Information processing device and method

Abstract

An information processing device according to one embodiment includes: an input unit that receives input of behavior data including environment information indicating an environment in which a person acts, and furthermore including one of person attribute information indicating an attribute of the person and behavior information indicating an action of the person; and a generation unit that generates information including the environment information included in the behavior data, furthermore including one of the person attribute information and the behavior information included in the behavior data, and furthermore including a question sentence asking for the other of the person attribute information and the behavior information included in the behavior data, inputs the generated information to a model, and generates information including a response of the information requested in the question sentence on the basis of a result of the input.

Description

Information processing device and method

[0001] FIELD Embodiments of the present invention relate to an information processing apparatus and method.

[0002] The technology for simulating the behavior of customers (agents) is expected to be useful for marketing purposes such as sales forecasting, by simulating customer behavior in various retail stores, etc., and observing the behavior of potential customers to change or rearrange product types.

[0003] Non-Patent Document 1 discloses a method for constructing a Large Language Model (hereinafter referred to as LLM) that simulates the behavior of an agent in a game space by creating training data for the agent's behavior using random simulation or Monte Carlo tree search methods, and then using this created training data to fine-tune the LLM.

[0004] Jiannan Xiang, Tianhua Tao, Yi Gu, Tianmin Shu, Zirui Wang, Zichao Yang and Zhiting Hu, “Language Models Meet World Models: Embodied Experiences Enhance Language Models” arXiv:2305.10626 2023.

[0005] The method disclosed in Non-Patent Document 1 does not take into consideration the attributes of the agent, and therefore is unable to estimate a simulation of the agent's behavior that is typical of the person.

[0006] Furthermore, the method disclosed in Non-Patent Document 1 uses random simulation results as learning data, so it learns behavioral data in unrealistic contexts and generates behaviors in unnatural contexts.

[0007] The present invention has been made in light of the above circumstances, and an object of the present invention is to provide an information processing apparatus and method that are capable of generating an appropriate behavior sequence.

[0008] An information processing device according to one aspect of the present invention comprises an input unit that accepts input of behavioral data that includes environmental information that indicates the environment in which a person behaves, and further includes one of person attribute information that indicates the attributes of the person and behavioral information that indicates the behavior of the person; and a generation unit that generates information that includes the environmental information included in the behavioral data, further includes one of the person attribute information and the behavioral information included in the behavioral data, and further includes a question that inquires about the other of the person attribute information and the behavioral information included in the behavioral data, inputs the generated information into a model, and generates information that includes an answer to the information asked in the question based on the result of this input.

[0009] An information processing method according to one aspect of the present invention is a method performed by an information processing device, comprising: an input unit of the information processing device accepting input of behavioral data including environmental information indicating the environment in which a person behaves, and further including one of person attribute information indicating the attributes of the person and behavioral information indicating the behavior of the person; a generation unit of the information processing device generating information including the environmental information included in the behavioral data, further including one of the person attribute information and the behavioral information included in the behavioral data, and further including a question inquiring about the other of the person attribute information and the behavioral information included in the behavioral data; inputting the generated information into a model; and generating information including an answer to the information asked in the question based on the result of this input.

[0010] According to the present invention, an appropriate behavior sequence can be generated.

[0011] FIG. 1 is a diagram showing an application example of a behavior simulation device according to an embodiment of the present invention. FIG. 2 is a conceptual diagram showing an example of the configuration of various data and various models during sentence generation model learning and inference by the behavior simulation device. FIG. 3 is a flowchart showing an example of a sentence generation model learning processing routine executed by the behavior simulation device. FIG. 4 is a flowchart showing a first example of a sentence generation model inference processing routine executed by the behavior simulation device. FIG. 5 is a flowchart showing a second example of a sentence generation model inference processing routine executed by the behavior simulation device. FIG. 6 is a conceptual diagram showing an example of the configuration of various data and various models during behavior dataset expansion executed by the behavior simulation device. FIG. 7 is a flowchart showing an example of a behavior dataset expansion processing routine executed by the behavior simulation device. FIG. 8 is a block diagram showing an example of the hardware configuration of a behavior simulation device according to an embodiment of the present invention.

[0012] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below with reference to the accompanying drawings, in which: FIG. 1 is a diagram showing an example of an application of a behavior simulation device according to an embodiment of the present invention;

[0013] The behavior simulation device 100 shown in FIG. 1 is configured as a computer equipped with a CPU (Central Processing Unit), RAM (Random Access Memory), and ROM (Read Only Memory) in which a program for executing a behavior simulation routine, which will be described later, is stored, and is functionally configured as follows.

[0014] As shown in FIG. 1, the behavioral simulation device 100 according to this embodiment is configured to include a memory unit 1, an input sentence / correct answer sentence creation unit 2, an answer sentence generation unit 3, a parameter update unit 4, a data extension determination unit (selection unit) 5, and an output unit 6.

[0015] 2 is a conceptual diagram showing an example of the configuration of various data and various models during sentence generation model learning and inference by the behavior simulation device. The storage unit 1 includes a storage device such as a non-volatile memory, and stores a behavior dataset A and a sentence generation model x.

[0016] The behavioral dataset A includes one or more behavioral data a. The behavioral data a is composed of two or more pieces of information selected from the group consisting of person attribute information b, environmental information c, and behavioral information d. The person attribute information b may be any text information (hereinafter sometimes referred to as text) that indicates a person's attributes, such as a sentence describing the person's age, gender, or height. The environmental information c may be any text information that indicates the surrounding environment in which the person is behaving, such as words that describe the characteristics of a retail store, such as an unmanned convenience store in an office. The behavioral information d may be any text information that indicates a person's behavior, such as a sentence describing products that the person picked up or purchased in a retail store.

[0017] The sentence generation model x may be any neural network that inputs a sentence as text information and outputs a sentence as text information. For example, the sentence generation model x may be a neural network as described in the literature "Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gray, Benjamin Chess, Jack Clark, Christopher Berner, Sam McCandlish, Alec Radford, Ilya Sutskever, Dario Amodei, "Language Models are Few-Shot Learners" This is "GPT-3" disclosed in "arXiv:2005.14165".

[0018] The input sentence / correct answer sentence creation unit 2 outputs an input sentence f and a correct answer sentence g when it receives behavioral data a including all of person attribute information b, environmental information c, and behavioral information d from the storage unit 1. The input sentence f may be configured to include environmental information c and person attribute information b, and further include a question sentence e that asks about the behavioral information d, or may be configured to include environmental information c and behavioral information d, and further include person attribute information b, and any text information including this information will do.

[0019] The correct answer sentence g may be any text indicating the behavioral information d when the input sentence f includes the environmental information c and the person attribute information b as described above. Also, the correct answer sentence g may be any text information indicating the person attribute information b when the input sentence f includes the environmental information c and the behavioral information d as described above.

[0020] Furthermore, when the input sentence / correct answer sentence creation unit 2 receives behavioral data a composed of environmental information c and person attribute information b, or composed of environmental information c and behavioral information d, from the storage unit 1, it outputs an input sentence f and a correct answer sentence g. In this case, the input sentence f includes two pieces of information included in the behavioral data a and further includes a question sentence asking about the remaining information, and any text information including these may be used. The correct answer sentence g may be any text information indicating the person attribute information b or behavioral information d, which is included in the behavioral data a and is asked about in the question sentence.

[0021] Furthermore, the input sentence / correct answer sentence creation unit 2 receives behavioral data a including environmental information c and person attribute information b from the storage unit 1, and when it receives a generated answer sentence h including the behavioral information, it outputs only the input sentence f. This output input sentence f includes the environmental information c and further includes the behavioral information included in the generated answer sentence h, and any text information including these may be used.

[0022] Furthermore, the input sentence / correct answer sentence creation unit 2 receives behavioral data a including environmental information c and behavioral information d from the storage unit 1, and when it receives a generated answer sentence h including person attribute information, it outputs only the input sentence f. This output input sentence f includes the environmental information c and further includes the person attribute information included in the generated answer sentence h, and any text information including these may be used.

[0023] The answer sentence generation unit 3 receives the input sentence f from the input sentence / correct answer sentence creation unit 2, receives the sentence generation model x from the memory unit 1, inputs the received input sentence f to the sentence generation model x, and outputs the generated answer sentence h output from the sentence generation model x as the result of generating an answer sentence to the question sentence in the input sentence f.

[0024] The parameter update unit 4 receives the generated answer sentence h from the answer sentence generation unit 3, receives the correct answer sentence g from the input sentence / correct answer sentence creation unit 2, and receives the sentence generation model x from the memory unit 1, updates the parameters of the neural network of the sentence generation model x so that the generated answer sentence h satisfies the constraint that it approaches the correct answer sentence g, and outputs the sentence generation model x with the updated parameters to the memory unit 1 for storage.

[0025] The constraint that the generated answer sentence h approaches the correct answer sentence g may be any learning method that is set so that the generated answer sentence h and the correct answer sentence g approach or become the same. For example, the parameter update unit 4 calculates the cross-entropy error from the generated answer sentence h and the correct answer sentence g as in the following formula (1), and updates the parameters of the sentence generation model x so as to reduce this error or set it to zero.

[0026]

[0027] Here, k in formula (1) is the index of the generated answer sentence h and the correct answer sentence g, and y ｋ is the value in the generated answer sentence h, and t ｋ is the value in the correct answer sentence g. ｋ Only the index value of the correct class is "1", and the other index values ​​are "0".

[0028] The data extension determination unit 5 receives from the storage unit 1 behavioral data a consisting of two of the three pieces of information consisting of person attribute information b, environmental information c, and behavioral information d, receives from the answer sentence generation unit 3 a plurality of second generated answer sentences h based on a plurality of first generated answer sentences h based on the behavioral data a, receives a correct answer sentence g from the input sentence / correct answer sentence creation unit 2, and calculates a score indicating the degree of word agreement between each of the plurality of second generated answer sentences h and the correct answer sentence g. Based on the calculated score, the data extension determination unit 5 selects a generated answer sentence h from the plurality of first generated answer sentences h that is suitable for interpolating the behavioral data a, and interpolates the behavioral data a using information included in the selected generated answer sentence h, thereby outputting behavioral data a consisting of the three pieces of information consisting of person attribute information b, environmental information c, and behavioral information d.

[0029] The score indicating the degree of word agreement between the generated answer sentence h and the correct answer sentence g can be any score that indicates the word agreement rate, for example, the BLUE score disclosed in "Chin-Yew Lin and Eduard Hovy, "Automatic Evaluation of Summaries Using N-gram Co-Occurrence Statistics" In NAACL2003".

[0030] Furthermore, the method for selecting the generated answer sentence h may be any method that selects the generated answer sentences with the highest scores indicating the degree of word agreement, such as a method of selecting the top three generated answer sentences. The output unit 6 receives the generated answer sentence h from the answer sentence generation unit 3, and outputs and stores the generated answer sentence h in the memory unit 1 as an inference result based on the sentence generation model x.

[0031] <Functions of Behavior Simulation Device> Next, we will explain the functions of the behavior simulation device 100 according to this embodiment. The behavior simulation device 100 executes a sentence generation model learning processing routine (hereinafter sometimes referred to as a learning processing routine), a sentence generation model inference processing routine (hereinafter sometimes referred to as an inference processing routine), and a behavior dataset expansion processing routine, all of which will be described below.

[0032] <<Sentence Generation Model Learning Processing Routine>> First, the learning processing routine will be described. Fig. 3 is a flowchart showing an example of the sentence generation model learning processing routine executed by the behavior simulation device. In step S101-1, the input sentence / correct answer sentence creation unit 2 receives behavior data a including all of the person attribute information b, environmental information c, and behavior information d from the storage unit 1. The input sentence / correct answer sentence creation unit 2 outputs an input sentence f, which is text including the person attribute information b and environmental information c included in the behavior data a and further including a question sentence inquiring about the behavior information d, and a correct answer sentence g, which is text indicating the behavior information d included in the received behavior data a, and transmits the input sentence f to the answer sentence generation unit 3 and transmits the correct answer sentence g to the parameter update unit 4.

[0033] In step S101-2, the input sentence and correct answer sentence creation unit 2 receives behavioral data a including all of the person attribute information b, environmental information c, and behavioral information d from the storage unit 1. The input sentence and correct answer sentence creation unit 2 outputs an input sentence f, which is text including the behavioral information d and environmental information c included in the behavioral data a and further including a question sentence inquiring about the person attribute information b, and a correct answer sentence g, which is text indicating the person attribute information b included in the received behavioral data a, and transmits the input sentence f to the answer sentence generation unit 3 and transmits the correct answer sentence g to the parameter update unit 4.

[0034] In step S102, as a first process, the answer sentence generation unit 3 accepts the input sentence f transmitted from the input sentence / correct answer sentence creation unit 2 in step S101-1 above, accepts the sentence generation model x from the memory unit 1, inputs the input sentence f to the sentence generation model x, outputs a generated answer sentence h output from this sentence generation model x as the generated answer sentence for the question sentence in the input sentence f, and transmits this generated answer sentence h to the parameter update unit 4.

[0035] Also, in this step S102, as a second process, the answer sentence generation unit 3 accepts the input sentence f transmitted from the input sentence / correct answer sentence creation unit 2 in the above step S101-2, accepts the sentence generation model x from the memory unit 1, inputs the input sentence f to the sentence generation model x, outputs the generated answer sentence h output from this sentence generation model x as the generated answer sentence for the question sentence in the input sentence f, and transmits this generated answer sentence h to the parameter update unit 4.

[0036] In step S103, as a first process, the parameter update unit 4 accepts the generated answer sentence h transmitted from the answer sentence generation unit 3 by the first process of step S102 above, accepts the correct answer sentence g corresponding to the generated answer sentence h from the input sentence / correct answer sentence creation unit 2, that is, the correct answer sentence g transmitted from the input sentence / correct answer sentence creation unit 2 by step S101-1 above, calculates a constraint such that the generated answer sentence h approaches the correct answer sentence g, updates the parameters of the sentence generation model x so that this constraint is satisfied, and transmits the sentence generation model x with the updated parameters to the storage unit 1.

[0037] Also, in this step S103, as a second process, the parameter update unit 4 accepts the generated answer sentence h transmitted from the answer sentence generation unit 3 by the second process of the above step S102, accepts the correct answer sentence g corresponding to the generated answer sentence h from the input sentence / correct answer sentence creation unit 2, that is, the correct answer sentence g transmitted from the input sentence / correct answer sentence creation unit 2 by the above step S101-2, calculates a constraint such that the generated answer sentence h approaches the correct answer sentence g, updates the parameters of the sentence generation model x so that this constraint is satisfied, and transmits the sentence generation model x with the updated parameters to the memory unit 1.

[0038] <<Sentence Generation Model Inference Processing Routine>> Next, a first example of the inference processing routine will be described. Fig. 4 is a flowchart showing a first example of the sentence generation model inference processing routine executed by the behavior simulation device. First, the processes of step S101-1, step S101-2, and step S102 described in the learning processing routine are performed in this order.

[0039] Next, in step S203, the output unit 6 accepts the generated answer sentence h transmitted from the answer sentence generation unit 3 in step S102, and outputs and stores this generated answer sentence h in the memory unit 1 as the inference result based on the sentence generation model x.

[0040] Next, a second example of the inference processing routine will be described. FIG. 5 is a flowchart showing a second example of the sentence generation model inference processing routine executed by the behavior simulation device. First, in step S201-1, the input sentence / correct answer sentence creation unit 2 receives behavior data a including person attribute information b and environmental information c from the storage unit 1. The input sentence / correct answer sentence creation unit 2 outputs an input sentence f, which is text including the person attribute information b and environmental information c included in the behavior data a and further including a question sentence asking about behavior information d, and transmits this input sentence f to the answer sentence generation unit 3.

[0041] In step S201-2, the input sentence and correct answer sentence creation unit 2 receives behavioral data a including environmental information c and behavioral information d from the storage unit 1. The input sentence and correct answer sentence creation unit 2 outputs an input sentence f, which is text including the behavioral information d and environmental information c included in this behavioral data a and further including a question sentence asking about person attribute information b, and transmits this input sentence f to the answer sentence generation unit 3.

[0042] Next, the processes are performed in the order of step S102 described in the above learning processing routine and step S203 described in the above first example of the inference processing routine. However, in step S102 in the second example of the inference processing routine, as a first process, the answer sentence generation unit 3 receives the input sentence f transmitted from the input sentence / correct answer sentence creation unit 2 in step S201-1 above, receives the sentence generation model x from the storage unit 1, inputs the input sentence f to the sentence generation model x, outputs a generated answer sentence h output from this sentence generation model x as a generated answer sentence for the question sentence in the input sentence f, and transmits this generated answer sentence h to the parameter update unit 4.

[0043] Furthermore, in step S102 in the second example of the inference processing routine, as a second process, the answer sentence generation unit 3 accepts the input sentence f transmitted from the input sentence / correct answer sentence creation unit 2 in step S201-2 above, accepts the sentence generation model x from the memory unit 1, inputs the input sentence f to the sentence generation model x, outputs the generated answer sentence h output from this sentence generation model x, and transmits this generated answer sentence h to the parameter update unit 4.

[0044] <<Behavior Data Set Expansion Processing Routine>> Next, the behavior data set expansion processing routine will be described. Fig. 6 is a conceptual diagram showing an example of the configuration of various data and various models when behavior data set expansion is executed by the behavior simulation device. Fig. 7 is a flowchart showing an example of the behavior data set expansion processing routine executed by the behavior simulation device. The behavior data set expansion processing routine is executed as a process to interpolate the missing information when, for example, person attribute information b or behavior information d is missing for behavior data a included in behavior data set A in the previous stage of the learning processing routine.

[0045] In step S301-1, the input sentence / correct answer sentence creation unit 2 receives behavioral data a including person attribute information b and environmental information c from the storage unit 1. The input sentence / correct answer sentence creation unit 2 outputs an input sentence f, which is text including the person attribute information b and environmental information c included in this behavioral data a, and further including a question sentence inquiring about behavioral information d, and a correct answer sentence g, which is text indicating the person attribute information b included in the received behavioral data a, and transmits the input sentence f to the answer sentence generation unit 3, and transmits the correct answer sentence g to the data extension determination unit 5.

[0046] In step S301-2, the input sentence / correct answer sentence creation unit 2 receives behavioral data a including behavioral information d and environmental information c from the storage unit 1. The input sentence / correct answer sentence creation unit 2 outputs an input sentence f, which is text including the behavioral information d and environmental information c included in this behavioral data a and further including a question sentence inquiring about person attribute information b, and a correct answer sentence g, which is text indicating the behavioral information d included in the received behavioral data a, and transmits the input sentence f to the answer sentence generation unit 3 and transmits the correct answer sentence g to the data extension determination unit 5.

[0047] In step S302, as a first process, the answer sentence generation unit 3 accepts the input sentence f transmitted from the input sentence / correct answer sentence creation unit 2 in step S301-1 above, accepts the sentence generation model x from the memory unit 1, inputs the input sentence f to the sentence generation model x, outputs a generated answer sentence h output from this sentence generation model x as the result of generating an answer sentence to the question sentence in the input sentence f, and transmits this generated answer sentence h to the input sentence / correct answer sentence creation unit 2 as “generated answer sentence h (S302).”

[0048] Also, in this step S302, as a second process, the answer sentence generation unit 3 accepts the input sentence f transmitted from the input sentence / correct answer sentence creation unit 2 in the above step S301-2, accepts the sentence generation model x from the memory unit 1, inputs the input sentence f to the sentence generation model x, outputs the generated answer sentence h output from this sentence generation model x as the generated answer sentence to the question sentence in the input sentence f, and transmits this generated answer sentence h to the input sentence / correct answer sentence creation unit 2 as “generated answer sentence h (S302).”

[0049] In step S303-1, the input sentence and correct answer sentence creation unit 2 receives from the storage unit 1 the same behavioral data a as received in step S301-1 above, and receives the "generated answer sentence h (S302)" transmitted from the answer sentence generation unit 3 by the first process of step S302 above. The input sentence and correct answer sentence creation unit 2 outputs an input sentence f, which is text that includes the person attribute information b contained in this "generated answer sentence h (S302)" and the environmental information c contained in the received behavioral data a, and further includes a question sentence asking about behavioral information d, and transmits this input sentence f to the answer sentence generation unit 3.

[0050] In step S303-2, the input sentence and correct answer sentence creation unit 2 receives from the storage unit 1 the same behavioral data a as received in step S301-2 above, and receives the "generated answer sentence h (S302)" transmitted from the answer sentence generation unit 3 by the second process of step S302 above. The input sentence and correct answer sentence creation unit 2 outputs an input sentence f, which is text that includes behavioral information d included in this "generated answer sentence h (S302)" and environmental information c included in the received behavioral data a, and further includes a question sentence inquiring about person attribute information b, and transmits this input sentence f to the answer sentence generation unit 3.

[0051] In step S304, as a first process, the answer sentence generation unit 3 accepts the input sentence f transmitted from the input sentence / correct answer sentence creation unit 2 in step S303-1 above, accepts the sentence generation model x from the memory unit 1, inputs the input sentence f into the sentence generation model x, outputs the generated answer sentence h output from this sentence generation model x as the result of generating an answer sentence to the question sentence in the input sentence f, and transmits this generated answer sentence h to the data extension determination unit 5 as ``generated answer sentence h (S304).''

[0052] Also, in step S304, as a second process, the answer sentence generation unit 3 receives the input sentence f transmitted from the input sentence / correct answer sentence creation unit 2 in step S303-2 above, receives the sentence generation model x from the storage unit 1, inputs the input sentence f to the sentence generation model x, outputs a generated answer sentence h output from this sentence generation model x as the generated answer sentence to the question sentence in the input sentence f, and transmits this generated answer sentence h as "generated answer sentence h (S304)" to the data extension determination unit 5. The above steps S302 to S304 are repeated N times, which is 1 or more.

[0053] When the above steps S302 to S304 have been repeated N times (Yes in S305), in step S306, as a first process, the data extension determination unit 5 accepts from the memory unit 1 two of the three pieces of information, namely, person attribute information b, environmental information c, and behavioral information d, in this case behavioral data a consisting of person attribute information b and environmental information c, accepts N ``generated answer sentences h (S304)'' transmitted from the answer sentence generation unit 3 by the first process of step S304, and accepts from the input sentence / correct answer sentence creation unit 2 the correct answer sentence g corresponding to this generated answer sentence, i.e., the correct answer sentence g transmitted from the input sentence / correct answer sentence creation unit 2 by the above step S301-1.

[0054] In the first process of step S306, the data extension determination unit 5 calculates a score indicating the degree of word agreement between each of the N ``generated answer sentences h (S304)'' and the correct answer sentence g, and selects the top n ``generated answer sentences h (S304)'' with the calculated scores.

[0055] In the first process of step S306, the data extension determination unit 5 determines, as pseudo-correct data, the "generated answer sentence h (S302)" corresponding to the selected "generated answer sentence h (S304)," i.e., the "generated answer sentence h (S302)" used to create the input sentence f that was used to generate the selected "generated answer sentence h (S304)" among the "generated answer sentences h (S302)" generated by the first process of step S302. The data extension determination unit 5 interpolates the received behavioral data a using information contained in the determined "generated answer sentence h (S302)," thereby outputting behavioral data a consisting of three pieces of information: person attribute information b, environmental information c, and behavioral information d, and transmitting this behavioral data a to the storage unit 1.

[0056] Also, in step S306, as a second process, the data extension determination unit 5 accepts from the memory unit 1 two of the three pieces of information, namely, person attribute information b, environmental information c, and behavioral information d, in this case behavioral data a consisting of behavioral information d and environmental information c, accepts N ``generated answer sentences h (S304)'' transmitted from the answer sentence generation unit 3 by the second process of step S304 above, and accepts from the input sentence / correct answer sentence creation unit 2 the correct answer sentence g corresponding to this generated answer sentence, i.e., the correct answer sentence g transmitted from the input sentence / correct answer sentence creation unit 2 by step S301-2 above.

[0057] In the second process of step S306, the data extension determination unit 5 calculates a score indicating the degree of word agreement between each of the N ``generated answer sentences h (S304)'' and the correct answer sentence g, and selects the top n ``generated answer sentences h (S304)'' with the calculated scores.

[0058] In the second processing of step S306, the data extension determination unit 5 determines, as pseudo-correct data, the "generated answer sentence h (S302)" corresponding to the selected "generated answer sentence h (S304)," i.e., the "generated answer sentence h (S302)" used to create the input sentence f that was used to generate the selected "generated answer sentence h (S304)" among the "generated answer sentences h (S302)" generated by the second processing of step S302. The data extension determination unit 5 interpolates the received behavioral data a using the information contained in the determined "generated answer sentence h (S302)," thereby outputting behavioral data a consisting of three pieces of information: person attribute information b, environmental information c, and behavioral information d, and transmits this behavioral data a to the storage unit 1.

[0059] According to the behavior simulation device of one embodiment of the present invention described above, it is possible to obtain the effect of generating a behavior sequence that is characteristic of the individual and that has a natural context, taking into account the individual's attributes.

[0060] Fig. 8 is a block diagram showing an example of the hardware configuration of a behavior simulation device according to an embodiment of the present invention. In the example shown in Fig. 8, the behavior simulation device 100 according to the embodiment is configured, for example, as a server computer or a personal computer, and has a hardware processor 111A such as a CPU. A program memory 111B, a data memory 112, an input / output interface 113, and a communication interface 114 are connected to this hardware processor 111A via a bus 115.

[0061] The communication interface 114 includes, for example, one or more wireless communication interface units, and enables transmission and reception of information to and from a communication network. As the wireless interface, for example, an interface that adopts a low-power wireless data communication standard such as a wireless LAN (Local Area Network) is used.

[0062] An input device 300 and an output device 400, which are attached to the behavior simulation device 100 and used by a user or the like, are connected to the input / output interface 113. The input / output interface 113 takes in operation data input by a user or the like through the input device 300, such as a keyboard, touch panel, touchpad, or mouse, and outputs and displays output data to an output device 400, which may include a display device using liquid crystal or organic electroluminescence (EL) or an audio output device. The input device 300 and the output device 400 may be devices built into the behavior simulation device 100, or may be input devices and output devices of other information terminals that can communicate with the behavior simulation device 100 via a network.

[0063] The program memory 111B is a non-transitory tangible storage medium that is a combination of a non-volatile memory that can be written to and read from at any time, such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive), and a non-volatile memory such as a ROM, and stores programs necessary to execute various control processes, etc., according to one embodiment.

[0064] The data memory 112 is a tangible storage medium, for example, a combination of the above-mentioned nonvolatile memory and a volatile memory such as RAM, and is used to store various data acquired and created during various processing steps.

[0065] A behavior simulation device 100 according to one embodiment of the present invention can be configured as a software-based processing function unit, i.e., a data processing system or information processing device having the units shown in Fig. 1. A storage system used as a work memory or the like by the behavior simulation device 100 can be configured by using the data memory 112 shown in Fig. 8. However, these configured storage areas are not essential components within the behavior simulation device 100, and may be, for example, areas provided in an external storage medium such as a USB (Universal Serial Bus) memory, or a storage system such as a database server located in the cloud.

[0066] The processing function unit can be realized by having the hardware processor 111A read and execute a program stored in the program memory 111B, but the processing function unit may also be realized in various other forms, including an integrated circuit such as an application specific integrated circuit (ASIC) or a field-programmable gate array (FPGA).

[0067] Furthermore, the techniques described in the above embodiments can be stored as a program (software means) that can be executed by a computer on a recording medium such as a magnetic disk (floppy disk, hard disk, etc.), optical disk (CD-ROM, DVD, MO, etc.), or semiconductor memory (ROM, RAM, flash memory, etc.), and can also be distributed by transmitting it via a communication medium. The program stored on the medium also includes a configuration program that configures the software means (including not only execution programs but also tables and data structures) that the computer executes. The computer that realizes this device reads the program stored on the recording medium and, in some cases, configures the software means using the configuration program, and executes the above-mentioned processing by controlling the operation of this software means. Note that the term "recording medium" as used herein is not limited to a storage medium for distribution, but also includes storage media such as a magnetic disk or semiconductor memory installed inside the computer or in a device connected via a network.

[0068] The present invention is not limited to the above-described embodiments, and various modifications can be made in the implementation stage without departing from the spirit of the invention. Furthermore, the embodiments may be implemented in appropriate combinations, in which case the combined effects can be obtained. Furthermore, the above-described embodiments include various inventions, and various inventions can be extracted by combining selected elements from the disclosed elements. For example, if the problem can be solved and the desired effect can be obtained even if some elements are deleted from all elements shown in the embodiments, the configuration from which these elements are deleted can be extracted as an invention.

[0069] 100... Behavior simulation device 1... Storage unit 2... Input sentence / correct answer sentence creation unit 3... Answer sentence generation unit 4... Parameter update unit 5... Data extension determination unit 6... Output unit

Claims

1. An information processing device comprising: an input unit that accepts input of behavioral data including environmental information that indicates the environment in which a person behaves, and further including one of person attribute information that indicates the attributes of the person and behavioral information that indicates the behavior of the person; and a generation unit that generates information that includes the environmental information included in the behavioral data, further including one of the person attribute information and the behavioral information included in the behavioral data, and further including a question that inquires about the other of the person attribute information and the behavioral information included in the behavioral data, inputs the generated information into a model, and generates information that includes an answer to the information asked in the question based on the result of this input.

2. The information processing device of claim 1, wherein the input unit accepts input of behavioral data including the environmental information, the behavioral information, and the person attribute information; the generation unit generates correct answer information for the information asked in the question; and the information generated by the generation unit and including the answer to the information asked in the question is compared with the correct answer information to update the parameters of the model.

3. The input unit accepts input of behavioral data including the environmental information and further including one of the behavioral information and the person attribute information; the generation unit generates first input information including the environmental information included in the behavioral data, further including one of the person attribute information and the behavioral information included in the behavioral data, and further including a question inquiring about the other of the person attribute information and the behavioral information included in the behavioral data, and correct answer information indicating the one of the person attribute information and the behavioral information included in the behavioral data; inputs the first input information to a model and generates first answer information including an answer to information asked in the question based on a result of this input; generates second input information including the first answer information and further including a question inquiring about the one of the person attribute information and the behavioral information included in the behavioral data; inputs the second input information to a model and generates second answer information including an answer to information asked in the question included in the second input information based on a result of this input; 2. The information processing device according to claim 1, further comprising a selection unit that calculates a value indicating a degree of agreement between the content of the second answer information and the content of the correct answer information, and selects, based on the calculated value, pseudo-answer information from the first answer information that is a pseudo-answer to information asked in a question included in the first input information that was used to generate the second input information used to generate the generated second answer information.

4. A method performed by an information processing device, comprising: an input unit of the information processing device accepting input of behavioral data including environmental information indicating the environment in which a person behaves, and further including one of person attribute information indicating the attributes of the person and behavioral information indicating the behavior of the person; a generation unit of the information processing device generating information including the environmental information included in the behavioral data, further including one of the person attribute information and the behavioral information included in the behavioral data, and further including a question inquiring about the other of the person attribute information and the behavioral information included in the behavioral data, inputting the generated information into a model, and generating information including an answer to the information asked in the question based on the result of this input.

Images