Information processing device, information processing method, and program

Abstract

To obtain qualitative data using an avatar from another avatar in a three-dimensional virtual space.SOLUTION: An information processing device (1) includes an avatar control unit (111) and an answer acquisition unit (112). The avatar control unit (111) controls an avatar (CA) in a three-dimensional virtual space (VS) by using a machine learning model (LM) that has machine-learned to control the avatar (CA) to converse with another avatar (OA) and ask the other avatar a question in the conversation. The answer acquisition unit (112) obtains an answer to the question.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and a program for controlling an avatar in a three-dimensional virtual space. [Background technology]

[0002] There are known techniques for acquiring information about avatars in a three-dimensional virtual space. For example, Patent Document 1 discloses an information management device that acquires static information and dynamic information held by an avatar and stores the avatar, static information, and dynamic information in association with each other. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2011-216073 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the information management device described in Patent Document 1 acquires information on how long an avatar stayed at a certain location and how it moved from a certain location to another certain location. In other words, the data that the information management device described in Patent Document 1 can acquire are avatar position data and time data. Therefore, the information management device described in Patent Document 1 has a problem in that it cannot acquire qualitative data such as the content of conversations.

[0005] One aspect of the present invention aims to realize a technology for using an avatar to acquire qualitative data from other avatars in a three-dimensional virtual space. [Means for solving the problem]

[0006] In order to solve the above problem, an information processing device according to one embodiment of the present invention includes an avatar control unit that controls an avatar in a three-dimensional virtual space using a machine learning model that has been trained to control the avatar so that the avatar converses with other avatars and asks questions to the other avatars during the conversation, and an answer acquisition unit that acquires answers to the questions.

[0007] In order to solve the above problem, an information processing method according to one aspect of the present invention includes an avatar control process that controls an avatar in a three-dimensional virtual space using a machine learning model that has been trained to control the avatar so that the avatar converses with another avatar and asks the other avatar a question during the conversation, and an answer acquisition process that acquires an answer to the question.

[0008] The information processing device according to each aspect of the present invention may be realized by a computer, in which case the program of the information processing device that realizes the information processing device on a computer by causing the computer to operate as each part (software element) of the information processing device also falls within the scope of the present invention. [Effects of the Invention]

[0009] According to one aspect of the present invention, an avatar can be used to obtain qualitative data from other avatars in a three-dimensional virtual space. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a block diagram showing a configuration of an information processing device according to a first embodiment of the present invention. [Figure 2] FIG. 1 is a diagram showing an example of a three-dimensional virtual space according to a first embodiment of the present invention. [Figure 3] 1 is a flowchart showing a flow of processing executed by an information processing device according to a first embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0011] [Embodiment 1] Hereinafter, one embodiment of the present invention will be described in detail.

[0012] (Overview of information processing device 1) The information processing device 1 according to this embodiment is a device that controls an avatar CA in a three-dimensional virtual space VS using a machine learning model LM that has undergone machine learning. The information processing device 1 causes the avatar CA to converse with another avatar OA that is different from the avatar CA. The information processing device 1 also causes the avatar CA to ask a question to the other avatar OA during the conversation. The information processing device 1 also acquires an answer ANS to the question QUE.

[0013] As an example, the information processing device 1 acquires answers ANS to questions QUE created by a person who takes a survey. In this configuration, the content of the conversation, the content of the survey, and the avatar CA may be created by the person who takes the survey.

[0014] Furthermore, the question QUE may be made up of multiple questions. For example, the questions QUE may be configured to ask predetermined questions in sequence, or may be configured to select one of multiple questions as the next question depending on the answer to a certain question.

[0015] The information processing device 1 controls an avatar CA in the three-dimensional virtual space VS by, for example, transmitting and receiving data to and from a server that provides the three-dimensional virtual space VS via the Internet. The three-dimensional virtual space VS has a plurality of platforms, and the information processing device 1 controls an avatar CA in one of a plurality of worlds that constitute a certain platform.

[0016] (Configuration of information processing device 1) The configuration of the information processing device 1 will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the configuration of the information processing device 1 according to this embodiment.

[0017] As shown in FIG. 1, the information processing device 1 includes a control unit 11, a storage unit 12, an input / output unit 13, and a communication unit .

[0018] The storage unit 12 stores data referenced by the control unit 11. Examples of the storage unit 12 include a flash memory, a hard disk drive (HDD), a solid state drive (SSD), or a combination of these.

[0019] Examples of data stored in storage unit 12 include a machine learning model LM, a question QUE, an answer ANS, avatar information CAI indicating information about avatar CA, and avatar information OAI indicating information about another avatar OA. Storage unit 12 may associate avatar information OAI indicating another avatar OA, a question QUE asked to the other avatar OA, and an answer ANS from the other avatar OA to the question QUE. Furthermore, storing a machine learning model LM in storage unit 12 means that parameters defining the machine learning model LM are stored in storage unit 12.

[0020] Specific examples of the machine learning model LM include a convolution neural network (CNN), a recurrent neural network (RNN), or a combination thereof. The machine learning model LM may also be a non-neural network model such as a random forest or a support vector machine.

[0021] Examples of avatar information CAI include at least one of information indicating the appearance of the avatar CA, information indicating the behavior (movements and speech content) of the avatar CA, and information indicating the position of the avatar CA.

[0022] "Information indicating the behavior of the avatar CA" includes the momentary movements of the avatar CA (e.g., raising a hand, bowing, etc.) and the movements of the avatar CA over a given period of time (e.g., staying in a booth set up in a three-dimensional virtual space for one minute or more, or watching a display set up in a three-dimensional virtual space for three minutes or more). Furthermore, "movement" includes not only movements that change state, but also movements that remain in a continuous state (e.g., standing, sitting, etc.). The same applies to "information indicating the behavior of other avatars OA" described below.

[0023] An example of avatar information OAI is information transmitted from a server that provides the three-dimensional virtual space VS, and indicates whether other avatars OA exist in the three-dimensional virtual space. Furthermore, if the avatar information OAI indicates that other avatars OA exist in the three-dimensional virtual space, the avatar information OAI may further include at least one of information indicating the appearance of the other avatars OA existing in the three-dimensional virtual space, information indicating the behavior of the other avatars OA, and information indicating the location of the other avatars OA.

[0024] Another example of avatar information OAI is an image displayed to a user using a three-dimensional virtual space. In this case, the information processing device 1 analyzes the image to determine whether or not other avatars OA exist. If the information processing device 1 determines that other avatars OA exist, it further analyzes the image to recognize at least one of the appearances, behaviors, and positions of the other avatars OA.

[0025] The input / output unit 13 is an interface that receives data input from a user or outputs data supplied from the control unit 11 to the user. The input / output unit 13 supplies data indicating the received input to the control unit 11. Examples of the input / output unit 13 include a keyboard, a mouse, a touchpad, a liquid crystal display, and a speaker.

[0026] As one example, the input / output unit 13 receives from the user the contents of the conversation, the contents of the questionnaire, and an operation to create an avatar CA. As another example, the input / output unit 13 displays the answer ANS to the user.

[0027] The communication unit 14 is an interface for transmitting and receiving data via a network. Examples of the communication unit 14 include, but are not limited to, communication chips for various communication standards such as Ethernet (registered trademark), Wi-Fi (registered trademark), and wireless communication standards for mobile data communication networks.

[0028] As one example, the communication unit 14 outputs a signal for controlling the avatar CA to a server that provides the three-dimensional virtual space VS. As another example, the communication unit 14 acquires at least one of the avatar information OAI and the answer ANS from the server that provides the three-dimensional virtual space VS.

[0029] (Control unit 11) The control unit 11 controls each component included in the information processing device 1. As shown in FIG. 1 , the control unit 11 includes an avatar control unit 111, an answer acquisition unit 112, a learning unit 113, an acquisition unit 114, and an output unit 115.

[0030] The avatar control unit 111 controls the avatar using a machine learning model LM that has been trained to control the avatar CA to converse with other avatars OA in the three-dimensional virtual space VS and ask questions to the other avatars OA during the conversation. Examples of methods for the avatar CA to converse with other avatars OA include voice chat and text chat.

[0031] As an example, the avatar control unit 111 controls the avatar CA by outputting control information indicating the content of controlling the avatar CA to a server that provides the three-dimensional virtual space VS via the communication unit 14. Specific processing by the avatar control unit 111 will be described in a processing example of the learning unit 113, which will be described later.

[0032] The answer acquisition unit 112 acquires answers ANS to the questions QUE. The answer acquisition unit 112 stores the acquired answers ANS in the storage unit 12. The answer acquisition unit 112 also determines whether the number of answers ANS has reached a predetermined value. For example, the answer acquisition unit 112 determines whether the number of answers has reached a preset target.

[0033] The learning unit 113 uses the learning data to train the machine learning model LM. An example of the processing by the learning unit 113 will be described later.

[0034] The acquisition unit 114 acquires data via the input / output unit 13 or the communication unit 14. The acquisition unit 114 stores the acquired data in the storage unit 12. As an example, the acquisition unit 114 acquires the content of the conversation, the content of the questionnaire, and avatar information CAI via the input / output unit 13. As another example, the acquisition unit 114 acquires avatar information OAI via the communication unit 14.

[0035] The output unit 115 outputs the data via the input / output unit 13 or the communication unit 14. As an example, the output unit 115 outputs the answer ANS to the input / output unit 13 and presents the answer ANS to the user.

[0036] (Example of 3D virtual space VS) An example of the three-dimensional virtual space VS will be described with reference to Fig. 2. Fig. 2 is a diagram showing an example of the three-dimensional virtual space VS according to this embodiment.

[0037] The three-dimensional virtual space VS contains an avatar CA controlled by the information processing device 1. In the three-dimensional virtual space VS, the information processing device 1 controls the avatar CA so that it can converse and move.

[0038] In addition, one or more other avatars OA controlled by other users may exist in the three-dimensional virtual space VS. In Fig. 2, multiple other avatars OA (other avatar OA1 and other avatar OA2) exist in the three-dimensional virtual space VS.

[0039] Furthermore, a display DP and an object OB for displaying images may be installed in the three-dimensional virtual space VS. In FIG. 2, a display DP1 and a display DP2 are installed in the three-dimensional virtual space VS. A user using the three-dimensional virtual space VS may display an image on the display DP. In this case, a user using the three-dimensional virtual space VS can view the image displayed on the display DP through an avatar (avatar CA, another avatar OA1, and another avatar OA2 in FIG. 2). Examples of an object OB include a chair, a desk, and a flag indicating the venue.

[0040] (Processing example 1 of the learning unit 113) An example of the processing performed by the learning unit 113 will be described.

[0041] The learning unit 113 trains the machine learning model LM using training data including the avatar information OAI and the question QUE. In this case, the learning unit 113 trains the machine learning model LM so that it receives the avatar information OAI as input and outputs control information for controlling the avatar CA.

[0042] As an example, the learning unit 113 trains the machine learning model LM so that when the avatar information OAI input to the machine learning model LM indicates that another avatar OA exists in the three-dimensional virtual space VS, the learning unit 113 outputs control information that controls the avatar CA to converse with the other avatar OA and ask the other avatar OA a question QUE during the conversation.

[0043] In this case, the avatar control unit 111 inputs the avatar information OAI acquired by the acquisition unit 114 to the machine learning model LM. Then, the avatar control unit 111 controls the avatar CA based on the control information output from the machine learning model LM.

[0044] For example, based on the control information output from the machine learning model LM, the avatar control unit 111 controls the avatar CA to have a conversation with another avatar OA1 shown in Fig. 2. Then, the avatar control unit 111 controls the avatar CA to ask the other avatar OA1 a question QUE "What do you think of XXX?" in the conversation.

[0045] Furthermore, the avatar control unit 111 may move the avatar CA to the vicinity of another avatar OA with which the avatar CA is to converse, and then control the avatar CA to converse with the other avatar OA. For example, the avatar control unit 111 may move the avatar CA so that the position of the avatar CA is within a predetermined range from the other avatar OA, and then control the avatar CA to converse with the other avatar OA. The predetermined range is not particularly limited, but may be a range of distance equivalent to that when a person converses with another person.

[0046] For example, the avatar control unit 111 refers to information indicating the position of the other avatar OA1, and moves the avatar CA so that the position of the avatar CA is within a predetermined range from the other avatar OA1 shown in Fig. 2. Then, the avatar control unit 111 controls the avatar CA to converse with the other avatar OA1 and ask the question QUE.

[0047] With this configuration, the information processing device 1 can indicate to the user of the other avatar OA that the avatar CA wishes to converse with the other avatar OA. Also, by indicating to the user of the other avatar OA that the avatar CA wishes to converse with the other avatar OA, the information processing device 1 can smoothly converse with the user of the other avatar OA.

[0048] (Processing example 2 of the learning unit 113) Another example of the processing by the learning unit 113 will be described.

[0049] The learning unit 113 trains the machine learning model LM using learning data including information indicating at least one of the appearance, speech, behavior, and position of other avatars OA. In this case, the learning unit 113 trains the machine learning model LM so that it receives avatar information OAI including information indicating at least one of the appearance, speech, behavior, and position of other avatars OA as input and outputs control information for controlling the avatar CA.

[0050] As an example, when the input avatar information OAI indicates that there are multiple other avatars OA in the three-dimensional virtual space VS, the learning unit 113 trains the machine learning model LM to output control information that controls the avatar CA to converse with any of the multiple other avatars OA and ask the other avatar OA a question QUE in the conversation, based on at least one of the appearance, behavior, and position of each of the multiple other avatars OA.

[0051] Specifically, the learning unit 113 trains the machine learning model LM to output control information that controls the avatar CA to ask a question to another avatar OA that is predicted to give a valid answer (for example, a long answer to a free conversation question, an answer that is not all the same to a multiple-choice question, etc.) among multiple other avatars OA.

[0052] For example, if the appearance of another avatar OA indicates that the other avatar OA belongs to a generation that is interested in the content of the question QUE, if the behavior of another avatar OA indicates that the other avatar OA is looking at an image related to the content of the question QUE or is making a statement related to the content of the question QUE, or if the location of the other avatar OA indicates that the other avatar OA is near a display DP that is displaying an image related to the content of the question QUE, then the machine learning model LM is trained to predict that the other avatar OA will give a valid answer.

[0053] Furthermore, the learning unit 113 may further train the machine learning model LM using learning data including information indicating at least one of the appearance, behavior, and location of another avatar OA, and the answer of the other avatar OA. In this case, the machine learning model LM is trained to predict that another avatar OA that has the same appearance, behavior, and location as the other avatar OA that gave a valid answer will give a valid answer.

[0054] In this case, the avatar control unit 111 inputs the avatar information OAI acquired by the acquisition unit 114 into the machine learning model LM. Then, the avatar control unit 111 controls the avatar CA based on the control information output from the machine learning model LM. That is, when multiple other avatars OA exist in the three-dimensional virtual space VS, the avatar control unit 111 controls the avatar CA to converse with one of the multiple other avatars OA based on at least one of the appearance, speech and behavior, and position of each of the multiple other avatars OA.

[0055] For example, the avatar control unit 111 controls the avatar CA to converse with the other avatar OA1 that the machine learning model LM predicts will give a valid answer, out of the other avatars OA1 and OA2 shown in Fig. 2. Then, the avatar control unit 111 controls the avatar CA to ask the other avatar OA1 the question QUE "Are you interested in XXX?" in the conversation.

[0056] In addition, the avatar control unit 111 may change the question QUE to another avatar OA1 that the machine learning model LM predicts will give a valid answer by referring to information indicating at least one of the appearance, behavior, and position of the other avatar OA1.

[0057] For example, suppose a user asks either question QUE1 or question QUE2. In this case, if the appearance of the other avatar OA1 indicates that the user belongs to a generation that is interested in the content of question QUE2, if the behavior of the other avatar OA1 indicates that the user has been looking at an image related to the content of question QUE2 for a predetermined period of time or more (or indicates that the other avatar OA1 has been staying near an image related to the content of question QUE2 for a predetermined period of time or more), or indicates that the other avatar OA1 has made a statement related to the content of question QUE2, or if the location of the other avatar OA1 indicates that the other avatar OA1 is near a display DP that is displaying an image related to the content of question QUE2, then the avatar control unit 111 controls the avatar CA to ask question QUE2 to the other avatar OA1.

[0058] In this way, the learning unit 113 trains the machine learning model LM so that when there are multiple other avatars OA, the learning unit 113 outputs control information that controls the avatar CA to ask a question to another avatar OA that is predicted to give a valid answer based on information indicating at least one of the appearance, behavior, and position of the other avatars OA.

[0059] Furthermore, when there are multiple other avatars OA, the avatar control unit 111 controls the avatar CA to ask a question to another avatar OA that is predicted to give a valid answer based on information indicating at least one of the appearance, speech and behavior, and position of the other avatars OA.

[0060] With this configuration, when there are multiple other avatars OA, the information processing device 1 can prioritize a question to an avatar OA that is predicted to give a valid answer among the multiple other avatars OA.

[0061] (Processing example 3 of the learning unit 113) Another example of the processing by the learning unit 113 will be described.

[0062] In this example as well, the learning unit 113 trains the machine learning model LM using learning data including information indicating at least one of the appearance, speech, and behavior, and position of other avatars OA. In this example as well, the learning unit 113 trains the machine learning model LM so that it receives avatar information OAI including information indicating at least one of the appearance, speech, and behavior, and position of other avatars OA as input and outputs control information for controlling the avatar CA.

[0063] As an example, the learning unit 113 trains the machine learning model LM to output control information that controls the avatar CA to change the question QUE in the conversation when the input avatar information OAI includes information indicating at least one of the appearance, behavior, and position of another avatar OA.

[0064] Specifically, the learning unit 113 trains the machine learning model LM so as to output control information for controlling the avatar CA to ask a question QUE that is predicted to provide a valid answer from among a plurality of questions QUE.

[0065] For example, if the appearance of another avatar OA indicates that it is from a generation that is interested in the content of question QUE2 out of questions QUE1 and QUE2, if the behavior of another avatar OA indicates that it is looking at an image related to the content of question QUE2 or that it is making a statement related to the content of question QUE2, or if the position of another avatar OA indicates that it is near a display DP that is displaying an image related to the content of question QUE2, then the machine learning model LM is trained to predict that the other avatar OA will give a valid answer to question QUE2.

[0066] In this case, the avatar control unit 111 inputs the avatar information OAI acquired by the acquisition unit 114 to the machine learning model LM. Then, the avatar control unit 111 controls the avatar CA based on the control information output from the machine learning model LM. In other words, the avatar control unit 111 controls the avatar CA to change the content of the question based on at least one of the appearance, behavior, and position of another avatar OA.

[0067] For example, the avatar control unit 111 controls the avatar CA to ask another avatar OA1 shown in Figure 2 the question QUE2 ``What do you think of YYY?'', which the machine learning model LM predicts will provide a valid answer, out of the questions QUE1 ``What do you think of XXX?'' and QUE2 ``What do you think of YYY?''

[0068] In addition, when there are multiple other avatars OA in the three-dimensional virtual space VS, the avatar control unit 111 may decide which of the multiple other avatars OA to converse with based on at least one of the appearance, behavior, and position of each of the multiple other avatars OA.

[0069] For example, if the appearance of the other avatar OA1 indicates that the other avatar OA1 is of a generation that is interested in the content of questions QUE1 and QUE2, the behavior of the other avatar OA1 indicates that the other avatar OA1 is looking at images related to the content of questions QUE1 and QUE2 (or that the other avatar OA1 has stayed near images related to the content of questions QUE1 and QUE2 for more than a predetermined time), or that the other avatar OA1 is making statements related to the content of questions QUE1 and QUE2, or the location of the other avatar OA1 indicates that the other avatar OA1 is near a display DP that is displaying images related to the content of questions QUE1 and QUE2, the avatar control unit 111 decides to have a conversation with the other avatar OA1 from among the multiple other avatars OA.The avatar control unit 111 then controls the avatar CA to ask a question from questions QUE1 and QUE2 that the machine learning model LM predicts will provide a valid answer.

[0070] In this way, the learning unit 113 trains the machine learning model LM to output control information that controls the avatar CA to ask other avatars OA questions QUE that it predicts will produce valid answers based on information indicating at least one of the appearance, behavior, and position of other avatars OA.

[0071] Furthermore, the avatar control unit 111 controls the avatar CA to ask the other avatar OA a question QUE that is predicted to produce a valid answer based on information indicating at least one of the appearance, speech and behavior, and position of the other avatar OA.

[0072] With this configuration, the information processing device 1 can prioritize and ask the other avatar OA a question QUE that is predicted to have a valid answer among a plurality of questions QUE.

[0073] (Processing example 4 of the learning unit 113) Another example of the processing by the learning unit 113 will be described.

[0074] The learning unit 113 trains the machine learning model LM by using learning data including one or more questions QUE posed to another avatar OA and answers ANS to the one or more questions QUE. In this case, the learning unit 113 trains the machine learning model LM so that it receives as input one or more questions QUE and the answers ANS to the one or more questions QUE and outputs control information for controlling the avatar CA.

[0075] As an example, the learning unit 113 trains the machine learning model LM to output control information that controls the avatar CA to change the question QUE in the conversation based on one or more input questions QUE and the answers ANS to each of the one or more questions QUE.

[0076] Specifically, the learning unit 113 trains the machine learning model LM to output control information that controls the avatar CA to ask a question QUE among multiple questions QUE that another avatar OA predicts will give a valid answer to.

[0077] For example, if the answers ANS of another avatar OA to one or more questions QUE are invalid answers (e.g., all "I don't know," all "I think so," etc.), the machine learning model LM is trained to change the question QUE to another avatar OA to a question QUE with content different from the one or more questions QUE already asked.

[0078] In this case, the avatar control unit 111 inputs one or more questions QUE and the answers ANS of another avatar OA to each of the one or more questions QUE, which have been acquired by the answer acquisition unit 112, to the machine learning model LM. Then, the avatar control unit 111 controls the avatar CA based on the control information output from the machine learning model LM. In other words, the avatar control unit 111 controls the avatar CA to change the content of the question based on the answers ANS.

[0079] For example, assume that the avatar control unit 111 controls the avatar CA to ask the question QUE1 "Are you interested in XXX?", the question QUE2 "What do you think about XXX?", and the question QUE3 "Do you want to buy XXX?" to another avatar OA1 shown in Fig. 2. Also assume that the answer acquisition unit 112 acquires the answer ANS1 "I don't know" to the question QUE1, the answer ANS2 "I don't know" to the question QUE2, and the answer ANS3 "I don't know" to the question QUE3.

[0080] In this case, the avatar control unit 111 inputs questions QUE1 to QUE3 and answers ANS1 to ANS3 into the machine learning model LM, and controls the avatar CA to ask the modified question QUE4, "What do you think about YYY?"

[0081] In this way, the learning unit 113 trains the machine learning model LM to output control information that controls the avatar CA to ask a question QUE among multiple questions QUE that another avatar OA is predicted to answer effectively.

[0082] Furthermore, the avatar control unit 111 controls the avatar CA to change the question to a question QUE that is predicted to have a valid answer based on the answer ANS.

[0083] With this configuration, the information processing device 1 can change a question QUE to be posed to another avatar OA that repeatedly gives invalid answers to a question QUE that is predicted to give a valid answer.

[0084] (Processing example 5 of the learning unit 113) Another example of the processing performed by the learning unit 113 will now be described.

[0085] The learning unit 113 trains the machine learning model LM using learning data including avatar information CAI of the avatar CA.

[0086] As an example, the learning unit 113 trains the machine learning model LM using learning data including the appearance of the avatar CA, the words and actions of the avatar CA, the timing at which the avatar CA speaks to other avatars OA, the timing at which the avatar CA asks questions to other avatars OA, and the content of what the avatar CA says to other avatars OA, as well as the answers of the other avatars.

[0087] In other words, when the avatar information OAI input to the machine learning model LM indicates that another avatar OA exists in the three-dimensional virtual space VS, the machine learning model LM is trained to converse with the other avatar OA, ask the other avatar OA a question QUE in the conversation, and output control information that controls the avatar CA so as to obtain an appropriate answer.

[0088] With this configuration, the information processing device 1 can control the avatar CA so as to obtain an appropriate response from another avatar OA.

[0089] (Processing executed by information processing device 1) The flow of processing executed by the information processing device 1 will be described with reference to Fig. 3. Fig. 3 is a flowchart showing the flow of processing executed by the information processing device 1 according to the supplementary embodiment.

[0090] (Step S11) In step S11, the acquisition unit 114 acquires avatar information OAI. The avatar control unit 111 references the avatar information OAI and determines whether or not other avatars OA exist in the three-dimensional virtual space VS.

[0091] If it is determined in step S11 that there is no other avatar OA (step S11: NO), the information processing device 1 executes step S11 again. The information processing device 1 may execute step S11 again after a predetermined time has elapsed.

[0092] (Step S12) If it is determined in step S11 that another avatar OA exists (step S11: YES), in step S12, the avatar control unit 111 refers to the avatar information OAI and determines whether or not a plurality of other avatars OA exist.

[0093] (Step S13) If it is determined in step S12 that there are multiple other avatars OA (step S12: YES), in step S13, the avatar control unit 111 selects one of the multiple other avatars OA to talk to.

[0094] (Step S14) If it is determined in step S12 that there are no other avatars OA (step S12: NO), or after executing step S13, in step S14, the avatar control unit 111 moves the avatar CA to the vicinity of the other avatar OA with which the avatar CA is to be conversed.

[0095] (Step S15) In step S15, the avatar control unit 111 causes the avatar CA to start a conversation with another avatar OA.

[0096] (Step S16) In step S16, the avatar control unit 111 requests the avatar CA to provide an answer ANS to the question QUE.

[0097] (Step S17) In step S17, the avatar control unit 111 determines whether or not the answer ANS to the question QUE requested by the avatar CA in step S16 has been approved.

[0098] (Step S18) If approval is not obtained in step S17 (step S17: NO), in step S18, avatar control unit 111 refers to avatar information OAI and determines whether there is another avatar OS with which the user has not had a conversation.

[0099] If it is determined in step S18 that there is no other avatar OS with which the user is not having a conversation (step S18: NO), the information processing device 1 executes step S18 again.

[0100] On the other hand, if it is determined in step S18 that there is another avatar OS with which the user is not having a conversation (step S18: YES), the information processing device 1 returns to the process of step S13.

[0101] (Step S19) If approval is obtained in step S17 (step S17: YES), in step S19, the avatar control unit 111 causes the avatar CA to ask a question QUE to the other avatar OA.

[0102] (Step S20) In step S20, the answer acquisition unit 112 acquires the answer ANS to the question QUE.

[0103] (Step S21) In step S21, the response acquisition unit 112 determines whether the number of responses has reached a target.

[0104] In step S21, if it is determined that the number of responses has not reached the target (step S21: NO), the information processing device 1 executes the process of step S18.

[0105] On the other hand, if it is determined in step S21 that the number of responses has reached the target (step S21: YES), the information processing device 1 ends the processing shown in FIG.

[0106] (Effects of information processing device 1) In this way, the information processing device 1 according to this embodiment controls the avatar CA using the machine learning model LM that has been trained to control the avatar CA so that the avatar CA converses with other avatars OA in the three-dimensional virtual space VS and asks questions to the other avatars OA during the conversation. The information processing device 1 also acquires answers ANS to questions QUE.

[0107] Therefore, the information processing device 1 uses the avatar CA to ask questions to other avatars OA in conversation in the three-dimensional virtual space VS, and can therefore obtain qualitative data from other avatars OA in the three-dimensional virtual space VS using the avatar CA.

[0108] Furthermore, the information processing device 1 controls the avatar CA using the machine learning model LM. Therefore, the information processing device 1 can reduce the manual labor and time burden.

[0109] [Software implementation example] The functions of the information processing device 1 (hereinafter referred to as the "device") can be realized by a program that causes a computer to function as the device, and a program that causes a computer to function as each control block of the device (particularly each part included in the control unit 11).

[0110] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The control device and storage device execute the program, thereby realizing the functions described in each of the above embodiments.

[0111] The program may be non-transitory and may be recorded on one or more computer-readable recording media. The recording media may or may not be included in the device. In the latter case, the program may be supplied to the device via any wired or wireless transmission medium.

[0112] Furthermore, some or all of the functions of the control blocks can be realized by logic circuits. For example, an integrated circuit in which a logic circuit that functions as each of the control blocks is formed is also included in the scope of the present invention. In addition, the functions of the control blocks can also be realized by, for example, a quantum computer.

[0113] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention.

[0114] [summary] An information processing device according to aspect 1 of this embodiment includes an avatar control unit that controls an avatar using a machine learning model that has been trained to control the avatar so that the avatar converses with another avatar in a three-dimensional virtual space and asks the other avatar a question during the conversation, and an answer acquisition unit that acquires an answer to the question.

[0115] With the above configuration, it is possible to use an avatar in a three-dimensional virtual space to obtain qualitative data from other avatars.

[0116] In an information processing device according to aspect 2 of this embodiment, the avatar control unit according to aspect 1 may move the avatar so that the position of the avatar is within a predetermined range from the other avatar, and then control the avatar to converse with the other avatar.

[0117] The above configuration allows smooth conversation with users of other avatars.

[0118] In an information processing device according to aspect 3 of this embodiment, the avatar control unit according to aspect 1 or 2 may, when there are multiple other avatars in the three-dimensional virtual space, control the avatar to converse with one of the multiple other avatars based on at least one of the appearance, behavior, and position of each of the multiple other avatars.

[0119] With the above configuration, when there are multiple other avatars, a question can be asked preferentially to an avatar that is predicted to give a valid answer.

[0120] In the information processing device according to aspect 4 of this embodiment, the avatar control unit according to any of aspects 1 to 3 may control the avatar to change the content of the question based on at least one of the appearance, behavior, and position of the other avatar.

[0121] With the above configuration, it is possible to ask other avatars questions that are predicted to have valid answers, with priority given to the questions among a plurality of questions.

[0122] In the information processing device according to aspect 5 of this embodiment, the avatar control unit according to any one of aspects 1 to 4 may control the avatar to change the content of the question based on the answer.

[0123] With the above configuration, a question QUE to be posed to another avatar who repeatedly gives invalid answers can be changed to a question QUE that is predicted to give a valid answer.

[0124] The information processing device according to aspect 6 of the present embodiment is in any one of aspects 1 to 5, and may further include a learning unit that uses learning data to train the machine learning model.

[0125] The above configuration reduces the manual labor and time burden.

[0126] In the information processing device according to aspect 7 of this embodiment, the learning unit according to aspect 6 may train the machine learning model using learning data including information indicating at least one of the appearance, behavior, and position of other avatars.

[0127] With the above configuration, when there are multiple other avatars, a question can be asked preferentially to an avatar that is predicted to give a valid answer.

[0128] In the information processing device according to aspect 8 of this embodiment, the learning unit according to aspect 6 or 7 may train the machine learning model using learning data including one or more questions posed to a certain other avatar and answers to each of the one or more questions.

[0129] With the above configuration, a question QUE to be posed to another avatar who repeatedly gives invalid answers can be changed to a question QUE that is predicted to give a valid answer.

[0130] In the information processing device according to aspect 9 of this embodiment, the learning unit according to any of aspects 6 to 8 may train a machine learning model using learning data including the appearance of the avatar, the words and actions of the avatar, the timing at which the avatar speaks to the other avatar, the timing at which the avatar asks the other avatar a question, the content of the words spoken by the avatar to the other avatar, and the answers of the other avatar.

[0131] With the above configuration, it is possible to control an avatar so that an appropriate response is obtained from another avatar.

[0132] The information processing method according to aspect 10 of this embodiment includes an avatar control process that controls an avatar in a three-dimensional virtual space using a machine learning model that has been trained to control the avatar so that the avatar converses with another avatar and asks the other avatar a question during the conversation, and an answer acquisition process that acquires an answer to the question.

[0133] With the above configuration, it is possible to use an avatar in a three-dimensional virtual space to obtain qualitative data from other avatars.

[0134] A program according to aspect 11 of this embodiment is a program for causing a computer to function as an information processing device according to any one of aspects 1 to 8, and causes the computer to function as the avatar control unit and the answer acquisition unit.

[0135] With the above configuration, it is possible to use an avatar in a three-dimensional virtual space to obtain qualitative data from other avatars. [Explanation of symbols]

[0136] 1. Information processing equipment 111 Avatar control unit 112 Answer acquisition part 113 Learning Department 114 Acquisition Department 115 Output section CA Avatar ANS Answer OA Other Avatars QUE Question VS 3D virtual space LM machine learning model

Claims

1. an avatar control unit that controls an avatar using a machine learning model that has been trained by machine learning to control the avatar so that the avatar converses with another avatar and asks the other avatar a question in the conversation in a three-dimensional virtual space; an answer acquisition unit that acquires an answer to the question; An information processing device comprising:

2. the avatar control unit moves the avatar so that the position of the avatar is within a predetermined range from the other avatar, and then controls the avatar to converse with the other avatar.

2. The information processing apparatus according to claim 1, wherein:

3. When a plurality of the other avatars exist in the three-dimensional virtual space, the avatar control unit controls the avatar to converse with any one of the plurality of other avatars based on at least one of appearance, speech and behavior, and position of each of the plurality of other avatars.

3. The information processing apparatus according to claim 1, wherein the information processing apparatus is a computer.

4. the avatar control unit controls the avatar to change the content of the question based on at least one of the appearance, speech, behavior, and position of the other avatar; 3. The information processing apparatus according to claim 1, wherein the information processing apparatus is a computer.

5. the avatar control unit controls the avatar to change the content of the question based on the answer.

3. The information processing apparatus according to claim 1, wherein the information processing apparatus is a computer.

6. a learning unit that uses learning data to train the machine learning model; 3. The information processing apparatus according to claim 1, further comprising:

7. the learning unit trains the machine learning model using learning data including information indicating at least one of appearance, speech and behavior, and position of other avatars; The information processing device according to claim 6 .

8. the learning unit trains the machine learning model by using learning data including one or more questions posed to a certain one of the other avatars and answers to the one or more questions; The information processing device according to claim 6 .

9. the learning unit trains a machine learning model using learning data including an appearance of the avatar, a behavior of the avatar, a timing when the avatar speaks to the other avatar, a timing when the avatar asks a question to the other avatar, a content of the speech of the avatar to the other avatar, and a response of the other avatar; The information processing device according to claim 6 .

10. an avatar control process for controlling an avatar in a three-dimensional virtual space by using a machine learning model that has been trained to control the avatar so that the avatar converses with another avatar and asks the other avatar a question during the conversation; an answer acquisition process for acquiring an answer to the question; An information processing method comprising:

11. 2. A program for causing a computer to function as the information processing device according to claim 1, the program causing the computer to function as the avatar control unit and the answer acquisition unit.

Images