JP2021043723A - Information processing apparatus, information processing method, and program - Google Patents

Abstract

To generate a response model for reproduction of a response from a specific person of a predetermined age.SOLUTION: An information processing apparatus is characterized by having: acquiring means for acquiring information on a response of a specific person together with age information of the specific person upon the response; setting means for setting a predetermined age; extracting means for extracting, from the acquired response information, information on a response corresponding to the predetermined age, based on the predetermined age and the age information; and generating means for generating a learning model for the response of the specific person from the extracted response information.SELECTED DRAWING: Figure 2

Description

The present invention relates to a technique for realizing a response imitating a specific person.

In recent years, along with the improvement of the processing performance of personal computers and smartphones, a technology has been developed in which chatbots and avatars generated on personal computers and smartphones automatically respond to user's questions and inquiries. As a technique related to a chatbot, Patent Document 1 discloses that a long response message is divided into appropriate lengths in a chatbot system that automatically responds to question sentences from terminals of a plurality of users. .. In communication with chatbots and avatars, users are not only able to quickly obtain the necessary information, but are also able to enjoy the communication itself. Therefore, in order to realize more human-like communication with artificial intelligence (AI), conversations in various cases are learned by a computer. Machine-learned avatars can automatically respond as if a real person were responding.

Japanese Unexamined Patent Publication No. 2009-3533

However, communication with a chatbot system or an avatar in the prior art may not give sufficient satisfaction to the user. For example, when communicating with an avatar that imitates a celebrity such as an actor, it is desirable to communicate with an avatar that reproduces the celebrity who was most active in the past, not the avatar that imitates the current celebrity. There is. In addition, when communicating by reproducing the deceased person with an avatar, it may be desired to communicate with the deceased person of a desired age instead of the deceased person immediately before his death.

The present invention has been made in view of the above problems, and an object of the present invention is to generate a learning model for realizing a response imitating a specific person of a predetermined age.

The present invention includes acquisition means for acquiring response information of a specific person together with age information of the specific person at the time of the response, setting means for setting a predetermined age, and the predetermined age and the age information. Based on this, it has an extraction means for extracting response information corresponding to the predetermined age from the response information, and a generation means for generating a learning model of the response of the specific person from the extracted response information. It is characterized by that.

According to the present invention, it is possible to realize a response that imitates a specific person of a predetermined age.

Device and system configuration diagram in the first embodiment Functional configuration diagram in the first embodiment Diagram showing Q & A data and age information Conceptual diagram of the response model in the first embodiment Diagram showing the concept of neural networks Flow chart of response model generation in the first embodiment Flow chart of communication using the response model in the first embodiment Conceptual diagram of the response model in the second embodiment Display example of avatar using response model by age Conceptual diagram of a model that selects a response model by age

(First Embodiment)
Hereinafter, the first embodiment will be described with reference to the drawings. FIG. 1 is a configuration diagram of an apparatus and a system according to the present embodiment. The information processing device 1100 is, for example, a general computer, and generates a response model (learning model) for realizing a response imitating a specific person.

The information processing device 1100 is composed of a CPU 1101, a DRAM 1102, a secondary storage device 1103, a network IF 1104, and an I / O controller 1105. The CPU 1101 is a central processing unit, and performs various processes according to instructions and instructions of a computer program. The DRAM 1102 is a volatile memory and temporarily stores computer programs and various data. The secondary storage device 1103 is a non-volatile storage device such as a hard disk drive (HDD) or a solid state disk (SSD), and stores computer programs and various data. The network 1400 is a wired cable or a wireless network, and various data are transmitted and received to and from an external device via the network 1400. The computer program and various data are acquired from the secondary storage device 1103 or the network 1400, and are temporarily stored in the DRAM 1102. The CPU 1101 performs information processing based on a computer program and various data stored in the DRAM 1102. The I / O controller 1105 is a controller that controls input / output between an externally connected input device 1106, a monitor 1107, and a speaker 1108. The input device 1106 is, for example, a keyboard, mouse, or microphone device, and receives user input. The user can use the input device 1106 to ask a question or ask a question to a virtual specific person reproduced on the information processing device 1100 by text input or voice input. The monitor 1107 is, for example, a general liquid crystal display, and displays the processing result of the information processing apparatus 1100. Computer graphics of a specific person reproduced on the information processing apparatus 1100 may be displayed on the monitor 1107. Further, the response of a specific person may be displayed as text on the monitor 1107. The speaker 1108 may output the response of a specific person reproduced on the information processing device 1100 by voice.

While visually recognizing the computer graphics of a specific person displayed on the monitor 1107, the user can ask a question or the like on the input device 1106 and listen to the response output by the speaker 1108. Communication with the specific person reproduced on the information processing device 1100 may be performed without using the input device 1106, the monitor 1107, and the speaker 1108. For example, it is possible to communicate on Saturday using a client PC or a smartphone 1300 connected to the information processing device 1100 via the network 1400.

FIG. 2 is a functional configuration diagram according to the present embodiment. Each function shown in FIG. 2 is realized by the information processing apparatus 1100 performing processing based on a computer program and various data stored in the DRAM 1102. The storage means 2001 is realized by the DRAM 1102 and the secondary storage device 1103, and holds Q & A data and age information. Q & A data is data that collects responses to questions asked to a specific person. The Q & A data may be text format data or voice format data. Further, as a voluntary remark of a specific person, data including a response to silence may be used. The Q & A data may be collected from the conversation log of the mobile phone or smartphone, or may be collected from the chat log recorded on the personal computer. The age information is information indicating at what age a specific person made each response included in the Q & A data. The age information is stored in association with each response included in the Q & A data. FIG. 3 is a diagram showing Q & A data and age information in this embodiment. As shown in the figure, the Q & A data 3001 contains answers to many questions. In addition, age information 3002 indicating at what age each Q & A was made is associated. The storage means 2001 may store not only the Q & A data 3001 and the age information 3002 of one specific person but also the Q & A data 3001 and the age information 3002 of a plurality of specific persons. Further, the age information 3002 may be set in the age range such as 25 to 30 years old.

The acquisition means 2002 is a means for acquiring the Q & A data and the age information stored in the storage means 2001. Then, the setting means 2003 is a means for setting a predetermined age based on a user instruction or the like. The predetermined age may be set by the user as desired. For example, if you want to reproduce a virtual person when a specific person is 25 years old and communicate with him, set 25 years old. Further, the age range may be set such as 25 to 30 years old. The extraction means 2004 is a means for extracting the corresponding response information from the Q & A data based on the predetermined age set by the setting means 2003. For example, when 25 years old is set by the setting means 2003, the Q & A data 3001 shown in FIG. 3 includes the response information of 25 years old, 50 years old, and 51 years old, so refer to the age information 3002. , Extract response information only for 25 years old. The generation means 2005 generates a response model (learning model) of a specific person by using the response information extracted by the extraction means 2004. The response model, in the simplest form, may be the Q & A data itself extracted by the extraction means 2004. In that case, the model can answer only the questions included in the Q & A data stored in the storage means 2001. The determination means 2006 is a determination means for determining whether or not to perform communication using the generated response model and whether or not to stop the communication. The determination means 2006 stops communication using the response model generated by learning when there is no question from the user for a predetermined time or when the user gives an instruction to stop communication. The input means 2007 is a means for receiving a question from the user and inputting the question data into the response model. When there are a plurality of generated response models, a desired response model is selected based on the input from the input means 2007. The output means 2008 is a means for outputting the response of the response model to the question from the user.

FIG. 4 is a conceptual diagram of an example of the response model in the present embodiment. As described above, when the Q & A data itself extracted by the extraction means 2004 is used as the response model, only the questions included in the Q & A data stored in the storage means 2001 can be answered. Therefore, in order to respond to questions other than the questions included in the Q & A data stored in the storage means 2001, the semantic analysis model 4102 is introduced. The semantic analysis model 4102 is created in advance by using machine learning such as a neural network or a support vector machine. The semantic analysis model 4102 inputs question data in text format or voice format, and outputs a feature amount 4103 of the question that can identify the question. By performing machine learning using many question data with the same label attached to similar questions as teacher data, it is possible to create a semantic analysis model 4102 that outputs the same features for similar questions. When the Q & A data is relatively small, the feature amount 4103 may be a scalar value, but when the Q & A data is large, the feature amount 4103 may be a vector value. Here, the concept of a neural network, which is one of machine learning, will be described.

Since the principle of the neural network itself is known, it will be briefly described. FIG. 5 is a diagram illustrating a neural network. In FIG. 5, the intermediate layer is one layer, but it is desirable that the intermediate layer is composed of two or more layers. In the neural network shown in FIG. 5, the input layer has Mi nodes (n11, n12, ..., N1Mi), the intermediate layer has Mh nodes (n21, n22, ..., N2Mh), and the output layer. The (final layer) has Mo nodes (n31, n32, ..., N3Mo). The nodes of each layer are connected to all the nodes of the adjacent layers to form a three-layer hierarchical neural network that transmits information between the layers.

When an image is input to the input layer, the input layer is provided with nodes for the number of pixels so that the pixels and the nodes are on a one-to-one basis. Also, in the output layer, nodes for the number of pixels to be output are set. That is, in the present embodiment, since the block image of 16 pixels × 16 pixels is input and the pixel value of 16 pixels × 16 pixels is output, the number of nodes in the input layer and the output layer is 256. Data is passed from left to right in FIG. 5, that is, in the order of input layer, intermediate layer, and output layer. Each node in the input layer is connected to all the nodes in the middle layer, and the connections between the nodes have their own weights. The output value transmitted from one node to the other node through the join is enhanced or attenuated by the weight of the join. The set of weighting factors and bias values defined for such a connection is a parameter of the learning model. The activation function is not particularly limited, but a logistic sigmoid function, a Rectifier Unit (ReLU) function, or the like may be used. As a learning method, various proposed neural network learning methods may be applied. For example, the difference between the output obtained from the output layer when the student data is input to the input layer and the neural network is operated and the teacher data associated with the student data in advance is calculated, and the difference is calculated. Adjust the weighting coefficient and bias value to minimize it.

As shown in FIG. 4, the response model 4100 is created by connecting the semantic analysis model 4102 created as described above and the Q & A model 4101 created from the response information extracted by the extraction means 2004 in series. Can be done. The Q & A model 4101 is a set of Q & A data, but in order to connect with the semantic analysis model 4102, the input question is changed to the feature amount of the semantic analysis model 4102. As the processing of the response model 4100, first, when a question is input, the semantic analysis model 4102 outputs a feature amount 4103 that can identify the question. The Q & A model 4101 takes the feature amount 4103 output by the semantic analysis model 4102 as an input, and outputs a response corresponding to the feature amount 4103. Through the above processing, it is possible to realize a response model 4100 capable of answering questions other than the questions stored in advance. In the above embodiment, the semantic analysis model 4102 is generated by machine learning separately from the Q & A model 4101, but the response model 4100 may be created including the Q & A model 4101 at the machine learning stage.

FIG. 6 is a flowchart showing the generation of the response model in the present embodiment. Hereinafter, each step will be described. Each of the following steps is performed by at least one of the information processing device 1100, the client PC 1200, and the smartphone 1300.

S601 is a step in which the acquisition means 2002 acquires the Q & A data of a specific person and the corresponding age information from the storage means 2001. The specific person may be set in advance or may be set based on the user's instruction.

S602 is a step in which the setting means 2003 sets the age desired by the user. The age set here may be in the age range such as 25 to 30 years old instead of a specific age. In addition, the age may be calculated based on the date of birth of a specific person by specifying it in the Christian era or the era name.

S603 is a process in which the extraction means 2004 extracts the corresponding response information from the Q & A data of a specific person based on the predetermined age set in S602. For example, when 25 years old is set in S602, only the response information of 25 years old is extracted from the Q & A data of a specific person by referring to the age information acquired in S601.

S604 is a step of generating a response model using the response information extracted in S603. Since the details of generating the response model have been described above, they will be omitted here.

By the above processing, it is possible to generate a response model capable of reproducing the response of a specific person of a desired age by the user.

Next, a method of communicating using the generated response model using at least one of the information processing device 1100, the client PC 1200, and the smartphone 1300 will be described.

FIG. 7 is a flowchart of communication using the response model in the first embodiment. Hereinafter, each step will be described.

S701 is a step in which the input means 2007 selects a response model of a specific person of a desired age to communicate based on an input instruction from the user. Here, the selection is made on the premise that a response model of a plurality of specific persons or a plurality of ages has already been generated. However, if only one desired response model is generated, the selection in this step can be omitted.

S702 is a step of determining whether or not the determination means 2006 stops communication with the response model selected in S701. The determination is made based on a user's input instruction or the elapse of a preset predetermined time. If it is determined that communication will be stopped, the processing in this flow will be terminated. If it is determined not to stop the communication, that is, if it is determined to continue the communication, the process proceeds to S703.

S703 is a step in which the determination means 2006 determines whether or not there is a question from the user. Questions from the user are input in text format or voice format data using the input device 1106, the client PC 1200, and the smartphone 1300. Further, when a spontaneous remark from a specific person is desired, by setting the spontaneous mode in advance, even if there is no question from the user, the process proceeds as if the question was input. When the question is input, it is determined that there is a question, and the process proceeds to S704. If it is determined that there is no question, the process is returned to S702 in order to determine whether or not to stop the communication.

S704 is a step in which the output means 2008 outputs a response to the question input in S703. As for the response, the question input in S703 is input to the response model, and the response is output from the response model. The response is output from the speaker 1108, the monitor 1107, the client PC 1200 via the network 1400, and the smartphone 1300. After the output, the process is returned to S703 in order to receive a question from the next user.

As described above, according to the present embodiment, it is possible to create a response model of a specific person of a user's desired age and communicate using the generated response model.

(Second embodiment)
In the first embodiment, the semantic analysis model 4102 and the Q & A model 4101 are connected in series to generate the response model 4100. However, it may be better to create response models in different ways in order to properly respond to a variety of user questions in a way that is unique to the target person.

In general, it is considered that whether or not a person seems to be a specific person is determined by the content of the response and the tone such as the habit of the ending and the presence or absence of a dialect. Therefore, in the present embodiment, a response model incorporating a tone model is generated. Since the method of generating the response model is the same as that of the first embodiment, only the method of generating the response model in the present embodiment will be described below.

FIG. 8 is a conceptual diagram of the response model 8100 according to the present embodiment. In FIG. 8, the semantic analysis model 8102 is the same as the semantic analysis model 4102 in the first embodiment, and the feature amount 8103 is the same as the feature amount 4103 in the first embodiment. The response model 8100 in the present embodiment is realized by connecting the semantic analysis model 8102, the retrained standard response model 8101, and the tone model 8104 in series.

The retrained standard response model 8101 is generated by retraining the standard response model using the Q & A data 3001 and the age information 3002. The standard response model is a model capable of performing a general standard response, and can be generated by using a standard conversation log or the like as teacher data. It is not possible to learn the response unique to a specific person from a standard conversation log, but it is possible to collect a huge amount of Q & A data. Therefore, it is possible to generate a response model that can respond to various questions. In the present embodiment, the standard response model capable of responding to various questions is relearned using the Q & A data 3001 and the age information 3002. That is, re-learning is performed using the Q & A data 3001 corresponding to the age set by the user, and the response included in the Q & A data 3001 is performed for the question included in the Q & A data 3001. Then, it is learned to give a standard response to a question not included in the Q & A data 3001.

The tone model 8104 is a model in which the tone is learned by using the Q & A data 3001 corresponding to the age set by the user for the model that outputs the input text or voice as it is. That is, the input text or voice is trained to output the same content but with the tone of a specific person such as a habit of ending a word or a dialect.

In the response model 8100 in the present embodiment, as in the first embodiment, when a question is input, the semantic analysis model 8102 outputs a feature amount 8103 that can identify the question. The retrained standard response model 8101 takes the feature amount 8103 output by the semantic analysis model 8102 as an input, and outputs a response corresponding to the feature amount 8103. The response here is a specific person-like response to the question included in the Q & A data 3001, but is a standard response to the question not included in the Q & A data 3001. The output of the relearned standard response model 8101 is input to the tone model 8104, and is output with a response of a tone that seems to be a specific person. By the above processing, even if the output from the relearned standard response model 8101 is a standard response, the tone model 8104 outputs a response with a tone that seems to be a specific person, so that the user feels like a specific person. Can be done. If the output from the relearned standard response model 8101 is a question included in the Q & A data 3001, it is possible to make both the content and tone of the response look like a specific person.

The learning of the model in this embodiment can be realized by using each machine learning method such as the neural network and the support vector machine described above. As described above, according to the present embodiment, it is possible to generate a response model that appropriately responds to various questions of the user in a manner specific to the target specific person.

FIG. 9 shows an example of displaying an avatar using an age-specific response model. By displaying the image of the avatar by age shown in FIG. 9 on any of the monitor 1107, the client PC 1200, and the smartphone 1300, the user can feel as if they are facing different ages and avatars. .. Then, if a pre-created response model for each age is set for the avatar of each age, it is possible to respond to the user's question or question uniquely for each age, so that the users have different ages. You can get the feeling of having a conversation with your avatar.

FIG. 10 is an example of a response model for realizing the response by the avatar of FIG. 9, and is a conceptual diagram of a model in which the response model is selected according to age. As shown in the figure, the response model 10100 includes a standard response model 10101 retrained at age 25, a standard response model 10201 retrained at age 50, and a standard response model 10301 retrained at age 70. ing. Further, a 25-year-old tone model 10104 is connected to the standard response model 10101 that has been relearned at the age of 25. A 50-year-old tone model 10204 is connected to the standard response model 10201 relearned at the age of 50, and a 70-year-old tone model 10304 is connected to the standard response model 10301 relearned at the age of 70. The function of the model is similar to the response 8100 shown in FIG. The major difference is that the semantic analysis model and age selection 10102 for input (question) performs age selection in addition to semantic analysis of input (question), and features the model corresponding to the selected age. The quantity 10103 is output. The age selection may be preset or may be selected according to a user's setting instruction. Moreover, you may select not only a specific age but also the range of ages. With such a configuration, a response model corresponding to each age-specific avatar shown in FIG. 9 is selected, and an appropriate output (response) is made.

(Other Examples)
The present invention supplies a program that realizes one or more functions of the above-described embodiment to a system or device via a network or storage medium, and one or more processors in the computer of the system or device reads and executes the program. It can also be realized by the processing to be performed. It can also be realized by a circuit (for example, ASIC) that realizes one or more functions.

2002 Acquisition means 2003 Setting means 2004 Extraction means 2005 Generation means

Claims (11)

An acquisition means for acquiring the response information of the specific person together with the age information of the specific person at the time of the response, and
Setting means to set a predetermined age and
An extraction means for extracting response information corresponding to the predetermined age from the response information based on the predetermined age and the age information.
An information processing apparatus comprising: a generation means for generating a learning model of a response of a specific person from the extracted response information. The information processing device according to claim 1, wherein the setting means sets a predetermined age within an age range. The information processing device according to claim 1 or 2, wherein each of the responses included in the response information is associated with the age information of the specific person. The information processing apparatus according to any one of claims 1 to 3, wherein the response learning model is a model that outputs a response corresponding to the predetermined age in response to an input from a user. .. The information processing apparatus according to any one of claims 1 to 4, wherein the generation means generates a response learning model including an analysis model that analyzes the meaning of input from a user. The information processing apparatus according to any one of claims 1 to 5, wherein the generation means generates a learning model of a response including a tone model that reproduces the tone of the specific person of a predetermined age. .. Any one of claims 1 to 6, wherein the generation means generates a learning model of the response based on the extracted response information and a standard response model that performs a standard response. The information processing device described in. The information processing apparatus according to any one of claims 1 to 7, wherein the generation means uses a neural network to generate a learning model of the response. The information processing apparatus according to any one of claims 1 to 8, further comprising a display means for displaying an avatar associated with a learning model of the response of the specific person. An acquisition step in which the acquisition means acquires the response information of the specific person together with the age information of the specific person at the time of the response.
The setting means is a setting process for setting a predetermined age, and
An extraction step in which the extraction means extracts the response information corresponding to the predetermined age from the response information based on the predetermined age and the age information.
An information processing apparatus characterized in that the generation means includes a generation step of generating a learning model of the response of the specific person from the extracted response information. Computer,
An acquisition means for acquiring the response information of the specific person together with the age information of the specific person at the time of the response, and
Setting means to set a predetermined age and
An extraction means for extracting response information corresponding to the predetermined age from the response information based on the predetermined age and the age information.
A program that functions as an information processing device, which comprises a generation means for generating a learning model of the response of the specific person from the extracted response information.

Images