to the question from the robot, “What do you like?” In the first line (b1), when a person H utters “I like XX”, the robot Replied, “I like XX” or “I like △△ more than XX”. Both are utterances as a result of the sensor manager 18 recognizing “OO” included in the utterance of the person H. For this reason, in both the condition 2 and the condition 3, an utterance sentence excluding “OO” is defined in advance.

  Part (C) is similar to Part (A), because the question flag Fah is reset, that is, it is not `` Fah = 1 ''. ) Progresses regardless of the utterance from person H. Since the contents of the script data are as described in Part (A), redundant description will not be given here.

  The time or time tnext for determining the utterance timing in the embodiment of FIG. 6 is updated (shifted) by the group manager 20 based on an instruction from the sensor manager 18, for example, when there is an utterance interruption (interrupt) from the person H. ) May be necessary. For example, when the sensor manager 18 manages a flag (not shown) indicating whether or not the speech timing needs to be updated, and the group manager 20 is set (when “1”), T is rewritten. You can do it.

  In the robot dialogue system 10 of FIG. 1, the two robots R1 and R2 at the dialogue place 12 proceed with the dialogue according to the script illustrated in FIG. 6, but comprehensively control the utterances of these robots R1 and R2. It is the group manager 20 that does this.

  The flowchart shown in FIG. 7 shows the operation of the CPU 20a (FIG. 3) of the group manager 20. The operation of FIG. 7 is repeatedly executed at a speed of about the frame rate, for example.

  In the first step S1, the CPU 20a executes initialization such as reading script data as exemplified in FIG. 6 from the dialog database 22 (FIG. 1).

  In subsequent step S2, the CPU 20a updates the sensor data sent from the sensor manager 18 (FIGS. 1 and 3).

  In the sensor manager 18, based on the sensor signal from the sensor such as the microphone 14 or the camera 16 shown in FIG. 1, the non-linguistic parameter “NONVERBAL_PARAMETERS”, for example, “the eye does not calm down”, “yawned”, etc. Sensing. Further, as described above, the sensor manager 18 can also receive a signal indicating the intention of the person from the intention display device 17 and determine what kind of intention the person has expressed.

  Further, the sensor manager 18 is also based on the sensor signal, “whether there is a human voice”, “whether someone is speaking”, “interrupted by language” or “the robot responds non-verbally” Sensing whether or not a person is generating a behavior to be performed.

  In addition, information necessary to determine detailed parameters of robot behavior (for example, when looking at person H, the position of the person, when talking to the person, the inflection or end point of the person's utterance) ) Must be sensed.

  In addition, it is necessary to sense how individual robots behaved, such as “speaking,” “waiting,” “listening,” “sitting,” and so on. Sensing can be performed by input (for example, from the group manager 20) from software that manages the history of robot behavior.

  In step S2, sensing result data (sensing data) from the sensor manager 18 is updated.

  In the next step S3, the CPU 20a determines whether or not the dialog queue 23c has been updated. If “NO” in the step S3, the process proceeds to a step S5. If “YES”, the current time (t) is set to “0” and the execution time (tnext) is set to “−1” in step S4.

  When “NO” is determined in step S3, in the next step S5, based on the sensing data updated in step S2, the CPU 20a of the group manager 20 determines whether or not there is an interruption due to the utterance of the person H.

  When “YES” is determined in the step S5, the CPU 20a determines whether or not the speech recovery flag IS_RECOVER = Frc is set in the next step S7, that is, whether Frc = 1. Here, “YES” means that the robot R has returned from the utterance collision caused by the utterance of the person detected in step S5, and the robot R needs to insert some utterance. Therefore, in step S9, the script data for the return phase for returning from the dialog interruption state due to the interruption due to the utterance of the person H read in step S1 is loaded, and broadcast to the robot R, that is, the robot controller 24 in step S11. Therefore, after that, the robot R speaks according to the script data for the return phase.

  However, the script data to be loaded in step S9 (the same applies to steps S17 and S19 described later), as described in FIG. 6, is “ID_SPEAKER” who is the speaker, “ID_LISTNER” who is the listener, ) Includes information such as “ID_BYSTANDER”, utterance time “ONSET_TIME_SPEECH”, and non-language action “NONVERBAL_PARAMETERS”.

  For example, in response to a human speech interrupt, the first robot R1 utters “soka” to the person, the second robot R2 utters “first?” To the first robot, the first robot R1 speaks to the second robot R2 “Yes, (next utterance determined by script)”. In other words, an utterance that is light enough to respond to a human utterance interruption is inserted. Human utterances are not ignored, but the robot does not respond accurately to the utterances, but only returns the responses determined by the script.

  Then, after step S11, the process returns to step S2.

  If “YES” in the step S13, that is, if it is determined that the next utterance time predetermined by the script has been reached, the CPU 20a determines whether or not the above-described question flag Fah is set in a step S15, that is, to the person H. It is judged whether or not the question sentence is uttered.

If “YES” in the step S15, the script data for the inquiry phase stored in the dialog queue 23c is loaded from the head in the step S17, and is broadcast in the step S11.
In other words, human responses are not ignored.

  If “NO” is determined in step S15, that is, if it is determined not to speak a question to a person, the CPU 20a determines script data that is not for questioning (non-questioning script data in the next step S19). ).

  In step S21, the CPU 20a determines whether a script still remains. That is, it is determined whether the dialogue flag IS_CONVERSATION is reset. If "YES", the process proceeds to step S11. If "NO", the fact is broadcast in step S23, and the end process is performed in step S25, and this routine ends.

  Both the robots R1 and R2 are controlled by the robot controller 24 in accordance with the flowchart of FIG. The routine of FIG. 8 is also repeatedly executed at the frame rate, for example, similarly to the routine of FIG.

  In step S31, the CPU 24a of the robot controller 24 executes initialization. In the next step S33, the flag or data “IS_CONVERSATION (not yet interactive) received from the sensor manager 18 and broadcast from the group manager 20 as shown in FIG. ”,“ IS_ONSPEECH (who speaks) ”,“ CAN_BE_RECOVERED ” ”,“ ID_SPEAKER (who speaks) ”,“ ID_LISTNER (who listens) ”,“ ID_BYSTANDER (who is seated) ”,“ SPEECH_TEXT ”,“ ONSET_TIME_SPEECH ”,“ NONVERBAL_PARAMETERS ” Parameters: Specify non-verbal actions such as when to look at or whisper. That parameter) "the subject.

  In step S <b> 34, the CPU 24 a refers to the data “IS_CONVERSATION” and determines whether the dialogue still continues. If “YES”, in the next step S35, the CPU 24a of the robot controller 24 determines whether or not the data broadcast from the group manager 20 received in step S33 includes non-language operation script data. If “YES” is determined in the step S35, the CPU 24a issues an operation command to the actuator control board 24e of the corresponding robot R1 and / or R2 so as to execute the non-language operation command in the next step S36. give. Therefore, the robot R1 and / or R2 performs the broadcasted non-language operation.

  If “NO” in the previous step S35, it is determined in a next step S37 whether or not someone is speaking by referring to the data “IS_ONSPEECH”. If “YES”, it is determined in step S39 whether or not the speaker at that time is himself (the robot to be controlled). That is, it is determined whether or not the user is speaking by referring to the data “ID_SPEAKER”.

  If the user (the robot controlled by) is speaking, “YES” is determined in the step S39, and the process proceeds to the step S41. In step S41, the CPU 24a refers to the data “IS_INTERRUPTED_VERBAL” to determine whether or not there is a language interrupt (utterance blocking). If “YES” in the step S41, it means that there is an interrupt in a language. In the next step S43, the CPU 24a refers to the data “CAN_BE_RECOVERED” and determines whether or not the speech can be resumed.

  If “YES” in the step S43, the robot resumes the utterance in a step S46, but if “NO”, the robot is made to wait without causing the robot to speak in a step S46.

  If “NO” is determined in the step S41, that is, if there is no language-based interrupt, the CPU 24a determines whether or not the non-language operation is interrupted in a step S47. For example, it is determined whether or not the person has irregular non-language behavior, for example, the person H is moving greatly.

  “When the robot determines that the robot should speak while generating a non-verbal response, such as when the person H is performing irregular non-verbal movements or is listening to the robot's talk Is “YES” in step S47, and in step S48, the CPU 24a causes the robot R1 or R2 controlled by the CPU 24a to respond by a non-language operation based on the data “NONVERBAL_PARAMETERS”. If “NO” in the step S47, the process by the CPU 24a returns to the step S33.

  If “NO” in the step S39, that is, if the user (the robot controlled by) is not instructed to speak, the data “ID_LISTNER” is referred to and it is determined whether or not he is designated as a listener.

  If “YES” in the step S49, the CPU 24a controls the robot R controlled by the CPU 24a to be a listener in a step S51.

  If “NO” in the step S49, the data “ID_BYSTANDER” indicates that the robot R1 or R2 is a seated person, so that the robot controller 24, that is, the CPU 24a operates as a seated person on the robot R1 or R2. To do.

  When “NO” is determined in step S37, that is, when no one is speaking, in step S55, the CPU 24a refers to the data “ID_SPEAKER” and determines whether or not he is a speaker. If “YES”, control is performed so that the robot R1 or R2 controlled by the robot speaks in accordance with the content of the utterance sentence TEXT in step S57.

  If “NO” in the step S55, since the robot is neither a speaker nor a listener, the CPU 24a controls the robot to operate as a seated person in a step S59.

  If “NO” in the step S34, an end process is executed in a step S61, and then this routine is ended.

  As described above, the robot R only utters the content determined in the script at the timing determined in the script broadcast from the group manager 18 and does not utter spontaneously.

  On the other hand, the robot R does not interpret the speech from the person H. Even if an utterance comes from a person H, the robot R only hits the hand and does not respond to the utterance from that person. The talk also speaks as decided in the script.

  If such a dialogue system of the robots R1 and R2 as scripted by the robot controller 24 is used, a dialogue sequence is established as shown in FIG.

  For example, you can clearly distinguish between a robot talking to a person and a robot talking to a robot, and at the same time neglecting humans and robots who are occupants in the conversation Based on the history of conversation and the intention estimation of the participants, gestures such as utterances, gaze distribution, and interaction are generated for dialogue between robots and inquiries to people. That is, since the robot R can hear the utterance of the person H, it can ask the person H to ask the person H to speak. This makes it possible to make people feel like they are also participating in dialogues between robots. As a result, the contents included in the dialogue between the robots can be impressed on the person.

  Specifically, the scenes shown in FIGS. 10 to 14 can be intentionally made to appear by a script. However, the specific examples of FIGS. 10 to 14 are not directly related to the intention display of the person H.

  In the example of FIG. 10, when the first robot R1 and the second robot R2 are talking about the topic (Topic or theme) A while repeating the utterance, when switching from the topic A to the topic B. The robot R1 and / or R2 controls the person H to ask a question on the script. The content of the question at this time may be, for example, related to the topic A that has already ended, or may be related to the topic B that starts from now. It may be about both. Then, when there is an answer from the person H, the robot R1 and / or R2 utters a conflict such as “Sod” according to the script.

  That is, the robot does not ask the person H during the dialogue between the robots, and the robot speaks to the person H at the timing of topic switching. If it does so, the person H can maintain the state with the consciousness which is participating in the dialogue between robots.

  However, the robot may interrupt the response from the person H. For example, robot R2 “XX is XX, so it is △△”, robot R1 “Yes, but XX is □□?”, Robot R2 “So, but (But into topic B) Can also be interrupted.

  In the example of FIG. 11, when the robots R1 and R2 are engaged in a conversation on a certain topic while repeating questions and answers, the robot asks the person H the same question that the robot has made. For example, the robot R1 and / or R2 causes the person H to ask a question such as "How are you?" Then, there is an answer from person H. In response to this, the robots R1 and R2 speak the same conflict as in FIG.

  In the example of FIG. 12, contrary to the case of FIG. 11, the same question that the robot R1 or R2 has made the person H is the other robot R2 or R1 when the robots are interacting with each other. In this case, the person H can have a “dialogue feeling” by including the same question as the question to himself / herself in the dialogue between the robots.

  In the example of FIG. 13, when the robots R1 and R2 are engaged in a conversation on a certain topic while repeating questions and answers, for example, the second robot R2 explains something to the first robot R1, and from the first robot R1 Make an utterance that describes your thoughts. The robot R1 or R2 asks the person H for an agreement, for example, “Ne” so that the person H agrees to the impression. In this case as well, if it is a story that can be convinced to some extent, it is possible to utter consent or encourage consent in the mind, so that person H is aware that he is participating in a dialogue between robots. Can last.

In the example of FIG. 14, for example, the second robot R2 to the first robot R1
The person asked, “I ’m not right” or “I ’m not right” by asking a question or explanation for the first robot R1. Encourage them to make corrections and supplementary utterances. However, if there is no utterance from a person, the second robot R2 corrects or supplements the utterance of the first robot instead. Since people actively participate in the dialogue, the feeling of dialogue is further enhanced.

  When the intention display by the person H who is a seated person detected by the intention display device 17 is performed, the human intention processing module 23 follows the flowchart shown in FIG. 15 from the dialog database 22 to the next dialog candidate pool 23a or out of context. A dialog (script data) is loaded as a response, and further a dialog (script data) is loaded from the next dialog candidate pool 23a into the dialog queue 23c.

  Specifically, in the first step S71 in FIG. 15, the CPU 20a that controls the human intention processing module 23 executes initialization and reads a dialog (script data) from the dialog database 22 into the next dialog candidate pool 23a. That is, a topic or theme dialog to be read is determined, and a group of dialogs of the topic or theme is read into the next dialog candidate pool 23a. At this time, it also loads constraints such as dialog dependencies, for example, which dialog should not be executed before, or that this dialog should be executed at the same time as a specific dialog. At the same time, in step S71, necessary dialogs are loaded from the next dialog candidate pool 23a into the dialog queue 23c.

  The dialog at this time is set so that, for example, one robot R1 is uttered or operated consistently in favor of the topic, and the other robot R2 is uttered or operated consistently in the opposite position. Each robot will not change its position from beginning to end.

  In the next step S73, the CPU 20a updates the sensor data from the sensor manager 18.

  In the next step S75, the CPU 20a determines whether or not the intention of the person to be approved is detected in the updated sensor data.

  When “YES” is determined in step S75, that is, when there is a person's intention that the robot R1 and / or R2 must approve, in step S77, the CPU 20a displays an approval dialog (mainly non-display) for the person's intention expression. Language action) is loaded into the dialog queue 23c.

  However, if the approval dialog interferes with each robot's scheduled dialog (including non-verbal actions), the approval dialog is not loaded. For example, if it is scheduled to see the person H next to any robot, if the approval dialog is “the robot sees the person H”, the approval dialog is not placed in the dialog queue 23c. . Further, when a dialog for viewing another robot is scheduled for one of the robots, if the approval dialog is “the robot sees the person H”, the robot immediately sees the other robot and the person immediately The robot must see H, and the robot performs an unnatural operation (interference). Accordingly, at this time, the approval dialog is not loaded into the dialog queue 23a.

  That is, in step S77, only dialogs that do not interfere with other dialogs are loaded into the dialog queue 23c as approval dialogs that approve human intentions. Therefore, according to the approval dialog loaded in the dialog queue 23c, the intention of the person H detected in step S75 is approved by, for example, a non-language operation such as “the robots R1 and R2 both see the person H”.

  If “NO” is determined in step S75, that is, if the intention of the person to be approved by the robot is not detected, the process proceeds to step S79 as it is.

  In step S79, it is determined whether it is a phase for confirming a person's intention. For example, if the previous utterance is in the form of a question sentence or a negative sentence for the person H, the intention of the person should be confirmed in the question sentence or the negative sentence script, and the timing to be confirmed is set (for example, In step S79, the question flag Fah in FIG. 6 is checked.

  In step S81, it is determined whether or not the intention of the person H has been recognized. For example, if it is clear that the person H has input the intention by operating the affirmative button 19a or the negative button 19n of the intention display device 17, the CPU 20a can recognize the intention of the person. On the other hand, when the person H does not operate the intention display device 17 and ignores the question sentence, the CPU 20a cannot definitely recognize the person's intention.

  In such a case, although not shown in FIG. 15, the human intention processing module 23 loads a dialog necessary for confirming the intention of the person H to the dialog queue 23 c, and one of the robots R speaks. The determination of the intention of the person H may be promoted by non-linguistic action or both. For example, the intention of the person H is confirmed by a non-linguistic action in which the robot R1 and / or R2 sees the person H and / or by an utterance such as “What do you think” of the robot R1 and / or R2? can do. Then, after executing such a decision confirmation dialog, it is only necessary to return to the previous step S73 and determine whether or not the intention can be recognized again.

  However, when there is no intention expression from the person H for a predetermined time, for example, 5 seconds, it can be considered that the intention of the person H is either “affirmation” or “denial”.

  If the intention of the person H can be definitely determined in this way, the process proceeds from step S81 to the next step S83. In step S83, the CPU 20a controls the human intention processing module 23 to select a dialog (script data) for approving the intention from the next dialog candidate pool 23a based on the intention of the authorized person H. To the dialog queue 23c. Accordingly, at this time, the robots R1 and R2 execute a dialog for approving the intention of the person H in accordance with the approval dialog loaded in the dialog queue 23c. As described above, the dialog at this time is loaded into the dialog queue 23c depending on whether the person H made a positive intention or a negative intention.

  In step S85, since the CPU 20a has loaded the dialog into the dialog queue 23c in step S83, the CPU 20a reads the next dialog candidate from the dialog database 22 and loads it into the next dialog candidate pool 23a.

  In the previous step S79, if it is not the person intention recognition phase, “NO” is determined, and in the next step S87, the CPU 20a determines whether or not the sensor manager 18 has detected no intention outside the context of the person H that cannot be ignored. To do. If there is no intention display outside the context of the person H, “NO” is determined in the step S87, and the process proceeds to the next step S89.

  In step S89, it is determined whether to continue the dialogue. For example, when there are no dialogs that can be employed in the dialog database 22 or when there is no time left, “NO” is determined in the step S89, and the operation of the human intention processing module 23 in FIG.

  When “YES” is determined in the step S89, that is, when the dialogue is to be continued, in a step S91, it is determined whether or not the dialog to be executed remains in the dialog queue 23c, that is, whether or not the dialog has ended.

  If “YES” in the step S91, similarly to the step S71, in a step S93, a group of dialogs (script data) of another topic is selected and read from the dialog database 22, and loaded into the next dialog candidate pool 23a. At the same time, the dialog loaded in the next dialog candidate pool 23a is loaded into the dialog queue 23c for the next dialogue between the robots R1 and R2. Thereafter, the process proceeds to the previous step S85, and further dialogs are loaded from the dialog database 22 into the next dialog candidate pool 23a.

  If “NO” in the step S91, if the dialog is not ended, the process returns to the previous step S73.

  Furthermore, when it is determined “YES” in step S87, that is, when there is an intention display outside the context of the person H. In step S95, the CPU 20a selects an appropriate dialog among the dialogs loaded in the out-context response dialog pool 23b shown in FIG. 1 based on the intention of the person H outside the context at that time, and the selected dialog is displayed. Is loaded into the dialog queue 23c. Therefore, the robot R1 and / or R2 executes the utterance and / or the non-verbal action in response to the intention expression of the person H outside the context described above.

  In the next step S97, the CPU 20a deletes or holds the contents of the current dialog queue 23c and the next dialog candidate pool 23b based on the intention of the person. That is, based on the intention of the person H at that time, a dialog that is deemed unnecessary is deleted from the dialog queue 23c and the next dialog candidate pool 23b, or the execution of the dialog that is considered unnecessary is suspended.

  As described above, in the human intention processing module 23, for example, the robot R 1 and R 2 utterances or non-languages for approving the intention of the person H who is a seated person expressed using the intention display device 17. By executing the operation, in the dialogue system 10 including the robots R1 and R2, the person H clearly has a sense (interaction feeling) participating in the dialogue even if nothing is spoken. It reaches.

  Hereinafter, various examples of the approval dialog for the intention display of the person H will be described with reference to FIGS. However, it should be noted that all dialogs are executed according to a dialog (script data) preset in the dialog database 22 on the assumption of the following.

  “Topic” surrounded by an ellipse is the “topic” described in FIGS. 10 to 14, and Topic-1, Topic-2,... Are different topics. Topic indicates utterances in favor, and “Anti-Topic” surrounded by a shaded oval indicates utterances in the opposite position to Topic. “Follow-Topic” indicates an utterance in a position to follow Topic. Furthermore, in the person H, “affirmation” indicated by a white circle means that the sensor manager 18 has detected an affirmative intention of the person H, for example, when the affirmation button 17a of the intention display device 17 is operated. To do. On the other hand, “denial” indicated by a shaded circle means that the sensor manager 18 has detected an intention to deny the person H. In the description of the specific examples below, “person's intention display” is not limited to the intention that person H has explicitly or implicitly displayed, but is not necessarily the actual intention of person H. Regardless, it should be understood that this refers to the intention to be certified by the sensor manager 18 as described above.

  In the example of FIG. 16, with the initiative of dialogue, the robot R1 speaks in favor of Topic-1, and the robot R2 speaks in the opposite position to the robot R1.

  In accordance with the dialog after the utterance of the question sentence included in Topic-1 (Anti-Topic-1), the CPU 20a of the group manager 20 causes the robots R1 and R2 to execute a non-language operation of watching the person H together. The script data (instruction for the non-language operation) held in the queue 23c is broadcast. Accordingly, the person H understands that his / her intention is required.

  After the person H makes an affirmative intention in response to such non-linguistic operations of the robots R1 and R2, for example, in step S83 of FIG. 15, the robots R1 and R2 are shown by a horizontally long rectangular frame in FIG. Execute the approval dialog. In this example, since the person H has made a positive intention expression, the robot R1 speaks, for example, “That's right” for approval of the intention expression. On the other hand, since the robot R2 is in the opposite position and the person H has made an affirmative intention, for example, “Sokana” is an affirmative intention of the person H to approve the intention. Speaks that condemn the display.

  That is, here, the robots R1 and R2 execute utterances that approve the positive intention display of the person H in their respective positions.

  Note that the utterance in this approval dialog may be either the utterance of robots R1 and R2, and the utterances of robots R1 and R2 may be executed simultaneously. Further, in this approval dialog, the dialogue between the robots R1 and R2 may be executed not only once but a plurality of times as in the embodiment. It should be pointed out in advance that these facts also apply to the following examples in FIGS.

  After that, the robot R1 having the initiative speaks a connection utterance for returning to the conversation in the same Topic-1 or another Topic-2, for example, “Yes, that's it”. Thereafter, the robots R1 and R2 continue Topic-1 (or 2) (Anti-Topic-1 (or 2)).

  In the example of FIG. 16, the robot R1 approves the positive intention expression of the person H by performing the utterance synchronized with the positive intention expression of the person H, and the utterance that the robot R2 does not synchronize with the positive intention expression of the person H. Approves the positive intention expression of person H by doing.

  In the example of FIG. 17, with the initiative of dialogue, the robot R1 speaks in a position in favor of Topic-1, and the robot R2 speaks in a position following the robot R1. After the utterance of Topic-1 (Follow-Topic-1), the robot R1 and R2 perform a non-linguistic action in which both the robots R1 and R2 see the person H, thereby prompting the person H to express his intention. .

  Then, when the person H gives an affirmative intention, for example, in step S83 of FIG. 15, the robots R1 and R2 execute an approval dialog as shown by a horizontally long rectangular frame in FIG. In this example, the robot R1 speaks, for example, “That's right” for the approval of the intention display because the person H has made a positive intention display. On the other hand, since the robot R2 is in a position to follow the robot R1, the robot R2 speaks, for example, “Yes, right” for the approval of the positive intention display of the person H.

  That is, here, the robots R1 and R2 execute utterances that approve the positive intention display of the person H in their respective positions.

  Thereafter, the robot R1 having the initiative takes a synchronized utterance such as “Yes, that's right” to return to the conversation in the same Topic-1 or another Topic-2. Thereafter, the robots R1 and R2 continue Topic-1 (or 2) and Follow-Topic-1 (or 2).

  In the example of FIG. 17, the robot R1 approves the positive intention display of the person H by performing an utterance synchronized with the positive intention display of the person H, and the robot R2 in the position following the robot R1 also confirms the positive of the person H. The positive intention expression of the person H is approved by performing an utterance synchronized with the intention expression.

  In the example of FIG. 18, with the initiative of dialogue, the robot R1 speaks in favor of Topic-1, and the robot R2 speaks in the opposite position to the robot R1. After the utterance of Topic-1 (Anti-Topic-1), as in the case of FIG. 16, the robots R1 and R2 perform a non-language operation in which the person H sees the person H, thereby prompting the person H to express his intention. .

  After the person H makes a negative intention, for example, in step S83 of FIG. 15, the robots R1 and R2 execute an approval dialog as shown by a horizontally long rectangular frame in FIG. In this example, since the person H has made a negative intention, the robot R1 utters the intention to deny the intention of the person H, for example, “Yes, different” in order to approve the intention. On the other hand, since the robot R2 is in the opposite position and the person H has expressed a negative intention, the person who speaks, for example, “Da Nene” is synchronized with the negative intention. Approve the intention to deny person H.

  That is, here, the robots R1 and R2 execute utterances that approve the intention to deny the person H from their respective positions.

  After that, the robot R1 having the initiative has a connection utterance for returning to the conversation in the same Topic-1 or another Topic-2, for example, a reversal utterance such as “No, but it is”. Thereafter, the robots R1 and R2 continue Topic-1 (or 2) (Anti-Topic-1 (or 2)).

  In the example of FIG. 18, the robot R1 approves the intention of denying the person H in a position of approval by performing an utterance that does not synchronize with the intention expression of the person H, and the robot R2 performs an utterance that does not synchronize with the intention expression of the person H. Approves the intention to deny person H from the opposite standpoint.

  In the example of FIG. 19, with the initiative of dialogue, the robot R1 speaks in favor of Topic-1, and the robot R2 speaks in the opposite position to the robot R1. After the utterance of Topic-1 (Anti-Topic-1), as in the case of FIG. 16, the robots R1 and R2 perform a non-language operation in which the person H sees the person H, thereby prompting the person H to express his intention. .

  After the person H gives an affirmative intention, for example, in step S83 of FIG. 15, the robots R1 and R2 execute an approval dialog as shown by a horizontally long rectangular frame in FIG. In this example, the robot R1 speaks, for example, “That's right” for the approval of the intention display because the person H has made a positive intention display. On the other hand, since the robot R2 is in the opposite position and the person H has made an affirmative intention, for example, “Sokana” is an affirmative intention of the person H to approve the intention. Speaks that condemn the display.

  That is, here, the robots R1 and R2 execute utterances that approve the positive intention display of the person H in their respective positions.

  Thereafter, the initiative of the dialogue is transferred to the robot R2, and the robot R2 utters a connection utterance for returning to the dialogue in Topic-1 (or 2), for example, “No, that's it”. Thereafter, the robots R2 and R1 continue Topic-1 (or 2) (Anti-Topic-1 (or 2)).

  In the example of FIG. 19, the robot R2 approves the positive intention expression of the person H in a position of approval by performing an utterance synchronized with the positive intention expression of the person H, and the robot R2 displays the positive intention expression of the person H. Approves the positive intention of person H from the opposite standpoint by uttering out of sync.

  In the example of FIG. 20, with the initiative of dialogue, the robot R1 speaks in favor of Topic-1, and the robot R2 speaks in the opposite position to the robot R1. After the utterance of Topic-1 (Anti-Topic-1), as in the case of FIG. 16, the robots R1 and R2 perform a non-language operation in which the person H sees the person H, thereby prompting the person H to express his intention. .

  After the person H makes a negative intention, for example, in step S83 of FIG. 15, the robots R1 and R2 execute an approval dialog as shown by a horizontally long rectangular frame in FIG. In this example, since the robot R1 has made a negative intention expression, the person R speaks, for example, “No, different” in order to approve the intention expression. On the other hand, since the robot R2 is in the opposite position and the person H has expressed his intention to deny it, for example, "Don't?" Speaks in sync with the intention.

  That is, here, the robots R1 and R2 execute utterances that approve the intention to deny the person H from their respective positions.

  Thereafter, the initiative of the dialogue is transferred to the robot R2, and the robot R2 utters a connection utterance for returning to the dialogue in the same Topic-1 (or 2), for example, “No, but it”. Thereafter, the robots R2 and R1 continue Topic-1 (or 2) (Anti-Topic-1 (or 2)).

  In the example of FIG. 20, the robot R2 approves an intention to deny the person H from a position of approval by performing an utterance that does not synchronize with the intention to deny the person H, and the robot R2 displays an intention to deny the person H. Approves the intention to deny person H from the opposite position by performing a synchronized utterance.

  In the example of FIG. 21, with the initiative of dialogue, the robot R1 speaks in favor of Topic-1, and the robot R2 speaks in the opposite position to the robot R1. After the utterance of Topic-1 (Anti-Topic-1), as in the case of FIG. 16, the robots R1 and R2 perform a non-language operation in which the person H sees the person H, thereby prompting the person H to express his intention. .

  After the person H gives an affirmative intention, the robot R1 executes, for example, a dialog for enlarging and approving the person H's affirmative intention in step S83 of FIG. First, since the person H makes an affirmative intention and is in a position of approval, he speaks, for example, “That's right” so as to be in line with the person H. And the utterance which expands and interprets the positive intention display of the person H is performed with respect to the robot R2.

  For example, if Topic-1 is a topic about luxury goods, when Robot R1 speaks “Banana is delicious” in Topic-1, when Robot H1 makes a positive intention, Robot R1 First speaks to person H, "That's right." After that, toward the robot R <b> 2, the positive intention expression of the person H is interpreted and expanded, for example, “I like fruit”. On the other hand, the robot R2 speaks to the robot R1 to reinforce and approve the enlarged interpretation of the robot R1, for example, “Do you like fruit?”.

  That is, when the person H makes an affirmative intention with respect to the conversation between the robots R1 and R2 in the Topic, the robot R1 who has the initiative of the dialogue expands and approves the affirmative intention expression of the person H. By executing the dialog and subsequently executing the dialog in which the robot R2 approves the expanded interpretation, the person H feels that the expanded interpretation has actually been spoken. In the above example, the person H has a feeling that he / she actually uttered “I like fruits”.

  Thereafter, the robots R1 and R2 execute an approval dialog as shown by a horizontally long rectangular frame in FIG. In this example, since the robot R2 has approved the expanded interpretation of the robot R1 with respect to the positive intention of the person H, the robot R1 responds to the approval of the expanded interpretation. The robot R2 speaks to the robot R2 because, for example, the person H gives an affirmative intention, and as an opposite position, speaks, for example, "Terrible" that condemns the affirmative intention of the person H. .

  That is, here, the robots R1 and R2 execute utterances that approve the positive intention display of the person H on the premise of the expanded interpretation in their respective positions.

  Note that the utterance in this approval dialog may be either the utterance of robots R1 and R2, and the utterances of robots R1 and R2 may be executed simultaneously. Further, in this approval dialog, the dialogue between the robots R1 and R2 may be executed not only once as in the embodiment but also a plurality of times, or may be omitted. It should be pointed out in advance that these facts also apply to the following examples in FIGS.

  Thereafter, the robot R1 having the initiative speaks a connection utterance for starting an assertion in the same Topic-1 or another Topic-2, for example, “Yes, that's it”. Thereafter, the robots R1 and R2 continue Topic-1 (or 2) (Anti-Topic-1 (or 2)).

  In the example of FIG. 21, the robot R1 and R2 approves the human H's affirmative intention expression in an expanded manner, and then the robot R1 makes an utterance that synchronizes with the human H, so that the human H's positive affirmation The display is approved, and the robot R2 approves the positive intention expression of the person H in the opposite position by performing an utterance that is not synchronized with the positive intention expression of the person H.

  In the example of FIG. 22, with the initiative of dialogue, the robot R1 speaks in favor of Topic-1, and the robot R2 speaks in a position following the robot R1. After the utterance of Topic-1 (Follow-Topic-1), the robot R1 and R2 perform a non-linguistic action in which both the robots R1 and R2 see the person H, thereby prompting the person H to express his intention. .

  After the person H gives an affirmative intention, the robot R1 executes, for example, a dialog for enlarging and approving the person H's affirmative intention in step S83 of FIG. First, since the person H expresses a positive intention and is in a position to follow, the person H speaks, for example, “That's right” so as to synchronize with the person H. And the utterance which expands the positive intention display of the person H is performed with respect to the robot R2.

  For example, when Topic-1 is the topic of luxury goods, when Robot R1 speaks “Banana is delicious” in Topic-1, and when person H makes a positive intention, R1 first speaks to person H, "That's right." After that, toward the robot R <b> 2, the positive intention expression of the person H is interpreted and expanded, for example, “I like fruit”. On the other hand, the robot R2 speaks to the robot R1 to reinforce and approve the expanded interpretation of the robot R1, for example, “I understand the story”.

  In other words, when the person H makes an affirmative intention for the dialogue between the robots R1 and R2 in the Topic (Follow-Topic), the robot R1 having the initiative in the dialogue expands the affirmative intention of the person H. By executing a dialog for interpreting and approving, and subsequently executing a dialog for robot R2 to approve the expanded interpretation, person H feels that the expanded interpretation has actually been spoken. In the above example, the person H has a feeling that he / she actually uttered “I like fruits”.

  Thereafter, the robots R1 and R2 execute an approval dialog as shown by a horizontally long rectangular frame in FIG. In this example, since the robot R2 approves the expanded interpretation of the robot R1 for the positive intention expression of the person H, the robot R1 responds to the approval of the expanded interpretation. The robot R2 is in sync with the positive intention display of the person H as a position to follow the robot R1, because the robot R2 has expressed the positive intention. For example, say "Yes".

  That is, here, the robots R1 and R2 execute utterances that approve the positive intention display of the person H on the premise of the expanded interpretation in their respective positions.

  Thereafter, the robot R1 having the initiative speaks a connection utterance for returning to the conversation in the same Topic-1 or another Topic-2, for example, “Yes, that's it”. Thereafter, the robots R1 and R2 continue Topic-1 (or 2) (Anti-Topic-1 (or 2)).

  In the example of FIG. 22, the robot R1 and R2 approve the human H's positive intention expression in an expanded manner, and then the robot R1 makes an utterance that synchronizes with the human H, so that the human H's positive intention The display is enlarged and interpreted and approved, and the robot R2 also approves the positive intention display of the person H from the standpoint of following the robot R1 by uttering in synchronization with the positive intention display of the person H.

  In the example of FIG. 23, with the initiative of dialogue, the robot R1 speaks in favor of Topic-1, and the robot R2 speaks in a position opposite to the robot R1. After the utterance of Topic-1 (Anti-Topic-1), as in the case of FIG. 16, the robots R1 and R2 perform a non-language operation in which the person H sees the person H, thereby prompting the person H to express his intention. .

  After the person H makes a negative intention, the robot R2 executes a dialog for enlarging and approving the negative intention of the person H, for example, in step S83 of FIG. First, since the person H expresses a negative intention and he is in the opposite position, he speaks, for example, “That's right” so as to be in line with the person H. And the utterance which expands the intention display of the negative of the person H is performed with respect to the robot R2.

  For example, when Topic-1 is a topic about luxury goods, when Robot R1 speaks "Banana is delicious" in Topic-1, and when person H makes a negative intention, the robot R2 first speaks to person H, "That's right." Then, toward the robot R1, the intention expression of the negative of the person H is expanded and interpreted, for example, “I hate fruit”. On the other hand, the robot R1 speaks to the robot R2 to reinforce and approve the expanded interpretation of the robot R2, for example, “I hate fruit?”.

  In other words, when the person H makes a negative intention to the dialogue between the robots R1 and R2 in the topic (Anti-Topic), the robot R2 opposed to the robot R1 expands the interpretation of the negative intention of the person H. Then, by executing the dialog to approve and subsequently executing the dialog in which the robot R1 approves the enlarged interpretation, it feels that the person H has actually spoken the enlarged interpretation. In the above example, the person H has a feeling that he / she actually uttered “I hate fruit”.

  Thereafter, the robots R2 and R1 execute an approval dialog as shown by a horizontally long rectangular frame in FIG. In this example, since the robot R1 approves the expanded interpretation of the robot R2 for the negative intention expression of the person H, the robot R2 responds to the approval of the expanded interpretation. The robot R1 is in a position opposed to the robot R2 because the person H has expressed a negative intention, and the robot R1 does not synchronize with the negative intention of the person H. It ’s terrible. ”

  That is, here, the robots R1 and R2 execute utterances that approve the intention to deny the person H on the premise of the expanded interpretation in their respective positions.

  Thereafter, the robot R1 having the initiative utters a connection utterance for returning to the conversation in the same Topic-1 or another Topic-2, for example, “No, but it”. Thereafter, the robots R1 and R2 continue Topic-1 (or 2) (Anti-Topic-1 (or 2)).

  In the example of FIG. 23, the robots R1 and R2 approve the interpretation of the intention of denying the person H, and then the robot R2 makes an utterance in sync with the person H, thereby deciding the intention of the person H in the opposite position. The display is expanded and interpreted, and the robot R1 approves the intention to deny the person H by performing an utterance that is not synchronized with the intention to deny the person H.

  In the example of FIG. 24, with the initiative of dialogue, the robot R1 speaks in favor of Topic-1, and the robot R2 speaks in the opposite position to the robot R1. After the utterance of Topic-1 (Anti-Topic-1), as in the case of FIG. 16, the robots R1 and R2 perform a non-language operation in which the person H sees the person H, thereby prompting the person H to express his intention. .

  After the person H gives an affirmative intention, the robot R1 executes, for example, a dialog for enlarging and approving the person H's affirmative intention in step S83 of FIG. First, since the person H expresses an affirmative intention and is in a position of approval, he speaks, for example, “That's right” so as to be in line with the person H. And the utterance which expands the positive intention display of the person H is performed with respect to the robot R2.

  For example, when Topic-1 is the topic of luxury goods, when Robot R1 speaks “Banana is delicious” in Topic-1, and when person H makes a positive intention, R1 first speaks to person H, "That's right." After that, toward the robot R <b> 2, the positive intention expression of the person H is interpreted and expanded, for example, “I like fruit”. On the other hand, the robot R2 speaks to the robot R1 to reinforce and approve the enlarged interpretation of the robot R1, for example, “Do you like fruit?”.

  That is, when the person H makes an affirmative intention with respect to the conversation between the robots R1 and R2 in the Topic, the robot R1 who has the initiative of the dialogue expands and approves the affirmative intention expression of the person H. By executing the dialog and subsequently executing the dialog in which the robot R2 approves the expanded interpretation, the person H feels that the expanded interpretation has been spoken. In the above example, the person H has a feeling that he / she actually said that he likes fruits.

  Thereafter, the robots R1 and R2 execute an approval dialog as shown by a horizontally long rectangular frame in FIG. In this example, since the robot R2 has approved the expanded interpretation of the robot R1 with respect to the positive intention of the person H, the robot R1 responds to the approval of the expanded interpretation. The robot R2 speaks to the robot R2 because, for example, the person H gives an affirmative intention, and as an opposite position, speaks, for example, "Terrible" that condemns the affirmative intention of the person H. .

  That is, here, the robots R1 and R2 execute utterances that approve the positive intention display of the person H on the premise of the expanded interpretation in their respective positions.

  Thereafter, the initiative shifts to the robot R2, and the robot R2 utters a connection utterance for starting a claim in Topic-1 (or 2), for example, “No, but it is”. Thereafter, the robots R2 and R1 continue Topic-1 (or 2) (Anti-Topic-1 (or 2)).

  In the example of FIG. 24, the robot R1 and R2 approves the human H's positive intention expression in an expanded manner, and then the robot R2 makes an utterance synchronized with the human H, and the human H's positive intention in the opposite position. The display is expanded and interpreted, and the robot R1 approves the positive intention expression of the person H by performing an utterance that is not synchronized with the positive intention expression of the person H.

  In the above-described embodiment, the intention display of the person H is input in accordance with, for example, the operation of the intention display device 17. However, for example, the image signal from the camera 16 may be processed, and for example, if the person H hits, an “affirmation” intention display may be provided, and if the head is shaken, a “no” intention display may be provided.

  In the example of FIG. 25, with the initiative of dialogue, the robot R1 speaks in favor of Topic-1, and the robot R2 speaks in a position opposite to the robot R1. After the utterance of Topic-1 (Anti-Topic-1), as in the case of FIG. 16, the robots R1 and R2 perform a non-language operation in which the person H sees the person H, thereby prompting the person H to express his intention. .

  After the person H makes a negative intention, the robot R2 executes a dialog for enlarging and approving the negative intention of the person H, for example, in step S83 of FIG. First, since the person H expresses a negative intention and he is in the opposite position, he speaks, for example, “That's right” so as to be in line with the person H. And the utterance which expands the intention display of the negative of the person H is performed with respect to the robot R2.

  For example, when Topic-1 is a topic about luxury goods, when Robot R1 speaks "Banana is delicious" in Topic-1, and when person H makes a negative intention, the robot R2 first speaks to person H, "That's right." Then, toward the robot R1, the intention expression of the negative of the person H is expanded and interpreted, for example, “I hate fruit”. On the other hand, the robot R1 speaks to the robot R2 to reinforce and approve the expanded interpretation of the robot R2, for example, “I hate fruit?”.

  In other words, when the person H makes a negative intention to the dialogue between the robots R1 and R2 in the topic (Anti-Topic), the robot R2 opposed to the robot R1 expands the interpretation of the negative intention of the person H. Then, by executing the dialog to approve and subsequently executing the dialog in which the robot R1 approves the enlarged interpretation, it feels that the person H has actually spoken the enlarged interpretation. In the above example, the person H has a feeling that he / she actually uttered “I hate fruit”.

  Thereafter, the robots R2 and R1 execute an approval dialog as shown by a horizontally long rectangular frame in FIG. In this example, since the robot R1 approves the expanded interpretation of the robot R2 for the negative intention expression of the person H, the robot R2 responds to the approval of the expanded interpretation. The robot R1 is in a position opposed to the robot R2 because the person H has expressed a negative intention, and the robot R1 does not synchronize with the negative intention of the person H. It ’s terrible. ”

  That is, here, the robots R1 and R2 execute utterances that approve the intention to deny the person H on the premise of the expanded interpretation in their respective positions.

  After that, the initiative shifts to the robot R2, and the robot R2 speaks the continuation of the assertion start in Topic-1 (or 2), for example, “No, but it is”. Thereafter, the robots R2 and R1 continue Anti-Topic-1 (or 2) (Topic-1 (or 2)).

  In the example of FIG. 25, after the robots R1 and R2 approve the interpretation of the intention of denying the person H, the intention of the person H to deny it in the opposite position by the robot R2 performing an utterance synchronized with the person H. The display is expanded and interpreted, and the robot R1 approves the intention to deny the person H by performing an utterance that is not synchronized with the intention to deny the person H.

  In the specific examples shown in FIGS. 16 to 25, when the intention of the person H is recognized, the intention is approved by the dialogue between the robots R1 and R2. However, step S83 in FIG. In the approval dialog shown by (1), the intention of the person H may be approved by the utterance of at least one agent (in the embodiment, at least one of the robots R1 and R2).

  Furthermore, in order to approve the intention of the authorized person H, the non-language operation in step S77 in FIG. 15 may be executed. In this case, step S77 also constitutes execution means.

  In the embodiment described above, each robot R1 and the dialog database 22 are stored in advance. However, instead of the database 22, for example, necessary script data may be sequentially supplied to the group manager 20 from the net.

  Furthermore, although the above-described embodiment is a robot interaction system using a robot that is a physical agent, the present invention is not limited to such a physical agent, but may be an avatar displayed on a display screen, for example. Alternatively, a dialogue system between agents such as characters may be applied. In this case, the robot controller 24 and the robot sensor 26 in FIG. 1 are replaced with a display controller (not shown) for displaying such avatars and characters. That is, the present invention is an agent dialogue system.

DESCRIPTION OF SYMBOLS 10 ... Robot dialogue system 12 ... Dialog place R1, R2 ... Robot 17 ... Intention display device 18 ... Sensor manager 20 ... Group manager 22 ... Dialog database 23 ... Human intention processing module 24 ... Robot controller

Claims (6)

An agent dialogue system in which a plurality of agents each having an utterance function interact according to a script in a dialogue place, and a person sits in the dialogue place,
Agent dialogue comprising: an authorization means (18) for authorizing a person's intention; and an execution means for causing at least one agent to execute at least one of an utterance and a non-verbal action for approving the intention of the person authorized by the intention authorization means. system.   The agent execution system according to claim 1, wherein the execution means includes a dialog execution means for causing a plurality of agents to execute a dialog in which the intention of the person is expanded and interpreted.   The agent dialogue according to claim 2, wherein the dialogue execution unit causes one agent to execute an utterance in which the intention of the person is expanded and executed, and causes at least one other agent to execute an utterance that approves the extended interpretation. system. A program executed by a computer of an agent interaction system in which a plurality of agents each having an utterance function interact according to a script at a conversation place, and a person sits at the conversation place, A program for an agent dialogue system that functions as an execution means for causing at least one agent to execute at least one of an utterance and a non-linguistic action for approving an intention of a person authorized by the intention recognition means.   5. The agent dialogue system program according to claim 4, wherein the execution means includes dialogue execution means for causing a plurality of agents to execute a dialogue in which the intention of the person is expanded and interpreted.   6. The agent dialogue according to claim 5, wherein the dialogue execution unit causes one agent to execute an utterance obtained by expanding the interpretation of the person's intention, and causes another at least one agent to execute an utterance that approves the extended interpretation utterance. System program.