JP7044156B2 - Generation program, generation method and information processing device - Google Patents

Description

The present invention relates to a generation program, a generation method and an information processing apparatus.

Conventionally, in order to realize natural communication between a robot and a human, information for knowing a user's hobbies and preferences may be collected. As a method of collecting information for knowing a user's hobbies and preferences, for example, there is a method of asking a user a question from a robot in a chat format and collecting the user's answer to the question.

Prior art, for example, selects an action response or a confirmed utterance response based on a certainty function and generates an acknowledgment when the confirmed utterance response is selected. Further, there is a technique for calculating a sentence PN value that quantifies the emotion that the user has regarding the dialogue based on the PN value set for the proper noun keyword or the event keyword included in the dialogue. In addition, a predetermined word is extracted from a sentence, a vector representing the characteristics of the word generated using a plurality of sentences is clustered, and a representative word that is close to the center of gravity of each cluster and a word included in each cluster are used. There is a technique to replace the extracted word with a representative word by using a table that stores the correspondence of. In addition, the text data is acquired by processing the multimodal input signal, the user's intention is acquired by the text data, the answer corresponding to the user's intention is acquired, and the answer is converted into the multimodal output signal. There is a technology to output.

Japanese Unexamined Patent Publication No. 2011-54088 Japanese Unexamined Patent Publication No. 2017-37601 Japanese Unexamined Patent Publication No. 2017-27168 Japanese Unexamined Patent Publication No. 2017-10517

However, in the prior art, the information generated from the answers to the questions may deviate from the actual situation of the respondents. For example, a question may be answered by the respondent as "like", but rather like it, not very much. In this case, if the information generated from the answer is treated in the same way as when the answer is "I really like it", it will be different from the actual situation of the respondent.

In one aspect, it is an object of the present invention to suppress the deviation of the generated information from the actual state of the respondents.

In one embodiment, the content information indicating the content of the utterance made by the respondent to the question and the appearance status of the filler as the state information indicating the state of the respondent at the time of the utterance are accepted, and the appearance of the filler is received. Refer to the question-specific ambiguity distribution information that records the ambiguity distribution corresponding to the situation, specify the ambiguity of the received content information based on the appearance status of the received filler, and use the received content information. , A generation program for generating information associated with the specified information according to the degree of ambiguity is provided.

According to one aspect of the present invention, it is possible to suppress the deviation of the generated information from the actual state of the respondent.

FIG. 1 is an explanatory diagram showing an embodiment of a generation method according to an embodiment. FIG. 2 is an explanatory diagram showing a system configuration example of the data collection system 200. FIG. 3 is a block diagram showing a hardware configuration example of the information processing apparatus 101. FIG. 4 is a block diagram showing a hardware configuration example of the robot 201. FIG. 5 is an explanatory diagram showing an example of the stored contents of the question-specific ambiguity distribution DB 220. FIG. 6 is an explanatory diagram showing an example of the stored contents of the user-specific ambiguity distribution DB 230. FIG. 7 is an explanatory diagram showing an example of the stored contents of the learning data DB 240. FIG. 8 is a block diagram showing a functional configuration example of the information processing apparatus 101. FIG. 9 is an explanatory diagram showing an example of detection of a section in which a filler appears. FIG. 10 is an explanatory diagram showing an example of calculating an emotion value representing the emotion of the respondent. FIG. 11 is an explanatory diagram showing an example of calculating a value representing the behavior of the respondent. FIG. 12 is an explanatory diagram (No. 1) showing a specific example of the ambiguity degree v. FIG. 13 is an explanatory diagram (No. 2) showing a specific example of the ambiguity degree v. FIG. 14 is an explanatory diagram (No. 3) showing a specific example of the ambiguity degree v. FIG. 15 is an explanatory diagram showing an example of correction of the ambiguity degree v. FIG. 16 is an explanatory diagram showing a specific example of the status determination question list 1600. FIG. 17 is a flowchart showing an example of the user-specific ambiguity information creation processing procedure of the information processing apparatus 101. FIG. 18 is a flowchart (No. 1) showing an example of the generation processing procedure of the information processing apparatus 101. FIG. 19 is a flowchart (No. 2) showing an example of the generation processing procedure of the information processing apparatus 101.

Hereinafter, embodiments of a generation program, a generation method, and an information processing apparatus according to the present invention will be described in detail with reference to the drawings.

(Embodiment)
FIG. 1 is an explanatory diagram showing an embodiment of a generation method according to an embodiment. In FIG. 1, the information processing apparatus 101 identifies the degree of ambiguity of content information indicating the content of an answer given by a respondent to a certain question, and associates the content information with information according to the specified degree of ambiguity. It is a computer that generates information.

The question may be a spoken question or may be a question of how to solve some problem. The answer is the answer to the question. For example, questions are asked by robots and smartphones that interact with humans. In this case, the respondent is a user such as a robot or a smartphone.

Here, in collecting information for knowing the hobbies and tastes of the user, it is conceivable to ask the user a question from a robot, a smartphone, or the like, and collect the user's answer to the question. However, respondents may answer a question with confidence or ambiguous answers.

For this reason, if information indicating only the content of the answer given by the respondent to the question is collected, the information may deviate from the actual situation of the respondent. In such a case, if all the answers are treated in the same way as information for determining the user's hobbies and preferences, the user's hobbies and preferences may be misjudged and hinder natural communication.

For example, if a user who answers vaguely "Well, do you like it ..." to the question "Do you like baseball?" Is treated in the same way as a user who answers "I like it!" With confidence. , Baseball, which I don't like so much, may be learned as something I like. As a result, for example, in communication with a robot, the topic of baseball, which is not so much liked, may continue, causing discomfort and annoyance to the user.

Therefore, in the present embodiment, the ambiguity of the content information indicating the content of the answer given by the respondent to a certain question is specified, and the information in which the content information is associated with the information according to the specified ambiguity. The generation method of suppressing the deviation of the generated information from the actual situation of the respondents will be described. Hereinafter, a processing example of the information processing apparatus 101 will be described.

(1) The information processing apparatus 101 receives content information indicating the content of the answer given by the respondent to a certain question and state information indicating the state of the respondent at the time of the answer. Here, the content of the answer is the answer of the respondent to the question. The state of the respondent is the emotion, behavior, appearance status of the filler, etc. of the respondent.

Respondents' emotions are, for example, joy, sadness, anger, calm, and so on. The behavior of the respondent is the line of sight, movement, etc. of the respondent. In addition, filler is a word that is inserted between utterances such as "um", "that", and "well". When answering a question, the more the respondents are worried about the answer, the more likely it is that fillers will occur.

In the example of FIG. 1, it is assumed that the content information 120 and the state information 130 are received. The content information 120 indicates the content of the answer given by the respondent 102 to the question X. The state information 130 indicates the ratio of the filler of the respondent 102 at the time of answering the question X as the state of the respondent 102 at the time of answering the question X.

(2) The information processing apparatus 101 specifies the degree of ambiguity of the received content information based on the received state information. Here, the ambiguity is a value indicating the ambiguity of the answer to the question, and is an index for discriminating the ambiguity of the respondent. In the following description, the larger the value of the ambiguity, the higher the degree of ambiguity, and the smaller the value of the ambiguity, the lower the degree of ambiguity.

The state information is information indicating the state of the respondent at the time of reply. That is, the information processing apparatus 101 specifies the ambiguity of the answer based on the state of the respondent at the time of answering the question. For example, when respondents give a positive answer to the question "Do you like ...?", The more joyful they are, the more confident they are, and the degree of ambiguity in the answer. Tends to be low. That is, the emotions of the respondents can be an index for judging the ambiguity of the answers.

In addition, when answering a question asked by a robot, it often takes a long time for the respondent's line of sight to turn to the robot. The more annoyed the answer, the more ambiguous the answer tends to be. That is, the behavior of the respondent can be an index for judging the ambiguity of the answer.

In addition, the more frequently the respondents' fillers such as "um" and "that" appear when answering a question, the more often they are worried about the answer. That is, the appearance status of the filler of the respondent can be an index for judging the ambiguity of the answer.

Specifically, for example, the information processing apparatus 101 may refer to the storage unit 110 to specify the degree of ambiguity associated with the received state information and the type of question. The storage unit 110 stores the ambiguity of the content of the answer given by the respondent to the question in association with the state information indicating the state of the respondent at the time of the answer and the type of the question.

That is, the information processing apparatus 101 stores information that can identify the degree of ambiguity of the answer according to the state of the respondent at the time of answering the question in association with the type of the question. Hold it in 110 in advance. Here, the type of question distinguishes the question.

For example, the type of question may be one that distinguishes each question. Further, the type of question may be one that distinguishes between an open question and a closed question. An open question is a question that has no fixed answer, such as "What kind of tea do you like?" A closed question is a question that has a fixed answer, such as "Are you a man?" In addition, the type of question may distinguish the subject (field) such as sports, music, and movies.

In the example of FIG. 1, the storage unit 110 stores the ambiguity of the content of the answer given to the question X in association with the state information indicating the state of the respondent at the time of answer and the question X. It is assumed that it has been done. However, the status of the respondents shall be the ratio of the filler. In FIG. 1, the ambiguity distribution 140 represents the distribution of ambiguity according to the state of the respondent at the time of answering the question X. In this case, the information processing apparatus 101 specifies the ambiguity b corresponding to the ratio a of the filler specified from the state information 130 as the ambiguity of the content information 120. This ambiguity b indicates the degree of ambiguity of the answer of the respondent 102 to the question X.

(3) The information processing apparatus 101 generates information in which the received content information is associated with information according to the specified ambiguity. Here, the information according to the degree of ambiguity may be, for example, information indicating whether or not the specified degree of ambiguity exceeds the standard. The standard of ambiguity can be set arbitrarily. For example, if the degree of ambiguity exceeds the standard, the degree of ambiguity of the answer is low and it is set to a value that can be judged to be a valid answer.

In the example of FIG. 1, the response information 160 in which the content information 120 is associated with the ambiguity information 150 is generated. The ambiguity information 150 is information according to the specified ambiguity degree b, and is, for example, information indicating whether or not the ambiguity degree b exceeds the standard. The response information 160 is, for example, learning data for learning the hobbies and tastes of the respondent 102.

As described above, according to the information processing apparatus 101, it is possible to generate information capable of determining the degree of ambiguity of the answer given by the respondent based on the state of the respondent at the time of answering a certain question. Therefore, it is possible to suppress the deviation of the generated information from the actual situation of the respondents.

For example, according to the answer information 160, when learning the hobbies and tastes of the respondent 102, it is possible to determine the degree of ambiguity of the answer given by the respondent 102 to the question X. Thereby, for example, if the degree of ambiguity of the answer is low, it can be treated as an effective answer in determining the hobbies and tastes of the respondent 102. On the other hand, if the degree of ambiguity of the answer is high, it can be treated that the answer is not effective in determining the hobby / preference of the respondent 102, and it is possible to prevent the respondent 102 from erroneously determining the hobby / preference. ..

(System configuration example of data collection system 200)
Next, a system configuration example of the data collection system 200 including the information processing apparatus 101 shown in FIG. 1 will be described. The data collection system 200 is applied to, for example, a system for realizing natural interpersonal communication by a robot, a smartphone, or the like.

FIG. 2 is an explanatory diagram showing a system configuration example of the data collection system 200. In FIG. 2, the data collection system 200 includes an information processing device 101 and a robot 201. In the data collection system 200, the information processing apparatus 101 and the robot 201 are connected via a wired or wireless network 210. The network 210 is, for example, a LAN (Local Area Network), a WAN (Wide Area Network), the Internet, or the like.

Here, the information processing apparatus 101 has a question-specific ambiguity distribution DB (Data) 220, a user-specific ambiguity distribution DB 230, and learning data DB 240. The information processing device 101 is, for example, a server. The storage contents of the question-specific ambiguity distribution DB 220, the user-specific ambiguity distribution DB 230, and the learning data DB 240 will be described later with reference to FIGS. 5 to 7.

The robot 201 is a computer that performs interpersonal communication. For example, the robot 201 is installed in a user's home, a hospital room, or the like, and is used for collecting information for knowing the user's hobbies and tastes. The robot 201 may be used by a specific user or may be used by a plurality of users. When a plurality of users use the robot 201, the user who is using the robot 201 is specified by face recognition, a login ID, or the like.

In the example of FIG. 2, only one robot 201 is shown, but the present invention is not limited to this. For example, the data collection system 200 may include a plurality of robots 201 used by different users. Further, the robot 201 may be, for example, a smartphone, a tablet-type PC (Personal Computer), or the like.

(Hardware configuration example of information processing device 101)
FIG. 3 is a block diagram showing a hardware configuration example of the information processing apparatus 101. In FIG. 3, the information processing apparatus 101 includes a CPU (Central Processing Unit) 301, a memory 302, an I / F (Interface) 303, a disk drive 304, and a disk 305. Further, each component is connected by a bus 300.

Here, the CPU 301 controls the entire information processing apparatus 101. The CPU 301 may have a plurality of cores. The memory 302 includes, for example, a ROM (Read Only Memory), a RAM (Random Access Memory), a flash ROM, and the like. Specifically, for example, the flash ROM stores the OS (Operating System) program, the ROM stores the application program, and the RAM is used as the work area of the CPU 301. The program stored in the memory 302 is loaded into the CPU 301 to cause the CPU 301 to execute the coded process.

The I / F 303 is connected to the network 210 through a communication line, and is connected to an external computer (for example, the robot 201 shown in FIG. 2) via the network 210. The I / F 303 controls the interface between the network 210 and the inside of the device, and controls the input / output of data from an external computer. For the I / F 303, for example, a modem, a LAN adapter, or the like can be adopted.

The disk drive 304 controls data read / write to the disk 305 according to the control of the CPU 301. The disk 305 stores the data written under the control of the disk drive 304. Examples of the disk 305 include a magnetic disk and an optical disk.

In addition to the above-mentioned components, the information processing device 101 may include, for example, an SSD (Solid State Drive), an input device, a display, and the like.

(Hardware configuration example of robot 201)
FIG. 4 is a block diagram showing a hardware configuration example of the robot 201. In FIG. 4, the robot 201 has a CPU 401, a memory 402, a display 403, an input device 404, an I / F 405, an audio signal processing unit 406, a speaker 407, a microphone 408, and a camera 409. .. Further, each component is connected by a bus 400.

Here, the CPU 401 controls the entire robot 201. The memory 402 includes, for example, a ROM, a RAM, a flash ROM, and the like. Specifically, for example, a flash ROM or ROM stores various programs, and RAM is used as a work area of the CPU 401. The program stored in the memory 402 is loaded into the CPU 401 to cause the CPU 401 to execute the coded process.

The display 403 displays data such as a cursor, an icon, a toolbox, a document, an image, and functional information. For example, the display 403 displays eyes, mouths, and the like that represent the facial expressions of the robot 201. As the display 403, for example, a liquid crystal display, an organic EL (Electroluminescence) display, or the like can be adopted.

The input device 404 has keys for inputting characters, numbers, various instructions, and the like, and inputs data. The input device 404 may be, for example, a touch panel type input pad, a numeric keypad, or the like, or may be a keyboard, a mouse, or the like.

The I / F 405 is connected to the network 210 through a communication line, and is connected to an external computer (for example, the information processing apparatus 101 shown in FIG. 2) via the network 210. The I / F 405 controls the interface between the network 210 and the inside of the device, and controls the input / output of data from an external computer.

The audio signal processing unit 406 is connected to the speaker 407 and the microphone 408. For example, the voice received by the microphone 408 is A / D converted by the voice signal processing unit 406. In addition, sound is output from the speaker 407.

The camera 409 is an image pickup device that captures an image (still image or moving image) and outputs image information. The camera 409 is provided at a position where a person looking at the display 403 can be imaged. For example, the camera 409 periodically captures images at predetermined time intervals.

In addition to the above-mentioned components, the robot 201 has, for example, a PMU (Power Management Unit) that supplies a power supply voltage to each component, a battery, various timers, various sensors, an infrared LED (Light Emitting Diode), and the like. You may do it.

(Memory contents of various DBs 220, 230, 240)
Next, the stored contents of various DBs 220, 230, 240 included in the information processing apparatus 101 will be described with reference to FIGS. 5 to 7. The various DBs 220, 230, and 240 are realized by, for example, a storage device such as the memory 302 and the disk 305 shown in FIG.

FIG. 5 is an explanatory diagram showing an example of the stored contents of the question-specific ambiguity distribution DB 220. In FIG. 5, the question-specific ambiguity distribution DB 220 has fields for question ID, TAG, Value, Question1, Question2, ambiguity, filler, emotional value, and behavior. By setting information in each field, question-specific ambiguity information (for example, question-specific ambiguity information 500-1,500-2) is stored as a record.

Here, the question ID is an identifier that uniquely identifies the question of Question1. The TAG indicates the subject for classifying the question. Value indicates a division for subdividing a question of a certain subject. Question 1 indicates a question. Question 2 indicates a question related to the question of Question 1.

In Question 2, for example, a question whose content is similar to that of Question 1 and whose listening method is different from that of Question 1 is set. However, if there is no question related to the question of Question 1, "-(Null)" is set in the field of Question 2.

The degree of ambiguity indicates the degree of ambiguity in the answer (unit:%). Filler indicates the percentage of filler corresponding to the degree of ambiguity (unit:%). The emotion value is a value representing the emotion of the respondent. Here, as the emotion value, an emotion value (joy) expressing the joy of the respondent is set. The higher the emotional value (joy), the stronger the joy of the respondents.

However, as the emotion value, an emotion value (sadness) representing the respondent's sadness, an emotion value (anger) representing the respondent's anger, and an emotion value (calmness) representing the respondent's calm may be set. .. Behavior is a value that represents the behavior of the respondent. Here, as a behavior, the time until the respondent's line of sight matches the robot 201 is set (unit: seconds).

For example, the question-specific ambiguity information 500-1 is the ambiguity of the answer when the question ID "Q1", TAG "sports", and Question 1 question "Do you like playing baseball?" The filler "0%" corresponding to the degree "10%", the emotion value "20" and the behavior "10 seconds" are shown. In addition, the question-specific ambiguity information 500-1 is a question similar in content to the Question 1 question, and the Question 2 question "Do you like playing baseball, if you like it or not?" ? ”Is shown.

If there is no value corresponding to the degree of ambiguity for each of the filler, emotion value, and behavior fields, "-" is set. For example, if there is no emotion value corresponding to the ambiguity "100%", "-" is set in the emotion value field. Further, the ambiguity information for each question may include, for example, information indicating the date and time when the ambiguity information for each question was registered. The storage unit 110 shown in FIG. 1 corresponds to, for example, the question-specific ambiguity distribution DB 220.

The question-specific ambiguity information is generated, for example, based on the result of asking a question (Question 1 question) to a plurality of subjects. More specifically, for example, the administrator of the data collection system 200 asks a question to a plurality of subjects and measures the filler, emotional value, and behavior of each subject when answering the question. In addition, each subject is asked to self-report the ambiguity of the answer to the question. Then, the manager generates question-specific ambiguity information based on the measurement results of the filler, emotional value, and behavior of each subject and the ambiguity self-reported by each subject.

FIG. 6 is an explanatory diagram showing an example of the stored contents of the user-specific ambiguity distribution DB 230. In FIG. 6, the user-specific ambiguity distribution DB 230 has fields for user ID, ambiguity, filler, emotion value, and behavior, and by setting information in each field, user-specific ambiguity information (for example, user-specific). The ambiguity information 600-1 to 600-11) is stored as a record.

Here, the user ID is an identifier that uniquely identifies the user of the robot 201. The degree of ambiguity indicates the degree of ambiguity in the answer (unit:%). Here, for the same user ID, user-specific ambiguity information is stored for every 10% of ambiguity from 0% to 100%. Filler indicates the percentage of filler corresponding to the degree of ambiguity (unit:%).

The emotion value is a value representing the emotion of the respondent. Here, as the emotion value, an emotion value (joy) representing the joy of the user of the robot 201 is set. However, as the emotion value, an emotion value (sadness) representing the sadness of the user, an emotion value (anger) representing the anger of the user, or an emotion value (calmness) representing the calmness of the user may be set. Behavior is a value that represents the behavior of the respondent. Here, as a behavior, a time until the user's line of sight matches the robot 201 is set (unit: seconds).

For example, the user-specific ambiguity information 600-1 indicates a filler "0%", an emotion value "10", and a behavior "0 seconds" corresponding to the ambiguity "0%" of the user U1.

The user-specific ambiguity information is information representing the characteristics of the user of the robot 201, and is created, for example, when the user starts using the robot 201. The specific processing content when creating the ambiguity information for each user will be described later with reference to FIGS. 16 and 17.

FIG. 7 is an explanatory diagram showing an example of the stored contents of the learning data DB 240. In FIG. 7, the learning data DB 240 has fields of user ID, question, answer, ambiguity, and ambiguity flag, and by setting information in each field, learning data (for example, learning data 700-1) is recorded. Remember as.

Here, the user ID is an identifier that uniquely identifies the user of the robot 201. The question is a question asked to the user. The answer is the content of the answer to the question. The degree of ambiguity indicates the degree of ambiguity in the answer. The ambiguity flag is information indicating whether or not the ambiguity exceeds the standard. The ambiguity flag "0" indicates that the degree of ambiguity does not exceed the standard. The ambiguity flag "1" indicates that the degree of ambiguity exceeds the standard.

For example, the learning data 700-1 includes the question "Do you like playing baseball?" Asked to the user U1, the answer "I like it", the ambiguity level "10%" of the answer, and the ambiguity flag. Indicates "0". The learning data may include a question ID.

(Example of functional configuration of information processing device 101)
FIG. 8 is a block diagram showing a functional configuration example of the information processing apparatus 101. In FIG. 8, the information processing apparatus 101 includes a reception unit 801, a specific unit 802, a generation unit 803, an output unit 804, and a storage unit 810. The reception unit 801 to the output unit 804 are functions that serve as control units. Specifically, for example, by causing the CPU 301 to execute a program stored in a storage device such as the memory 302 and the disk 305 shown in FIG. Alternatively, the function is realized by the I / F 303. The processing result of each functional unit is stored in a storage device such as a memory 302 or a disk 305. Further, the storage unit 810 is realized by, for example, a storage device such as a memory 302 and a disk 305. Specifically, for example, the storage unit 810 stores the question-specific ambiguity distribution DB 220, the user-specific ambiguity distribution DB 230, the learning data DB 240, and the like.

The reception unit 801 receives content information indicating the content of the answer given by the respondent RT to a certain question Q and status information indicating the state of the respondent RT at the time of the answer. Here, the content of the answer is the answer of the respondent RT to the question Q. The state of the respondent RT is the emotion, behavior, appearance status of the filler, and the like of the respondent RT. Respondent RT is a user of robot 201.

Specifically, for example, the reception unit 801 includes an acquisition unit 805 and an analysis unit 806. The acquisition unit 805 acquires voice data at the time of answering the question Q from the robot 201. The voice data at the time of answering the question Q is, for example, the voice data of the voice received by the microphone 408 of the robot 201 shown in FIG. 4 in the response section to the question Q. Further, the acquisition unit 805 acquires the image data at the time of answering the question Q from the robot 201. The image data at the time of answering the question Q is, for example, the image data of the image captured by the camera 409 of the robot 201 shown in FIG. 4 in the response section to the question Q.

The response section is, for example, a section from the end of the utterance of the question Q by the robot 201 until a predetermined time (for example, 10 seconds) elapses. However, the end time of the response section may be the timing when it is determined that the answer to the question Q by the user is completed, that is, the utterance of the user is completed. Further, the response section may be the utterance section of the user. The utterance section is a section spoken by the user. The utterance section can be detected by analyzing the voice data at the time of answering the question Q.

The analysis unit 806 detects the content of the answer given by the respondent RT to the question Q by analyzing the voice data at the time of answering the acquired question Q. Specifically, for example, the analysis unit 806 decides to detect the result of converting the spoken language of the user obtained by voice recognition into a character string as the content of the answer given by the respondent RT to the question Q. May be good. In this case, the reception unit 801 accepts the information indicating the content of the detected answer as the content information indicating the content of the answer given by the respondent RT to the question Q.

Further, the analysis unit 806 detects the section in which the filler of the respondent RT appears by analyzing the voice data at the time of answering the acquired question Q. Then, the analysis unit 806 calculates the appearance state of the filler at the time of answering the question Q based on the detected result. Specifically, for example, the analysis unit 806 calculates the ratio of the filler at the time of answering the question Q from the response section of the question Q and the section in which the detected filler appears. In this case, the reception unit 801 receives, for example, information indicating the calculated percentage of the filler as state information indicating the state of the respondent RT at the time of answering the question Q.

Here, an example of calculating the ratio of the filler will be described with reference to FIG.

FIG. 9 is an explanatory diagram showing an example of detection of a section in which a filler appears. In FIG. 9, the t-axis shows the passage of time. The time point tS indicates the start time point of the response interval to the question Q. The time point tE indicates the end time point of the response interval to the question Q. Here, it is assumed that the sections s1 and s2 in which the filler appears are detected in the response section to the question Q.

Further, the section length of the response section is set to "10 seconds", and the section length of each section s1 and s2 is set to "1 second". In this case, the analysis unit 806 calculates the filler ratio "20%" by dividing the total section length "2 seconds" of the sections s1 and s2 by the section length "10 seconds" of the response section. Then, the reception unit 801 receives the information indicating the calculated filler ratio "20%" as the state information indicating the state of the respondent RT at the time of answering the question Q.

Further, the analysis unit 806 calculates an emotion value representing the emotion of the respondent RT by analyzing the voice data at the time of answering the acquired question Q. The emotional value includes, for example, the joy, sadness, anger, and calmness of the respondent RT. Specifically, for example, the analysis unit 806 will calculate the emotional values representing the joy, sadness, anger, and calmness of the respondent RT from the frequency of the voice, the speech speed (pitch), the loudness of the voice, and the like. You may.

In this case, the reception unit 801 receives, for example, the calculated emotion value indicating the emotion of the respondent RT as the state information indicating the state of the respondent RT at the time of answering the question Q. Any existing technique may be used as a method for obtaining the emotional value of a person from voice data.

Further, the analysis unit 806 may calculate an emotion value representing the emotion of the respondent RT by analyzing the image data at the time of answering the acquired question Q. Specifically, for example, the analysis unit 806 detects a face image from the image data at the time of answering the acquired question Q, and determines the facial expression of the detected face image. Then, the analysis unit 806 calculates an emotion value representing the emotion of the respondent RT based on the determination result.

Here, an example of calculating an emotion value representing the emotion of the respondent RT who is the user of the robot 201 will be described with reference to FIG. Here, as an emotional value, a case of calculating an emotional value expressing the joy of the respondent RT will be described as an example.

FIG. 10 is an explanatory diagram showing an example of calculating an emotion value representing the emotion of the respondent. In FIG. 10, the t-axis shows the passage of time. The time point tS indicates the start time point of the response interval to the question Q. The time point tE indicates the end time point of the response interval to the question Q. Here, it is assumed that the face images p1 to p6 are detected in the response section to the question Q.

The analysis unit 806 determines the facial expressions of the detected facial images p1 to p6. Here, it is assumed that the face images p1 to p3 are determined to be calm (expressionless), and the face images p4 to p6 are determined to be joy. In this case, the analysis unit 806 divides, for example, the number "3" of the face images p4 to p6 determined to be joy by the number "6" of the face images p1 to p6 detected in the response section, so that the respondent RT The emotional value "50 (= 3/6 x 100)" that expresses the joy of Then, the reception unit 801 receives the calculated information indicating the emotion value "50" representing the joy of the respondent RT as the state information indicating the state of the respondent RT at the time of answering the question Q.

Further, the analysis unit 806 calculates a value representing the behavior of the respondent RT by analyzing the image data at the time of answering the acquired question Q. The values representing the behavior of the respondent RT include, for example, the time until the line of sight of the respondent RT meets the robot 201, and the time until the respondent RT becomes stationary (for example, there is no movement of the neck or arm). Time etc. can be mentioned. The line of sight of the respondent RT is aligned with the robot 201, for example, the line of sight of the respondent RT is directed toward the eyes of the robot 201 displayed on the display 403 of the robot 201.

In this case, the reception unit 801 receives, for example, information indicating the calculated behavior of the respondent RT as state information indicating the state of the respondent RT at the time of answering the question Q. Any existing technique may be used as a method for detecting the line of sight or movement of a person from an image obtained by capturing the person.

Here, an example of calculating a value representing the behavior of the respondent RT who is the user of the robot 201 will be described with reference to FIG. Here, as a value representing the behavior, a case where the time until the line of sight of the respondent RT matches the robot 201 is calculated will be described as an example.

FIG. 11 is an explanatory diagram showing an example of calculating a value representing the behavior of the respondent. In FIG. 11, the t-axis shows the passage of time. The time point tS indicates the start time point of the response interval to the question Q. The time point tE indicates the end time point of the response interval to the question Q. Here, it is assumed that the line of sight of the respondent as shown in the facial images p11 to p16 is detected in the response section to the question Q.

Further, among the face images p11 to p16, the face images p15 and p16 are the face images when the line of sight of the respondent RT matches the robot 201. In this case, the analysis unit 806 sets the time from the time point tS to the time point t1 when the line of sight of the respondent RT shown in the face image p15 is detected, and the time until the line of sight of the respondent RT matches the robot 201 "t1-". Calculated as "tS". Then, the reception unit 801 receives the information indicating the calculated time until the line of sight is met "t1-tS" as the state information indicating the state of the respondent RT at the time of answering the question Q.

The content information indicating the content of the answer given by the respondent RT to the question Q and the state information indicating the state of the respondent RT at the time of the answer may be generated by the robot 201. In this case, the reception unit 801 receives from the robot 201 the content information indicating the content of the answer given by the respondent RT to the question Q and the state information indicating the state of the respondent RT at the time of the answer. Further, the reception unit 801 may not include the acquisition unit 805 and the analysis unit 806.

Further, the answer to the question Q may be input by the operation of the user (respondent RT) using the input device 404 of the robot 201 shown in FIG. 4, for example. In this case, the reception unit 801 may receive the result input to the robot 201 from the robot 201 as content information indicating the content of the answer given by the respondent RT to the question Q.

The specifying unit 802 specifies the ambiguity level v of the received content information based on the received state information. Specifically, for example, the specific unit 802 refers to the storage unit 810 and specifies the degree of ambiguity v associated with the received state information and the type of the question Q. The storage unit 810 stores the ambiguity of the content of the answer given by the respondent to the question in association with the state information indicating the state of the respondent at the time of the answer and the type of the question. ing. The type of question is, for example, to distinguish each question.

More specifically, for example, the specific unit 802 refers to the question-specific ambiguity distribution DB 220 shown in FIG. 5 and acquires the question-specific ambiguity information corresponding to the question ID of the question Q. According to the acquired ambiguity information for each question, it is possible to specify the distribution of the ambiguity according to each state of the respondent RT at the time of answering the question Q.

For example, in a coordinate system in which the vertical axis is the ambiguity and the horizontal axis is the filler ratio, the question Q is plotted by plotting the correspondence between the ambiguity indicated by the acquired ambiguity information for each question and the filler ratio. It is possible to specify the distribution of ambiguity according to the appearance status of the filler at the time of answering. Further, by plotting the correspondence between the ambiguity and the emotion value indicated by the ambiguity information for each question, it is possible to specify the distribution of the ambiguity according to the emotion of the respondent RT at the time of answering the question Q. In addition, by plotting the correspondence between the ambiguity indicated by the ambiguity information for each question and the value representing the behavior, it is possible to specify the distribution of the ambiguity according to the behavior of the respondent RT when answering the question Q. can.

Then, the specific unit 802 determines the ambiguity corresponding to the state of the respondent RT indicated by the received state information based on the distribution of the ambiguity specified from the acquired question-specific ambiguity information, and the ambiguity of the received content information. Specify as degree v. As the ambiguity v of the content information, for example, the ambiguity v according to each state of the respondent RT (appearance status of the filler, emotion of the respondent RT, behavior of the respondent RT) can be specified.

A specific example of the ambiguity level v of the content information will be described later with reference to FIGS. 12 to 14.

Further, in the storage unit 810, the degree of ambiguity of the content of the answer given by the user to another question Q'different from a certain question Q for the user who is the respondent RT is determined by the user at the time of the answer. User-specific ambiguity information that is associated with the state information indicating the state may be stored. The user-specific ambiguity information is, for example, the user-specific ambiguity information in the user-specific ambiguity distribution DB 230 shown in FIG.

In this case, the specifying unit 802 may refer to the storage unit 810 to specify the second ambiguity level v'corresponding to the received state information and the user who is the respondent RT. .. Then, the specific unit 802 may correct the specified ambiguity v based on the specified second ambiguity v'.

More specifically, for example, the specific unit 802 refers to the user-specific ambiguity distribution DB 230 and acquires user-specific ambiguity information corresponding to the user ID of the respondent RT (user of the robot 201). According to the acquired user-specific ambiguity information, it is possible to specify the distribution of ambiguity according to each state of the respondent RT at the time of answering another question Q'different from a certain question Q.

For example, in a coordinate system in which the vertical axis is the ambiguity and the horizontal axis is the filler ratio, the respondents are plotted by plotting the correspondence between the ambiguity and the filler ratio indicated by the acquired ambiguity information for each user. It is possible to specify the distribution of ambiguity according to the appearance status of the RT filler. Further, by plotting the correspondence between the ambiguity and the emotion value indicated by the ambiguity information for each user, it is possible to specify the distribution of the ambiguity according to the emotion of the respondent RT. Further, by plotting the correspondence between the ambiguity indicated by the ambiguity information for each user and the value representing the behavior, it is possible to specify the distribution of the ambiguity according to the behavior of the respondent RT.

Then, the specifying unit 802 sets the ambiguity corresponding to the state of the respondent RT indicated by the received state information to the second ambiguity v, based on the distribution of the ambiguity specified from the acquired user-specific ambiguity information. 'Specify as. As the second ambiguity v', for example, the ambiguity according to each state of the respondent RT (appearance status of the filler, emotion of the respondent RT, behavior of the respondent RT) can be specified.

As a method for correcting the ambiguity v, for example, the specific unit 802 may calculate the average value of the ambiguity v and the second ambiguity v'as the corrected ambiguity v. Further, the specific unit 802 may set the maximum ambiguity (or the minimum ambiguity) of the ambiguity v and the second ambiguity v'as the corrected ambiguity v.

An example of correction of the degree of ambiguity v will be described later with reference to FIG.

The generation unit 803 generates learning data in which the received content information is associated with the specified ambiguity v or the corrected ambiguity v. Here, the information according to the degree of ambiguity v may be, for example, information indicating the degree of ambiguity v, or information indicating whether or not the degree of ambiguity v exceeds the standard.

Further, the generation unit 803 generates learning data in which the received content information and the content of the question Q are associated with the information corresponding to the specified ambiguity v (or the corrected ambiguity v). You may do it. Here, the content of the question Q is, for example, a character string representing the contents of the question Q, an image, or the like.

Further, the generation unit 803 adds information according to the specified ambiguity v (or corrected ambiguity v) to the received content information, the content of the question Q, and the information specifying the respondent RT. The associated learning data may be generated. Here, the information that identifies the respondent RT is, for example, the user ID of the user who is the respondent RT.

Further, the information that identifies the respondent RT may be information indicating the name, gender, age, etc. of the user who is the respondent RT. The name, gender, and age of the user who is the respondent RT are stored in a storage device such as a memory 302 or a disk 305 in association with the user ID of the user, for example.

Specifically, for example, the generation unit 803 includes the received content information, the content of the question Q, the information specifying the respondent RT, and the specified ambiguity v (or the corrected ambiguity v). , The learning data including the information indicating whether or not the ambiguity v exceeds the standard may be generated. The generated learning data is stored in, for example, the learning data DB 240 shown in FIG. 7.

In the example of FIG. 7, the case where the learning data includes only one of the ambiguities corresponding to each state of the respondent RT at the time of answering the question Q is taken as an example. Not limited to this. For example, the learning data may include all ambiguities according to each state of the respondent RT (filler appearance status, emotion of the respondent RT, behavior of the respondent RT) at the time of answering the question Q. .. Further, as the ambiguity according to the emotion of the respondent RT, for example, the ambiguity corresponding to each of the emotion values representing the joy, sadness, anger, and calmness of the respondent RT may be included. Further, the learning data may include only the maximum ambiguity (or the minimum ambiguity) among the ambiguities corresponding to each state of the respondent RT at the time of answering the question Q.

The output unit 804 outputs the generated information. The output format of the output unit 804 includes, for example, storage in a storage device such as a memory 302 and a disk 305, and transmission to another computer by the I / F 303. Specifically, for example, the output unit 804 may output the learning data (for example, the learning data 700-1) stored in the learning data DB 240. The learning data is used, for example, as information for learning the hobbies and tastes of the user of the robot 201.

The information processing apparatus 101 may control which question Q is given to the user from the robot 201, or the robot 201 may autonomously select the question Q. Further, in the robot 201, the question Q may be output from the speaker 407 shown in FIG. 4 or may be output using the display 403 shown in FIG.

In addition, the emotions of the respondent RT may appear differently when giving a positive (positive) answer to the question Q and when giving a negative (negative) answer. For example, when giving a positive answer, feelings of joy are more likely to occur than when giving a negative answer. On the other hand, when giving a negative answer, feelings of sadness are more likely to occur than when giving a positive answer.

Therefore, the specifying unit 802 may narrow down the question-specific ambiguity information used to specify the ambiguity v for the answer according to the content of the answer given by the respondent RT to the question Q. For example, the question-specific ambiguity information in the question-specific ambiguity distribution DB 220 may include a phrase W for determining whether the information is when a positive or negative answer is obtained.

Examples of the phrase W for determining that the information is obtained when a positive answer is obtained include "like" and "yes" included in the positive answer. Examples of the phrase W for determining that the information is obtained when a negative answer is obtained include "dislike" and "no" included in the negative answer.

That is, the information processing apparatus 101 distinguishes between the question-specific ambiguity information when a positive answer is obtained and the question-specific ambiguity information when a negative answer is obtained. In this case, the specific unit 802 refers to, for example, the question-specific ambiguity distribution DB 220, and is the question-specific ambiguity information corresponding to the question ID of the question Q, and is the answer given by the respondent RT to the question Q. The ambiguity information for each question including the phrase W is acquired. Then, the specifying unit 802 specifies the ambiguity level v of the received content information based on the acquired ambiguity level information for each question.

For example, the content of the answer given by the respondent RT to the question Q is "I really like it." In this case, among the question-specific ambiguity information corresponding to the question ID of the question Q, the question-specific ambiguity information including the phrase W "like" is acquired. Thereby, when the respondent RT gives a positive answer to the question Q, the ambiguity v can be specified by using the ambiguity information for each question when the positive answer is obtained.

Further, for example, the content of the answer given by the respondent RT to the question Q is "I hate it." In this case, among the question-specific ambiguity information corresponding to the question ID of the question Q, the question-specific ambiguity information including the phrase W "dislike" is acquired. Thereby, when the respondent RT gives a negative answer to the question Q, the ambiguity v can be specified by using the ambiguity information for each question when the negative answer is obtained.

Further, the information processing apparatus 101 may control the question to be asked after the question Q based on the ambiguity v specified by the specific unit 802. For example, the respondent RT may have an ambiguous answer because he / she could not hear the question Q well or could not grasp the intention of the question Q well.

Therefore, when the ambiguity level v exceeds the standard, the information processing apparatus 101 controls the robot 201 to ask a question whose content is similar to that of the question Q, and which is different from the question Q (question Q). For example, Question Q) of Question 2 may be output. As a result, when the respondent RT cannot hear the question Q well or cannot grasp the intention of the question Q well, he / she can naturally re-listen and hear the answer.

(Specific example of ambiguity v)
Next, a specific example of the ambiguity level v of the content information indicating the content of the answer given by the respondent RT to a certain question Q will be described with reference to FIGS. 12 to 14. Here, question Q is referred to as "question Q10". First, with reference to FIG. 12, a specific example of the degree of ambiguity v according to the appearance state of the filler at the time of answering the question Q10 will be described.

FIG. 12 is an explanatory diagram (No. 1) showing a specific example of the ambiguity degree v. In FIG. 12, the ambiguity distribution 1200 represents the ambiguity distribution according to the proportion of the filler at the time of answering the question Q10. The ambiguity distribution 1200 is specified, for example, from the filler of the ambiguity information for each question corresponding to the question ID "Q10" in the ambiguity distribution DB 220 for each question.

Here, it is assumed that the state information indicating the filler ratio “30%” is received as the state information indicating the state of the respondent RT at the time of answering the question Q10. In this case, the specifying unit 802 specifies the ambiguity "33%" corresponding to the filler ratio "30%" indicated by the state information as the ambiguity v (filler) of the content information.

Next, with reference to FIG. 13, a specific example of the degree of ambiguity v according to the emotional value representing the joy of the respondent RT at the time of answering the question Q10 will be described.

FIG. 13 is an explanatory diagram (No. 2) showing a specific example of the ambiguity degree v. In FIG. 13, the ambiguity distribution 1300 represents the distribution of ambiguity according to the emotional value that expresses the joy of the respondent when answering the question Q10. The ambiguity distribution 1300 is specified, for example, from the emotional value of the ambiguity information for each question corresponding to the question ID "Q10" in the ambiguity distribution DB 220 for each question.

Here, as the state information indicating the state of the respondent RT at the time of answering the question Q10, it is assumed that the state information indicating the emotion value “10” expressing the joy of the respondent RT is received. In this case, the specifying unit 802 specifies the ambiguity "70%" corresponding to the emotion value "10" indicated by the state information as the ambiguity v (emotion) of the content information.

Next, with reference to FIG. 14, a specific example of the degree of ambiguity v according to the value representing the behavior of the respondent RT at the time of answering the question Q10 will be described.

FIG. 14 is an explanatory diagram (No. 3) showing a specific example of the ambiguity degree v. In FIG. 14, the ambiguity distribution 1400 represents the distribution of ambiguity according to the time until the respondent's line of sight matches the robot 201 when answering the question Q10. The ambiguity distribution 1400 is specified, for example, from the behavior of the ambiguity information for each question corresponding to the question ID "Q10" in the ambiguity distribution DB 220 for each question.

Here, as the state information indicating the state of the respondent RT at the time of answering the question Q10, it is assumed that the state information indicating the time "15 seconds" until the line of sight of the respondent RT meets the robot 201 is received. In this case, the specifying unit 802 specifies the ambiguity "70%" corresponding to the time "15" indicated by the state information as the ambiguity v (time until the line of sight is met) of the content information.

(Example of correction of ambiguity v)
Next, with reference to FIG. 15, an example of correcting the ambiguity degree v of the content information indicating the content of the answer given by the respondent RT to a certain question Q will be described. Here, the question Q is referred to as "question Q10", and an example of correcting the ambiguity level v according to the appearance state of the filler at the time of answering the question Q10 will be described. Further, the respondent RT is set to the user U1 of the user ID "U1".

FIG. 15 is an explanatory diagram showing an example of correction of the ambiguity degree v. In FIG. 15, the ambiguity distribution 1500 represents the ambiguity distribution according to the ratio of the filler at the time of answering by the user U1 who is the respondent RT. The ambiguity distribution 1500 is specified, for example, from the filler of the ambiguity information for each user corresponding to the user ID "U1" in the ambiguity distribution DB 230 for each user.

Here, it is assumed that the state information indicating the filler ratio “30%” is received as the state information indicating the state of the respondent RT at the time of answering the question Q10. In this case, the specifying unit 802 specifies the ambiguity "10%" corresponding to the filler ratio "30%" indicated by the state information as the second ambiguity v'(filler).

Further, the ambiguity v corresponding to the filler ratio “30%” indicated by the state information is set to “33%” (see FIG. 12). In this case, the specific unit 802 calculates, for example, the average value “21.5%” of the ambiguity v “33%” and the second ambiguity v ′ “10” as the corrected ambiguity v. Thereby, the ambiguity level v can be corrected according to the characteristics of the user U1 who is the respondent RT.

In the example of FIG. 15, as a characteristic of the user U1, the ambiguity v is set from "33%" to "21.5" in consideration of the tendency that the ambiguity is low when the ratio of the filler is about 0 to 50%. It can be corrected to "%".

(Various processing procedures of the information processing device 101)
Next, various processing procedures of the information processing apparatus 101 will be described. First, a user-specific ambiguity information creation processing procedure for creating user-specific ambiguity information representing the characteristics of the user of the robot 201 will be described using the state determination question list 1600 as shown in FIG. The user-specific ambiguity information creation process is performed, for example, as one of the processes related to the initial setting when the user starts using the robot 201.

FIG. 16 is an explanatory diagram showing a specific example of the status determination question list 1600. In FIG. 16, the state determination question list 1600 includes state determination question information 1600-1 to 1600-11 representing the ambiguity (unit:%) and the question in association with each other. The question information 1600-1 to 1600-11 for each state determination is created by estimating in advance how much the degree of ambiguity will be when the user answers each question.

For example, since it can be estimated that all users can answer (for example, "yes") the question information for status determination without worrying about the question, the degree of ambiguity is "0%". On the other hand, since it can be estimated that all users have trouble answering the question in the state determination question information 1600-11, the degree of ambiguity is "100%".

In the example of FIG. 16, the case where one question is set for every 10% of the ambiguity from 0% to 100% is given as an example, but the present invention is not limited to this. For example, in the state determination question list 1600, a plurality of questions corresponding to a certain degree of ambiguity may be set.

FIG. 17 is a flowchart showing an example of the user-specific ambiguity information creation processing procedure of the information processing apparatus 101. In the flowchart of FIG. 17, first, the information processing apparatus 101 selects an unselected question that has not been selected from the status determination question list 1600 (step S1701).

Next, the information processing apparatus 101 controls the robot 201 to output the selected question (step S1702). Then, the information processing apparatus 101 starts acquiring the analysis target data from the robot 201 (step S1703). The analysis target data includes the voice data of the voice received by the microphone 408 of the robot 201 and the image data of the image captured by the camera 409 of the robot 201.

Next, the information processing apparatus 101 determines whether or not the user's answer to the output question has been completed (step S1704). For example, the information processing apparatus 101 may determine that the user's answer to the question has ended in response to the end of the response section to the output question.

Here, the information processing apparatus 101 waits for the answer to be completed (step S1704: No). Then, when the answer is completed (step S1704: Yes), the information processing apparatus 101 ends the acquisition of the analysis target data from the robot 201 (step S1705).

Next, the information processing apparatus 101 analyzes the voice data included in the acquired analysis target data to calculate the ratio of the filler at the time of answering the output question (step S1706). Next, the information processing apparatus 101 calculates an emotion value representing the user's emotion by analyzing the voice data (or image data) included in the acquired analysis target data (step S1707).

Next, the information processing apparatus 101 analyzes the image data included in the acquired analysis target data to calculate the time until the user's line of sight matches the robot 201 (step S1708). The order of processing in steps S1706 to S1708 may be changed.

Then, the information processing apparatus 101 creates user-specific ambiguity information representing the calculated ratio of the filler, the emotional value, and the time until the line of sight is matched, in association with the ambiguity of the selected question (step S1709). The user-specific ambiguity information includes the user ID of the user of the robot 201.

Next, the information processing apparatus 101 records the created user-specific ambiguity information in the user-specific ambiguity distribution DB 230 (step S1710). Then, the information processing apparatus 101 determines whether or not there is an unselected question that has not been selected from the status determination question list 1600 (step S1711).

Here, if there is an unselected question (step S1711: Yes), the information processing apparatus 101 returns to step S1701. On the other hand, when there is no unselected question (step S1711: No), the information processing apparatus 101 ends a series of processes according to this flowchart.

Thereby, the user-specific ambiguity information representing the user characteristics of the robot 201 can be created and registered in the user-specific ambiguity distribution DB 230.

In the state determination question list 1600, a plurality of questions corresponding to a certain degree of ambiguity may be set. In this case, in step S1709, as the information corresponding to the degree of ambiguity, the average value of each value (filler ratio, emotional value, time until the line of sight is met) when each of a plurality of questions is asked may be used. good.

Further, the status determination question list 1600 may be created for, for example, one for an open question and one for a closed question. In this case, it is possible to create user-specific ambiguity information that represents the characteristics of the user when asking an open question and user-specific ambiguity information that represents the characteristics of the user when asking a closed question.

Next, the generation processing procedure of the information processing apparatus 101 will be described with reference to FIGS. 18 and 19. The generation process is a process of generating learning data for the user of the robot 201. The learning data is used, for example, when learning a user's hobbies and tastes.

18 and 19 are flowcharts showing an example of the generation processing procedure of the information processing apparatus 101. In the flowchart of FIG. 18, first, the information processing apparatus 101 refers to the question-specific ambiguity distribution DB 220 and selects the question Q (step S1801). The selected question Q is any Question 1 question in the question-specific ambiguity distribution DB 220.

Next, the information processing apparatus 101 controls the robot 201 to output the selected question Q (step S1802). Then, the information processing apparatus 101 starts acquiring the analysis target data from the robot 201 (step S1803). Next, the information processing apparatus 101 determines whether or not the user's answer to the output question Q has been completed (step S1804).

Here, the information processing apparatus 101 waits for the answer to be completed (step S1804: No). Then, when the answer is completed (step S1804: Yes), the information processing apparatus 101 ends the acquisition of the analysis target data from the robot 201 (step S1805).

Next, the information processing apparatus 101 detects the content of the answer given by the respondent RT to the question Q by analyzing the voice data included in the acquired analysis target data (step S1806). As a result, the information indicating the content of the detected answer is accepted as the content information indicating the content of the answer given by the respondent RT to the question Q.

Next, the information processing apparatus 101 analyzes the voice data included in the acquired analysis target data to calculate the ratio of the filler at the time of answering the output question Q (step S1807). As a result, the information indicating the calculated ratio of the filler is accepted as the state information indicating the state of the respondent RT at the time of answering the question Q.

Next, the information processing apparatus 101 analyzes the voice data (or image data) included in the acquired analysis target data to calculate an emotion value representing the emotion of the respondent RT (step S1808). As a result, the calculated information indicating the emotion value is accepted as the state information indicating the state of the respondent RT at the time of answering the question Q.

Next, the information processing apparatus 101 analyzes the image data included in the acquired analysis target data to calculate the time until the line of sight of the respondent RT matches the robot 201 (step S1809). As a result, the calculated information indicating the time until the line of sight is met is accepted as the state information indicating the state of the respondent RT at the time of answering the question Q.

Then, the information processing apparatus 101 shifts to step S1901 shown in FIG. The order of processing in steps S1807 to S1809 may be changed.

In the flowchart of FIG. 19, first, the information processing apparatus 101 refers to the question-specific ambiguity distribution DB 220 and acquires the question-specific ambiguity information corresponding to the question ID of the selected question Q (step S1901). Then, the information processing apparatus 101 specifies the ambiguity corresponding to the state of the respondent RT indicated by the received state information as the ambiguity v of the received content information based on the acquired question-specific ambiguity information (step). S1902).

In step S1902, the information processing apparatus 101 may specify the ambiguity level v for all the state information among the state information received as a result of the processing of steps S1807 to S1809, or any of the state information is ambiguous. The degree v may be specified.

Next, the information processing apparatus 101 refers to the user-specific ambiguity distribution DB 230, and acquires the user-specific ambiguity information corresponding to the user ID of the respondent RT (step S1903). Then, the information processing apparatus 101 specifies the ambiguity corresponding to the state of the respondent RT indicated by the received state information as the second ambiguity v'based on the acquired user-specific ambiguity information (step S1904). ).

In step S1904, the information processing apparatus 101 may specify the second ambiguity level v'for all the state information among the state information received as a result of the processing of steps S1807 to S1809, or any one of them. A second degree of ambiguity v'may be specified for the state information.

Then, the information processing apparatus 101 corrects the specified ambiguity v based on the specified second ambiguity v'(step S1905). When the information processing apparatus 101 specifies the ambiguity v and the second ambiguity v'for all the state information received as a result of the processes of steps S1807 to S1809, the second ambiguity v corresponds to each of the second ambiguities v. The correction is made based on the ambiguity v'.

Next, the information processing apparatus 101 determines whether or not the corrected ambiguity v exceeds the threshold value (step S1906). The threshold value can be arbitrarily set, and is set to a value of, for example, about 30%. Here, when the corrected ambiguity v is less than the threshold value (step S1906: No), the information processing apparatus 101 sets the ambiguity flag to “0” (step S1907).

Next, the information processing apparatus 101 generates learning data including the received content information, the content of the selected question Q, the information specifying the respondent RT, the corrected ambiguity v, and the ambiguity flag. (Step S1908). Then, the information processing apparatus 101 records the generated learning data in the learning data DB 240 (step S1909), and ends a series of processes according to this flowchart.

Further, in step S1906, when the corrected ambiguity v exceeds the threshold value (step S1906: Yes), whether or not the information processing apparatus 101 has already selected the Question 2 question corresponding to the question Q selected in step S1801. (Step S1910). If there is no Question 2 question corresponding to Question Q, the information processing apparatus 101 ends a series of processes according to this flowchart.

Here, when the question of Question 2 is not selected (step S1910: No), the information processing apparatus 101 refers to the question-specific ambiguity distribution DB 220 and asks the question of Question 2 corresponding to the question Q selected in step S1801. Select as a new question Q (step S1911). Then, the information processing apparatus 101 returns to step S1802 shown in FIG. On the other hand, when the question of Question 2 has been selected (step S1910: Yes), the information processing apparatus 101 ends a series of processes according to this flowchart.

As a result, the degree of ambiguity of the answer to the question Q is low, and it is possible to collect learning data that can be said to be effective information for learning the hobbies and tastes of the respondent RT.

If the corrected ambiguity v exceeds the threshold value in step S1906 (step S1906: Yes), the information processing apparatus 101 may set the ambiguity flag to "1" and proceed to step S1908. In this case, even answers with a high degree of ambiguity can be collected as learning data. Further, the robot 201 may determine which question Q is selected in step S1801.

Further, in the user-specific ambiguity information creation process shown in FIG. 17, the user-specific ambiguity information showing the characteristics of the user when an open question is asked and the user representing the user characteristics when a closed question is asked. Another ambiguity information may be created. In this case, the information processing apparatus 101 may acquire user-specific ambiguity information in step S1903 depending on whether the question Q is an open question or a closed question. As a result, when the question Q is an open question, the ambiguity v can be corrected according to the characteristics of the user (respondent RT) when the open question is asked. On the other hand, when the question Q is a closed question, the ambiguity v can be corrected according to the characteristics of the user (respondent RT) when the closed question is asked.

As described above, according to the information processing apparatus 101 according to the embodiment, the content information indicating the content of the answer given by the respondent RT to a certain question Q and the state of the respondent RT at the time of the answer are displayed. The indicated status information and can be accepted. Then, according to the information processing apparatus 101, the ambiguity v of the received content information is specified based on the received state information, and the learned data in which the received content information is associated with the information corresponding to the specified ambiguity v. Can be generated.

As a result, it is possible to generate learning data capable of determining the degree of ambiguity of the answer given by the respondent RT based on the state of the respondent RT at the time of answering the question Q. Therefore, it is possible to prevent the generated learning data from being treated as information that deviates from the actual state of the respondent RT.

Further, according to the information processing apparatus 101, the degree of ambiguity of the content of the answer given by the respondent to the question corresponds to the state information indicating the state of the respondent at the time of the answer and the type of the question. With reference to the storage unit 810 to be attached and stored, the degree of ambiguity associated with the received state information and the type of question Q can be specified as the degree of ambiguity v of the received content information.

As a result, the ambiguity v of the content information can be specified based on the distribution of the ambiguity according to the state of the respondent at the time of answering, which is created in consideration of the type of the question Q, and the respondent RT can specify. It is possible to accurately identify the degree of ambiguity in the answers given.

For example, for each question Q, by preparing a distribution of ambiguity according to the state of the respondent at the time of answering, the characteristics of each question Q such as whether it is easy to answer are considered. Therefore, the degree of ambiguity v can be specified. Further, for each of the open question and the closed question, the distribution of the degree of ambiguity according to the state of the respondent at the time of answering may be prepared. Thereby, the ambiguity v can be specified in consideration of the tendency that the ambiguity of the answer tends to be higher in the open question than in the closed question.

Further, according to the information processing apparatus 101, it is possible to generate learning data in which the received content information is associated with information indicating whether or not the specified ambiguity v exceeds the standard.

This makes it possible to determine whether or not the learning data is effective for learning the hobbies and preferences of the user (respondent RT) of the robot 201.

Further, according to the information processing apparatus 101, it is possible to generate learning data in which the received content information and the content of the question Q are associated with the information corresponding to the specified ambiguity v.

As a result, it is possible to generate learning data that can identify not only the degree of ambiguity of the answer given by the respondent RT but also the content of the question Q given to the respondent RT.

Further, according to the information processing apparatus 101, learning data is generated in which the received content information, the content of the question Q, and the information specifying the respondent RT are associated with the information corresponding to the specified ambiguity v. be able to.

As a result, it is possible to generate learning data that can identify not only the degree of ambiguity of the answer made by the respondent RT and the content of the question Q, but also which user answered. Therefore, even when a plurality of users use the robot 201, the learning data for each user can be determined.

Further, according to the information processing apparatus 101, regarding the user who is the respondent RT, the ambiguity of the content of the answer given by the user to another question Q'different from the question Q is determined by the user at the time of the answer. The second ambiguity level v'corresponding to the received state information and the user who is the respondent RT is specified with reference to the storage unit 810 which is stored in association with the state information indicating the state of. Can be done. Then, according to the information processing apparatus 101, the specified ambiguity v is corrected based on the specified second ambiguity v', and the received content information corresponds to the information corresponding to the corrected ambiguity v. The attached learning data can be generated.

As a result, the ambiguity level v of the content information can be corrected in consideration of the characteristics of the user who is the respondent RT, and the ambiguity level of the answer made by the respondent RT can be accurately specified. For example, the ambiguity level v can be corrected in consideration of individual user characteristics such as whether emotions and emotions are likely to appear in voices and facial expressions. In addition, the ambiguity level v can be corrected in consideration of individual user characteristics such as whether or not a filler such as "um" is likely to be emitted in the middle of a conversation.

From these facts, according to the information processing apparatus 101 according to the embodiment, it becomes possible to collect effective learning data for determining the hobbies and preferences of the user of the robot 201, which is a natural relationship between the robot 201 and the user. It can be useful for realizing communication.

The generation method described in this embodiment can be realized by executing a program prepared in advance on a computer such as a personal computer or a workstation. This generation program can be used on a computer-readable recording medium such as a hard disk, flexible disk, CD (Compact Disc) -ROM, MO (Magnet-Optical disk), DVD (Digital Versaille Disc), or USB (Universal Serial Bus) memory. It is recorded and executed by being read from a recording medium by a computer. Further, this generation program may be distributed via a network such as the Internet.

The following additional notes are further disclosed with respect to the above-described embodiment.

(Appendix 1) Accepts content information indicating the content of the answer given by the respondent to a certain question and status information indicating the state of the respondent at the time of the answer.
Based on the received status information, the ambiguity of the received content information is specified.
Generates information in which the received content information is associated with the specified information according to the degree of ambiguity.
A generator characterized by having a computer perform processing.

(Appendix 2) A storage unit that stores the ambiguity of the content of the answer given by the respondent to the question in association with the state information indicating the state of the respondent at the time of the answer and the type of the question. The degree of ambiguity associated with the received state information and the type of the question is specified with reference to.
Generates information in which the received content information is associated with the specified information according to the degree of ambiguity.
The generation program according to Appendix 1, wherein the processing is executed by the computer.

(Appendix 3) The information according to the ambiguity is information indicating whether or not the specified ambiguity exceeds the standard.
The generation program according to Appendix 1, wherein the program is characterized by the above.

(Appendix 4) Generates information in which the received content information and the content of the certain question are associated with the specified information according to the degree of ambiguity.
The generation program according to Appendix 1, wherein the processing is executed by the computer.

(Appendix 5) Generates information in which the received content information, the content of the question, and the information specifying the respondent are associated with the specified information according to the degree of ambiguity.
The generation program according to Appendix 1, wherein the processing is executed by the computer.

(Appendix 6) Regarding the user who is the respondent, the ambiguity of the content of the answer given by the user to another question different from the question is set to the state information indicating the state of the user at the time of the answer. With reference to the second storage unit to be stored in association with each other, the second ambiguity associated with the received state information and the user who is the respondent is specified.
Based on the identified second ambiguity, the identified ambiguity is corrected.
Generates information in which the received content information is associated with the corrected information according to the ambiguity.
The generation program according to Appendix 2, wherein the processing is executed by the computer.

(Appendix 7) Output the generated information.
The generation program according to Appendix 1, wherein the processing is executed by the computer.

(Appendix 8) The state information indicates at least one of the appearance status of the filler at the time of the answer, the value expressing the emotion of the respondent, and the value expressing the behavior of the respondent.
The generation program according to Appendix 1, wherein the program is characterized by the above.

(Appendix 9) Accepts content information indicating the content of the answer given by the respondent to a certain question and status information indicating the state of the respondent at the time of the answer.
Based on the received status information, the ambiguity of the received content information is specified.
Generates information in which the received content information is associated with the specified information according to the degree of ambiguity.
A generation method characterized by the processing being performed by a computer.

(Appendix 10) A reception unit that accepts content information indicating the content of the answer given by the respondent to a certain question and status information indicating the state of the respondent at the time of the answer.
A specific unit that specifies the ambiguity of the received content information based on the status information received by the reception unit, and a specific unit.
A generation unit that generates information in which information corresponding to the degree of ambiguity specified by the specific unit is associated with the content information received by the reception unit.
An information processing device characterized by having.

101 Information processing device 102 Respondent 110, 810 Storage unit 120 Content information 130 Status information 140, 1200, 1300, 1400, 1500 Ambiguity distribution 150 Ambiguity information 160 Answer information 200 Data collection system 201 Robot 210 Network 220 Question-specific ambiguity Distribution DB
230 Ambiguity distribution DB by user
240 Learning data DB
300,400 Bus 301,401 CPU
302,402 Memory 303,405 I / F
304 Disk drive 305 Disk 403 Display 404 Input device 406 Audio signal processing unit 407 Speaker 408 Microphone 409 Camera 801 Reception unit 802 Specific unit 803 Generation unit 804 Output unit 805 Acquisition unit 806 Analysis unit 1600 Question list for status determination

Claims (10)

Accepts content information indicating the content of the utterance made by the respondent to the question and the appearance status of the filler as state information indicating the state of the respondent at the time of the utterance.
Refer to the information on the ambiguity distribution by question, which records the ambiguity distribution corresponding to the appearance status of the filler.
Based on the appearance status of the received filler, the degree of ambiguity of the received content information is specified.
Generates information in which the received content information is associated with the specified information according to the degree of ambiguity.
A generator characterized by having a computer perform processing. The generation program according to claim 1, wherein the state information further includes at least one of a value representing the emotion of the respondent and a value representing the behavior . Refer to the storage unit that stores the ambiguity of the content of the answer given by the respondent to the question in association with the state information indicating the state of the respondent at the time of the answer and the type of the question. , Identify the degree of ambiguity associated with the received status information and the type of question.
Generates information in which the received content information is associated with the specified information according to the degree of ambiguity.
The generation program according to claim 1, wherein the processing is executed by the computer . The information according to the degree of ambiguity is information indicating whether or not the specified degree of ambiguity exceeds the standard.
The generation program according to claim 1 . Generates information in which the received content information and the content of the question are associated with the specified information according to the degree of ambiguity.
The generation program according to claim 1, wherein the processing is executed by the computer . Generates information in which the received content information, the content of the question, and the information specifying the respondent are associated with the specified information according to the degree of ambiguity.
The generation program according to claim 1, wherein the processing is executed by the computer . For the user who is the respondent, the ambiguity of the content of the answer given by the user to another question different from the question is stored in association with the state information indicating the state of the user at the time of the answer. With reference to the second storage unit, the second degree of ambiguity associated with the received state information and the user who is the respondent is specified.
Based on the identified second ambiguity, the identified ambiguity is corrected.
Generates information in which the received content information is associated with the corrected information according to the ambiguity.
The generation program according to claim 3, wherein the processing is executed by the computer . Output the generated information,
The generation program according to claim 1, wherein the processing is executed by the computer . Accepts content information indicating the content of the utterance made by the respondent to the question and the appearance status of the filler as state information indicating the state of the respondent at the time of the utterance.
Refer to the information on the ambiguity distribution by question, which records the ambiguity distribution corresponding to the appearance status of the filler.
Based on the appearance status of the received filler, the degree of ambiguity of the received content information is specified.
Generates information in which the received content information is associated with the specified information according to the degree of ambiguity.
A generation method characterized by the processing being performed by a computer. A reception unit that accepts content information indicating the content of the utterance made by the respondent to the question, and the appearance status of the filler as state information indicating the state of the respondent at the time of the utterance.
A reference unit that records information on the ambiguity distribution by question, which records the ambiguity distribution corresponding to the appearance status of the filler, and a reference unit.
A specific unit that specifies the degree of ambiguity of the received content information based on the appearance status of the filler received by the reception unit, and a specific unit.
A generation unit that generates information in which information corresponding to the degree of ambiguity specified by the specific unit is associated with the content information received by the reception unit.
An information processing device characterized by having.

Images