Information processing device, information processing method, and information processing program

Abstract

We create machine learning models that are specialized for each user. [Solution] In the server device, an acquisition unit acquires user data related to a user, and an estimation unit estimates, based on the user data history, a first situation, which is the state the user was in when the user data included in the history was acquired from the user. Based on the user data, a detection unit detects, before and after the user exhibits the first situation, a change in emotion, and if a change in emotion is detected, a determination unit determines whether the first situation includes a difference that deviates from a general situation based on a comparison between information related to the first situation and information indicating an inference result output by a predetermined generative model in response to an input based on the first situation. If the first situation includes a difference that deviates from a general situation, a management unit stores difference information indicating the difference as personal feature data representing the user's personal information, and manages the personal feature data so that it is used as training data for a machine learning model corresponding to the user.

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and an information processing program. [Background technology]

[0002] A method for adjusting the parameters of a large-scale language model by re-training has been proposed so that the model can be used to support users' work. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 7649839 Summary of the Invention [Means for solving the problem]

[0004] The information processing device according to the present application comprises an acquisition unit that acquires user data related to a user; an estimation unit that estimates, based on a history of the user data, a first situation that is the state of the user at the time the user data included in the history was acquired from the user; a detection unit that detects, based on the user data, an emotional change before and after the user exhibits the first situation; a determination unit that, when the emotional change is detected, determines whether the first situation includes a difference that deviates from a general situation based on a comparison between information related to the first situation and information indicating an inference result output by a predetermined generative model in response to an input based on the first situation; and a management unit that, when the first situation includes the difference that deviates from a general situation, stores difference information indicating the difference as personal feature data representing personal information of the user, and thereby manages the personal feature data so that it is used as training data for a machine learning model corresponding to the user. [Brief explanation of the drawings]

[0005] [Figure 1]Figure 1 shows an overview of the proposed technology (1). [Figure 2] Figure 2 shows an overview of the proposed technology (2). [Figure 3] FIG. 3 is a diagram illustrating an example of the configuration of an information processing system according to the embodiment. [Figure 4] FIG. 4 is a diagram illustrating an example of the configuration of a server device according to the embodiment. [Figure 5] FIG. 5 is a flowchart showing an information processing procedure related to estimation of user characteristics. [Figure 6] FIG. 6 is a flowchart (1) showing an information processing procedure related to feature extraction. [Figure 7] FIG. 7 is a flowchart (2) showing the information processing procedure related to feature extraction. [Figure 8] FIG. 8 is a diagram illustrating a specific example of information processing according to the embodiment. [Figure 9] FIG. 9 is a hardware configuration diagram illustrating an example of a computer that realizes the functions of the server device. DETAILED DESCRIPTION OF THE INVENTION

[0006] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. In this specification and drawings, components having substantially the same functional configurations are designated by the same reference numerals, and redundant description will be omitted.

[0007] One or more embodiments (including examples, modifications, and application examples) described below can be implemented independently. However, at least a portion of the embodiments described below may be implemented in appropriate combination with at least a portion of another embodiment. These embodiments may include novel features that are different from each other. Therefore, these embodiments may contribute to solving different purposes or problems and may produce different effects from each other.

[0008] <Embodiment> 1. Introduction Improving productivity through generative AI (artificial intelligence) has become a global concern, and it is expected that efficiency will be further improved through the use of generative AI.

[0009] It is also expected that generative AI will be incorporated into our daily lives at an accelerating pace within the next few years. For example, companies will be able to use generative AI to conduct business more efficiently, and generative AI will also bring various conveniences to personal use.

[0010] In the evolution of generative AI, it is important to provide optimal solutions suited to each user (general solutions that can be applied to an unspecified number of users), but the pursuit of overly optimal solutions can hinder new discoveries and inventions. For this reason, we should consider a model for how to include elements that may at first glance be deemed unnecessary in the inference results of generative AI.

[0011] [2. Purpose and Solution] People can gain various insights by doing useless things. However, if generative AI continues to produce optimal solutions, people will lose opportunities to come into contact with unnecessary contacts, i.e., useless things, which may hinder new discoveries and inventions. Therefore, the inventors of this invention focused on a new method for generating machine learning models that can intentionally generate unnecessary contacts.

[0012] Specifically, the proposed technology of the present invention (hereinafter abbreviated as "proposed technology") extracts noise components (eigensolutions that are not general and may be judged as useless) that contradict general solution candidates in inference from user data about the user, and uses them as training data to be learned as the user's individuality. In other words, the proposed technology is characterized by capturing unnecessary information that is likely to be excluded from optimal solution candidates as the user's feature quantity.

[0013] In a world that has become overly optimized due to the advances in generative AI, this proposed technology will enable the provision of personalized machine learning models that understand "necessary waste" better than anyone else and can make suggestions based on that understanding. It could also be a tool for deriving new solutions that would be impossible to find if the route were too short, for example, in business.

[0014] [3. Overview of proposed technology] An overview of the proposed technology will be explained using Figures 1 and 2. The proposed technology will be explained by dividing it into a phase of extracting feature data that will serve as training data and a phase of utilizing a personalized model generated using the feature data. Furthermore, information processing will be explained for one user U1 out of the users U, which refers to various users, but similar information processing may also be performed for each of the other users U.

[0015] The following description will be given assuming that the entity that executes the information processing according to the proposed technology is the server device 100.

[0016] (3-1. Feature Extraction) 1 is a diagram (1) showing an overview of the proposed technology. The server device 100 sequentially acquires life data LF_DA1 as user data related to a user U1 from a terminal device 10 (e.g., a wearable device) of a user U, and accumulates the acquired life data LF_DA1 (step S11).

[0017] Since the terminal device 10 is equipped with various sensors, the life data LF_DA1 can be interpreted as sensor data detected by the sensors. Therefore, the life data LF_DA1 may include data related to the movements of the user U1, such as position, speed, acceleration, rotation, and posture.

[0018] The life data LF_DA1 may also include data on the vital signs of the user U1, such as heart rate, electrocardiogram, blood oxygen level, respiratory rate, body temperature, sweat rate, sleep state, and the like.

[0019] The life data LF_DA1 may also include data relating to the surrounding environment of the user U1, such as weather, temperature, humidity, atmospheric pressure, altitude, and the like.

[0020] The life data LF_DA1 may also include data related to the senses of sight, hearing, smell, etc. in clothing, diet, and living environment, as well as data related to the senses of sight, hearing, smell, etc. in places to go for work, daily life, and travel. Furthermore, the life data LF_DA1 may also include schedule information of the user U1.

[0021] Any of the above-mentioned life data LF_DA1 that can be acquired as sensor data may be actively provided by the user U. For example, if the terminal device 10 has a function that enables transmission and reception of information between the terminal device 10 and the server device 100, the user U1 can input the life data LF_DA1 to the server device 100 via the terminal device 10.

[0022] Returning to the explanation, the server device 100 analyzes the history LG1 of the life data LF_DA1 for a predetermined period (e.g., 10 minutes) up to the current time of processing, of the history LG of the life data LF_DA1, to estimate a first situation SU1, which is the situation of the user U1 when the life data LF_DA1 included in the history LG1 was acquired from the user U1 (step S12). For example, the server device 100 may estimate the characteristic behavior (also called action) BH or emotion EM of the user U1 as the first situation SU1. Note that any analytical method may be used to analyze the life data LF_DA1.

[0023] As a result, the server device 100 can obtain first situation data SU1_DA1 related to the first situation SU1, as shown in Fig. 1. The first situation data SU1_DA1 may be data indicating the first situation SU1 itself, or may be a history LG1 used to estimate the first situation SU1.

[0024] Furthermore, the server device 100 may be linked to a general inference model M, and generates input information (also called instruction information or prompt) IN according to the estimated first situation SU1 and inputs it into the general inference model M to obtain inference result data RE_DA1 indicating the inference result by the general inference model M (step S13). The general inference model M may be a generative AI, and specifically may be a large-scale language model (LLM) that performs natural language processing, such as a GPT (Generative Pre-trained Transformer) or a Transformer.

[0025] In this state, the server device 100 compares the first situation data SU1_DA1 with the inference result data RE_DA1, and determines whether or not the first situation data SU1_DA1 contains difference information DF that is different from the inference result data RE_DA1 (step S14).

[0026] The inference result data RE_DA1 is an optimal solution determined to be optimal by the general inference model M, and can be said to be a general solution that can be common to an unspecified number of users U. Therefore, the processing in step S14 corresponds to the processing of determining whether the first situation SU1 includes a difference that deviates from a general situation common to an unspecified number of users U, and, if a difference is included, extracting difference information DF indicating that difference. Furthermore, for this reason, the difference information DF is a noise component that contradicts the general solution candidate in the inference, or unnecessary information that is likely to be excluded from the optimal solution candidate.

[0027] Then, as shown in FIG. 1, the server device 100 stores the difference information DF as personal feature data PL_DA1 representing the personal information of user U1, thereby managing the personal feature data PL_DA1 so that it can be used as learning data LDA of the machine learning model corresponding to user U1 (step S15).

[0028] (3-2. Use of models for individuals) Fig. 2 is a diagram (2) illustrating an overview of the proposed technology. According to the example of Fig. 2, the server device 100 inputs personal feature data PL_DA1 to, for example, a pre-trained model for a generation AI (step S21).

[0029] Then, the server device 100 uses the personal feature data PL_DA1 to train a pre-trained model to learn differences that are different from the general public as the individuality of the user U1, thereby generating a unique inference model M11 (step S22). The unique inference model M11 is a machine learning model for the user U1.

[0030] That is, the unique inference model M11 is a machine learning model that has learned what "necessary waste" is for user U1, while it is likely to be wasteful in general, and makes proposals based on what it has learned, enabling it to make proposals specialized for user U1. Note that, as shown in Figure 2, the server device 100 may also input the life data LF_DA1 and the inference result data RE_DA1 into the pre-trained model as training data LDA.

[0031] Here, for example, suppose that user U1 inputs a request for route guidance to destination G to the unique inference model M11 (step S23). Since the general inference model M often outputs the shortest route to destination G as the optimal solution, user U1 cannot know the circuitous route that is excluded from the optimal solution.

[0032] However, this circuitous route may contain special events that may lead to new discoveries or encounters for user U1, even if the events may seem ordinary to many users U and do not cause any emotional changes or notice. Therefore, the unique inference model M11 may be able to infer the circuitous route excluded by the general inference model M as the optimal route for user U1 and may be able to suggest this circuitous route to user U1 as the optimal route (step S24). Therefore, by using the unique inference model M11, user U1 is more likely to make new discoveries and encounters.

[0033] As described above, the proposed technology is unique in that it can increase the probability of creating opportunities that would otherwise be merely chance encounters. For example, the proposed technology can achieve the unique effect of controlling events that user U1 would not have been able to experience unless he or she happened to choose a roundabout route, so that the probability of experiencing such an event can be increased.

[0034] In the following embodiments, when there is no need to limit the explanation to one user U1, the expression "user U" will be used. Furthermore, when not limiting to one user U1, the life data LF_DA1 will be referred to as "life data LF_DA," the first situation data SU1_DA1 as "first situation data SU1_DA," the inference result data RE_DA1 as "inference result data RE_DA," the personal feature data PL_DA1 as "personal feature data PL_DA," and the unique inference model M11 as "unique inference model M1."

[0035] [4. System Configuration] Fig. 3 is a diagram showing an example of the configuration of an information processing system according to an embodiment. Fig. 3 shows an information processing system 1 as an example of the information processing system according to an embodiment. As shown in Fig. 3, the information processing system 1 may include a terminal device 10, an external device 30, and a server device 100. Note that the information processing system 1 may include one or more terminal devices 10, one or more external devices 30, and one or more server devices 100.

[0036] The terminal device 10 is an example of an information processing terminal used by a user U. The terminal device 10 may be a smartphone, a wearable device, a tablet terminal, a notebook PC (Personal Computer), a desktop PC, a mobile phone, a PDA (Personal Digital Assistant), or the like. For example, an application for transmitting and receiving information to and from the server device 100 may be installed in the terminal device 10. Such an application may be a general-purpose application such as a web browser, or may be a dedicated application newly implemented in accordance with the proposed technology. Furthermore, such an application may have a function for inputting information to the general inference model M or the unique inference model M11, and for acquiring inference results from the general inference model M or the unique inference model M11.

[0037] The external device 30 is an information processing device that controls the learning of the general inference model M and predictions using the general inference model M. The functions of the external device 30 may be incorporated into the server device 100.

[0038] The server device 100, which is an example of an information processing device, is a central device that performs information processing according to the embodiment, and may be implemented as a cloud server device. It should be noted that a user U may use multiple terminal devices 10 of different types, such as a wearable device, a smartphone, and a laptop PC. It is assumed that a dedicated application is installed on each terminal device 10. In such a case, the life data LF_DA detected by the wearable device may be temporarily stored in the smartphone, which has a larger storage capacity. Furthermore, the life data LF_DA stored in the smartphone may be preprocessed by the laptop PC, which has higher-performance computing resources, before being sent to the server device 100.

[0039] 5. Server Device Configuration The server device 100 according to the embodiment will be described with reference to Fig. 4. Fig. 4 is a diagram showing an example of the configuration of the server device 100 according to the embodiment. As shown in Fig. 4, the server device 100 includes a communication unit 110, a storage unit 120, and a control unit 130.

[0040] (Communication unit 110) The communication unit 110 is realized by, for example, a network interface card (NIC), etc. For example, the communication unit 110 transmits and receives information to and from the terminal device 10 and the external device 30.

[0041] (Storage unit 120) The storage unit 120 is realized by, for example, a semiconductor memory element such as a random access memory (RAM) or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 120 may store, for example, data and programs related to the information processing according to the embodiment. As shown in FIG. 4 , the storage unit 120 may include a life data storage unit 121, a situation data storage unit 122, an inference result data storage unit 123, and a feature data storage unit 124.

[0042] (control unit 130) The control unit 130 is realized by a processor such as a CPU (Central Processing Unit) or an MPU (Micro Processing Unit) executing various programs (for example, the information processing program according to the embodiment) stored in a storage device inside the server device 100 using RAM as a working area. The control unit 130 is also realized by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

[0043] As shown in Fig. 4, the control unit 130 has an acquisition unit 131, an estimation unit 132, a determination unit 133, a detection unit 134, a discrimination unit 135, a management unit 136, and a learning unit 137, and realizes or executes the functions and actions of the information processing described below. The internal configuration of the control unit 130 is not limited to the configuration shown in Fig. 4, and may be other configurations as long as they perform the information processing described below. Furthermore, the connection relationship between the processing units of the control unit 130 is not limited to the connection relationship shown in Fig. 4, and may be other connection relationships. The following information processing, which will be described as being performed by the control unit 130, may be performed for each user U using information about each user U.

[0044] (Acquisition part 131) The acquisition unit 131 acquires various information related to requests, inputs, requests, etc. For example, the acquisition unit 131 acquires various information related to requests, inputs, requests, etc. from the terminal device 10 or the external device 30. Furthermore, the acquisition unit 131 may store the acquired information in the storage unit 120 and update the stored contents.

[0045] For example, the acquiring unit 131 sequentially acquires life data LF_DA as user data related to the user U from the terminal device 10 (for example, a wearable device), and stores the acquired life data LF_DA separately for each user U. The life data LF_DA may be stored in the life data storage unit 121.

[0046] (Estimation part 132) The estimation unit 132 analyzes the history LG of the life data LF_DA for a predetermined period (e.g., 10 minutes) up to the current time point for processing, of the history LG, to estimate a first situation SU1 that is the situation of the user U when the life data LF_DA included in the history LG was acquired from the user U. For example, the estimation unit 132 may estimate, as the first situation SU1, a characteristic behavior BH (first behavior) of the user U1 when the life data LF_DA included in the history LG was acquired from the user U.

[0047] Furthermore, the estimation unit 132 may estimate behavior patterns PT21 and emotion patterns PT22 in various situations as the second situation SU2 based on the life data LF_DA. For example, the estimation unit 132 may estimate various behavior patterns PT21 (second behaviors) based on the life data LF_DA, such as "eat at XX time on weekdays," "take XX route when commuting," "take XX route when walking," and "practice golf at XX time on weekends," and may also estimate emotion patterns PT22 for each behavior pattern PT21.

[0048] (Judgment unit 133) When a first situation SU1 is estimated when the life data LF_DA included in the history LG is acquired from the user U, the determination unit 133 determines whether the estimated first situation SU1 is different from a second situation SU2. For example, the determination unit 133 may determine whether the user U has exhibited a characteristic behavior BH (first behavior) that is different from the behavior pattern PT21 (second behavior). In other words, the determination unit 133 determines whether the user U has exhibited a situation that is different from their usual situation.

[0049] (Detection unit 134) The detection unit 134 detects a change in emotion before and after the user U exhibits a first situation SU1 based on the life data LF_DA. For example, when the user U exhibits a first situation SU1 that is different from the second situation SU2, the detection unit 134 may detect a change in emotion before and after the user U exhibits the first situation SU1. For example, when the user U exhibits a characteristic behavior BH (first behavior) that is different from the behavior pattern PT21 (second behavior), the detection unit 134 may detect a change in emotion before and after the user U exhibits the behavior BH based on the emotion EM when the user U exhibits the behavior BH.

[0050] (Discrimination unit 135) When an emotion change is detected, the determination unit 135 compares the first situation data SU1_DA with the inference result data RE_DA and determines whether or not the first situation data SU1_DA contains difference information DF that differs from the inference result data RE_DA. That is, the determination unit 135 determines whether or not the first situation SU1 includes a situation that is different from a general situation common to an unspecified number of users U.

[0051] The first situation data SU1_DA may be data indicating the first situation SU1 itself, or may be the history LG used to estimate the first situation SU1.

[0052] (Management Department 136) If the first situation SU1 includes a difference that deviates from a general situation, the management unit 136 extracts difference information DF indicating the difference. Then, the management unit 136 stores the difference information DF as personal feature data PL_DA representing personal information of the user U, thereby managing the personal feature data PL_DA so that it is used as learning data LDA of the machine learning model corresponding to the user U.

[0053] (Learning Section 137) The learning unit 137 inputs the personal feature data PL_DA into a pre-trained model for the generation AI and causes the pre-trained model to learn the difference information DF as the personality of the user U, thereby generating a unique inference model M1 specialized for the individual user U. In other words, the learning unit 137 may generate a unique inference model M1 for each user U.

[0054] [6. Operation Procedure] The operation procedure of the server device 100 will be described with reference to Figures 5 to 7. Figures 5 to 7 show scenes in which information processing according to the embodiment is performed for one user U1.

[0055] (6-1. Estimation of user characteristics) First, the operation procedure for estimating the characteristics of user U1 will be described. Fig. 5 is a flowchart showing the information processing procedure for estimating the user characteristics.

[0056] For example, the acquisition unit 131 sequentially acquires the life data LF_DA1 of the user U from the terminal device 10, and stores the acquired life data LF_DA1 in the life data storage unit 121 (step S501).

[0057] Based on the life data LF_DA1, the estimation unit 132 estimates behavior patterns PT21 in various situations and emotion patterns PT22, which are the emotions of user U1 for each behavior pattern PT21, as a second situation SU2 (a situation as a characteristic of user U1) (step S502).

[0058] Then, the estimation unit 132 stores second situation data SU2_DA1 indicating the second situation SU2 in the situation data storage unit 122 (step S503).

[0059] The estimation unit 132 may periodically estimate the user characteristics. For example, the estimation unit 132 may estimate the second situation SU2 every month using the life data LF_DA1 for that month, thereby updating the estimation result for the previous month with the latest estimation result.

[0060] (6-2. Feature Data Extraction (1)) Next, an operational procedure for extracting personal feature data PL_DA1 of user U1 will be described. Fig. 6 is a flowchart (1) showing an information processing procedure relating to feature extraction.

[0061] The acquisition unit 131 continues to acquire the life data LF_DA1 from the terminal device 10. In this state, the estimation unit 132 determines whether or not it is time to process feature extraction (step S601). For example, the estimation unit 132 may periodically (for example, every five minutes) determine whether or not it is time to process. While it is not time to process (step S601; No), the acquisition unit 131 waits until it is time to process.

[0062] When the processing timing arrives (step S601; Yes), the estimation unit 132 extracts the history LG1 of the life data LF_DA1 for a predetermined period (e.g., 5 minutes) up to the present time of the processing timing from the history LG of the life data LF_DA1 (step S602).

[0063] Then, the estimation unit 132 analyzes the life data LF_DA1 included in the history LG1 to estimate a first situation SU1, which is the situation of the user U1 when the life data LF_DA1 was acquired from the user U1 (the terminal device 10 of the user U1) (step S603). For example, the estimation unit 132 estimates, as the first situation SU1, a characteristic behavior BH of the user U1 and an emotion EM when the user U1 exhibited the behavior BH.

[0064] In response to the estimation of the first situation SU1, the determination unit 133 determines whether the estimated first situation SU1 is different from the second situation SU2 (step S604). Specifically, the determination unit 133 determines whether the user U1 has exhibited a first situation SU1 that is different from the second situation SU2. For example, the determination unit 133 may determine whether the first situation SU1 is different from the second situation SU2 that shares a common scene with the first situation SU1 (for example, a walking scene) among the second situations SU2 that have been estimated as the user characteristics. For example, the determination unit 133 may determine whether the user U1 has exhibited a characteristic behavior BH (first behavior) that is different from the behavior pattern PT21 (second behavior).

[0065] If the first situation SU1 is similar to or coincides with the second situation SU2 (step S604; No), the determination unit 133 may return the process to step S601.

[0066] On the other hand, if the user U1 indicates a first situation SU1 that is different from the second situation SU2 (step S604; Yes), the detection unit 134 detects a change in emotion before and after the user U1 indicates the first situation SU1 (step S605).

[0067] Based on the emotion change detection result, the determination unit 133 determines whether an emotion change has occurred in the user U1 before and after the user U1 exhibits the first situation SU1 (step S606).

[0068] If no emotional change has occurred in the user U1 (step S606; No), the determination unit 133 may return the process to step S601.

[0069] On the other hand, if an emotional change occurs in user U1 (step S606; Yes), the judgment unit 133 acquires first situation data SU1_DA1 indicating the first situation SU1 estimated in step S603 (step S607), and stores the acquired first situation data SU1_DA1 in the situation data storage unit 122 (step S608).

[0070] (6-3. Feature Data Extraction (2)) Next, the operation procedure for extracting the personal feature data PL_DA1 of the user U1 will be explained. Fig. 7 is a flowchart (2) showing the information processing procedure for feature extraction.

[0071] The discrimination unit 135 generates input information IN according to the first situation SU1 estimated in step S603 and inputs it to the general inference model M (step S701). The discrimination unit 135 acquires inference result data RE_DA1 indicating the inference result by the general inference model M and stores it in the inference result data memory unit 123 (step S702).

[0072] The determination unit 135 compares the first situation data SU1_DA1 with the inference result data RE_DA1, and determines whether the first situation SU1 includes a difference that deviates from a general situation (step S703). Specifically, the determination unit 135 compares the first situation data SU1_DA1 with the inference result data RE_DA1, and determines whether the first situation data SU1_DA1 contains difference information DF that differs from the inference result data RE_DA1.

[0073] If the first situation SU1 does not include a difference that deviates from the general situation (step S703; No), the determination unit 135 may return the process to step S601.

[0074] On the other hand, if the first situation SU1 includes a difference that deviates from the general situation (step S703; Yes), the management unit 136 extracts difference information DF that indicates the difference (step S704).

[0075] Then, the management unit 136 stores the difference information DF as personal feature data PL_DA1 in the feature data storage unit 124, thereby managing the personal feature data PL_DA1 so that it can be used in the learning data LDA of the unique inference model M11 (step S705).

[0076] 7, the learning unit 137 generates a unique inference model M11 for user U1 based on a pre-trained model for the generation AI and the personal feature data PL_DA1. For example, the learning unit 137 may input the personal feature data PL_DA1 to the pre-trained model, thereby causing the pre-trained model to learn the difference information DF as the personality of user U1.

[0077] In step S703, if the first situation SU1 does not include a difference that deviates from a general situation, the discrimination unit 135 may output information that the first situation SU1 does not include a difference to the management unit 136. The management unit 136 may also manage the information that the first situation SU1 does not include a difference so that it is used as learning data LDA for the unique inference model M11. As a result, the unique inference model M11 can be trained to clearly distinguish which situations are individual to the user U1 and which situations are general situations that are not individual to the user U1, thereby improving the accuracy of the unique inference model M11.

[0078] [7. Specific Examples] The information processing explained using Fig. 5 to Fig. 7 will be described in more detail. Fig. 8 is a diagram showing a specific example of the information processing according to the embodiment. In Fig. 8, the specific example will be described using a scene of the behavior of user U1.

[0079] For example, it is assumed that the estimation unit 132 estimates in step S502 a behavior pattern PT21 (second behavior) that user U1 has a characteristic of "taking route A when walking." Furthermore, it is assumed that in step S603, the estimation unit 132 estimates, as the first situation SU1, that "took route B on this walk" as the characteristic behavior BH of user U1, and estimates "happy" as the emotion EM when user U1 exhibits behavior BH.

[0080] In this example, in step S604, the determination unit 133 compares the second situation SU2 "When taking a walk, I take route A" with the first situation SU1 "I took route B on this walk," and thereby can determine that the first situation SU1 is different from the second situation SU2. That is, the determination unit 133 can determine that the user U1 "took route B, which is different from usual, on this walk."

[0081] Furthermore, if user U1 was emotionless until "taking route B," the emotion changes to "happy" when "taking route B," and accordingly, detection unit 134 can detect a change to positive emotion in step S605.

[0082] In response to an emotional change (change to a positive emotion) occurring in user U1, the judgment unit 133 acquires first situation data SU1_DA1 indicating behavior BH, "On this walk, I took route B, which is different from usual," and emotion EM, "I'm happy," as shown in FIG. 8.

[0083] Furthermore, according to the above example, in step S701, the discrimination unit 135 may generate, as input information IN according to the first situation SU1, "What emotions do people generally feel when taking an unfamiliar route?"

[0084] Figure 8 shows an example in which the general inference model M infers that the input information IN is a negative emotion such as "impatience," "anxiety," "irritation," or "confusion," and outputs inference result data RE_DA1. In this example, in step S703, the discrimination unit 135 determines that the first situation SU1 includes a difference that deviates from a general situation because the "negative emotion" inferred by the general inference model M differs from the "positive emotion" expressed by the user U1. For example, the discrimination unit 135 can determine that the emotion EM of "happiness" is a difference.

[0085] Therefore, in step S704, the management unit 136 may extract difference information DF such as "during this walk, I took route B, which was different from usual, and this caused my feelings to change to happiness."

[0086] In the above example, the general inference model M determines that people generally experience "negative emotions" when they take a wrong turn or when their usual route is unavailable for some reason and they have to take a different route than usual. However, user U1's emotions changed to "positive emotions" despite the fact that he "took route B, which is different from usual."

[0087] From this, the unique inference model M11 may be able to learn that "user U1 may feel positive emotions even when taking an unconventional route, so when providing directions, it is acceptable to suggest a detour route that is not the shortest." For example, if user U1 inputs a request for route guidance to destination G to the unique inference model M11, it is highly likely that a detour route will be suggested rather than the shortest route that the general inference model M could output. As a result, user U may be able to make new discoveries and encounters by taking a detour route.

[0088] Therefore, with the proposed technology, by accumulating individual features, it becomes possible to perform calculations to obtain a unique solution that is better than the optimal solution for an individual from data that would otherwise be a noise component when obtaining an optimal solution (general solution), thereby increasing the probability of creating opportunities that would otherwise be merely chance encounters.

[0089] 8, the reason for the change to "positive emotion" despite the fact that "the user took route B, which is different from usual," is that "the user found a flower on route B." However, it is difficult for the terminal device 10 to infer the cause based solely on the fact that "the user took route B, which is different from usual," and the server device 100 is in a state where it lacks information indicating the cause.

[0090] However, for example, the server device 100 can proactively infer the cause of the difference, triggered by the extraction of the difference information DF. For example, the server device 100 can infer the cause of the change to "positive emotion" despite "taking route B that is different from usual" based on information about the surrounding environment of route B (e.g., map information, event information, weather information). The server device 100 may also employ cause data CA_DA1 indicating the inferred cause as the difference information DF. As a result, the unique inference model M11 may learn that "since user U1 may feel positive emotion even when taking an unconventional route, when providing directions, it is acceptable to preferentially suggest routes with flowers in bloom, even if they are longer."

[0091] [8. Feature Candidates] Here, an example of what can be extracted as difference information DF and become personal feature amount data PL_DA is shown. In Fig. 8, the difference information DF is exemplified as "During this walk, I took route B, which is different from usual, and this caused my emotions to change to happiness," but the following information can also be extracted as difference information DF:

[0092] for example, Cleaning and organizing your desk when you're really busy Discover breakthroughs in everyday activities like walking and bathing For some reason, I like it when it's too hot (or too cold) for most people. Taking paid leave without any reason or purpose · I feel happy just watching the passing clouds Carrying unnecessary spares (e.g. bandages, mobile batteries, etc.) I would like to recommend the more obscure local characters. · Put the remote control somewhere you can't reach while sitting · Searching by accidentally tapping on a different candidate in predictive text It should be noted that the difference information DF is not limited to these examples.

[0093] [9. Modifications] The server device 100 can be implemented in various modes other than the above-described embodiment, so other embodiments of the server device 100 will be described below.

[0094] (9-1. Weighting) As described above, the proposed technology can increase the probability of creating opportunities that would otherwise be merely accidental encounters. Therefore, in order to create opportunities for accidental encounters, it is important to determine whether the user U accidentally exhibited the first situation SU1 or intentionally exhibited the first situation SU1 and assign a weight value according to the determination result to the personal feature data PL_DA.

[0095] Therefore, the determination unit 133 may determine whether the user U accidentally exhibited the behavior BH (first behavior) or intentionally exhibited the behavior BH (first behavior) based on the environmental data EN_DA corresponding to the history LG (history of user data) of the life data LF_DA. Furthermore, the management unit 136 may assign a weight value WV to the personal feature data PL_DA according to whether the user U accidentally exhibited the behavior BH or intentionally exhibited the behavior BH. This point will be described more specifically using the example of FIG. 7.

[0096] If the first situation SU1 includes a difference that deviates from the general situation (step S703; Yes), the determination unit 133 determines whether the user U1 accidentally exhibited the behavior BH or whether the user U1 intentionally exhibited the behavior BH based on the environmental data EN_DA1 corresponding to the history LG1 of the life data LF_DA1. The environmental data EN_DA1 may be data (e.g., event information or weather information) about the surrounding environment of the user U1 when the life data LF_DA1 included in the history LG1 was acquired from the user U1. The event information may also include information about events that the user U1 would not know unless they were actually at the event, such as construction information, traffic information, and entertainment information.

[0097] Furthermore, in step S704, the management unit 136 extracts difference information DF indicating the difference in response to the fact that the first situation SU1 includes a difference that deviates from a general situation. Then, in step S704, the management unit 136 stores the difference information DF as personal feature data PL_DA1 in the feature data storage unit 124, thereby managing the personal feature data PL_DA1 so that it is used in the learning data LDA of the unique inference model M11.

[0098] In this case, if the user U1 accidentally exhibits the behavior BH, the management unit 136 may assign a weight value WV11 corresponding to the fact that the user U1 accidentally exhibited the behavior BH to the personal feature data PL_DA1. Furthermore, if the user U1 intentionally exhibits the behavior BH, the management unit 136 may assign a weight value WV12 corresponding to the fact that the user U1 intentionally exhibited the behavior BH to the personal feature data PL_DA1.

[0099] As an example, when user U1 accidentally exhibits behavior BH, the management unit 136 may assign a weight WV11 to the personal feature data PL_DA1, which is higher than the weight WV12 when user U1 intentionally exhibits behavior BH. For example, an example of a case where user U1 accidentally exhibits behavior BH is a situation PT1 in which "since construction was occurring on my usual route A, I happened to take route B on this walk," and then "by chance found a flower on route B," which made me feel "happy." Another example of a case where user U1 intentionally exhibits behavior BH is a situation PT2 in which "because I wanted to see flowers, I intentionally chose route B, which is different from my usual route A."

[0100] When comparing situation PT1 and situation PT2, situation PT1 shows more of the individuality of user U1, and by placing more emphasis on this individuality in learning, the possibility of appropriately creating opportunities for serendipitous encounters increases. For this reason, as described above, when user U1 accidentally exhibits behavior BH, the management unit 136 may assign a higher weight value WV11 to the personal feature data PL_DA1 compared to when user U1 intentionally exhibits behavior BH.

[0101] (9-2. Acquiring response data) In the above embodiment, it has been described that the acquiring unit 131 acquires, as the life data LF_DA, sensor data detected by a sensor provided in the terminal device 10 of the user U. However, the acquiring unit 131 may acquire, as the life data LF_DA, answer data AN_DA of the user U to a question about the situation. For example, the acquiring unit 131 may acquire answer data AN_DA indicating the answer AN of the user U to the question Q by causing the terminal device 10 to output, via the general inference model M, a question Q about the current situation, such as "You seem very relaxed since a while ago. What are you doing now?" Such processing can improve the estimation accuracy of the first situation SU1.

[0102] (9-3. Completing processing on the edge side) In the above embodiment, an example has been shown in which the terminal device 10 and the server device 100 cooperate to perform the information processing according to the embodiment, and the main arithmetic processing (e.g., extraction, management, and learning of difference information DF) is performed on the server device 100 side. However, a configuration may be adopted in which all of the information processing according to the embodiment is completed on the terminal device 10 side. In such an example, the terminal device 10 functions as an example of an information processing device.

[0103] [10. Hardware Configuration] The information processing device may be realized, for example, by a computer 1000 configured as shown in Fig. 9. As an example, the hardware configuration of the server device 100 will be described. Fig. 9 is a hardware configuration diagram showing an example of a computer that realizes the functions of the server device 100. The computer 1000 has a CPU 1100, a RAM 1200, a ROM 1300, an HDD 1400, a communication interface (I / F) 1500, an input / output interface (I / F) 1600, and a media interface (I / F) 1700.

[0104] The CPU 1100 operates and controls each unit based on programs stored in the ROM 1300 or the HDD 1400. The ROM 1300 stores a boot program executed by the CPU 1100 when the computer 1000 starts up, programs that depend on the hardware of the computer 1000, and the like.

[0105] The HDD 1400 stores programs executed by the CPU 1100, data used by these programs, etc. The communication interface 1500 receives data from other devices via a predetermined communication network and sends the data to the CPU 1100, and transmits data generated by the CPU 1100 to other devices via the predetermined communication network.

[0106] The CPU 1100 controls an output device such as a display and an input device such as a keyboard via the input / output interface 1600. The CPU 1100 acquires data from the input device via the input / output interface 1600. The CPU 1100 also outputs generated data to the output device via the input / output interface 1600.

[0107] Media interface 1700 reads a program or data stored in recording medium 1800 and provides it to CPU 1100 via RAM 1200. CPU 1100 loads the program or data from recording medium 1800 onto RAM 1200 via media interface 1700 and executes the loaded program. Recording medium 1800 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or a PD (Phase Change Rewritable Disc), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory.

[0108] For example, when the computer 1000 functions as the server device 100, the CPU 1100 of the computer 1000 executes programs loaded onto the RAM 1200, thereby realizing the functions of the control unit 130. The CPU 1100 of the computer 1000 reads and executes these programs from the recording medium 1800, but as another example, the CPU 1100 may obtain these programs from another device via a predetermined communication network.

[0109] [11. Other] Furthermore, among the processes described in each of the above embodiments, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically using known methods. In addition, the information including the processing procedures, specific names, various data, and parameters shown in the above documents and drawings can be changed as desired unless otherwise specified. For example, the various information shown in each drawing is not limited to the information shown in the drawings.

[0110] Furthermore, the components of each device shown in the figure are conceptual functional components and do not necessarily have to be physically configured as shown in the figure. In other words, the specific form of distribution and integration of each device is not limited to that shown in the figure, and all or part of them can be functionally or physically distributed and integrated in any unit depending on various loads, usage conditions, etc.

[0111] Furthermore, the above-described embodiments can be combined as appropriate within the scope of not causing any contradiction in the processing content.

[0112] Although some of the embodiments of the present application have been described in detail above with reference to the drawings, these are merely examples, and the present invention can be implemented in other forms that include the aspects described in the "present invention" section and that have been modified and improved in various ways based on the knowledge of those skilled in the art. [Explanation of symbols]

[0113] 1. Information Processing Systems 10 Terminal Equipment 100 Server device 120 Storage section 121 Life data storage unit 122 Situation data storage unit 123 Inference result data storage unit 124 Feature data storage unit 130 control section 131 Acquisition Department 132 Estimation Department 133 Judgment section 134 Detector 135 Discrimination part 136 Management Department 137 Learning Department

Claims

1. an acquisition unit that acquires user data relating to a user; an estimation unit that estimates, based on the history of the user data, a first situation that is the situation of the user when the user data included in the history was acquired from the user; a detection unit that detects an emotional change before and after the user exhibits the first situation based on the user data; a determination unit that, when the emotion change is detected, determines whether or not the first situation includes a difference that deviates from a general situation based on a comparison between information about the first situation and information indicating an inference result output by a predetermined generative model in response to an input based on the first situation; a management unit that, when the first situation includes the difference that deviates from a general situation, stores difference information indicating the difference as personal feature data representing personal information of the user, thereby managing the personal feature data so that it is used as learning data for a machine learning model corresponding to the user; An information processing device comprising:

2. a determination unit that determines whether the user has exhibited a first situation that is different from a second situation that is estimated based on the user data and is a situation that is a characteristic of the user; Further preparation, the detection unit detects the emotion change when the user indicates the first situation that is different from the second situation. The information processing device according to claim 1 .

3. The estimation unit estimating, as the second situation, a second behavior indicating a behavior pattern of the user based on the user data; based on the history of the user data, estimating a first behavior, which is the current behavior of the user, as the first situation, and estimating the emotion of the user when the user exhibited the first behavior; the determination unit determines whether the user has exhibited the first behavior that is different from the second behavior; When the user exhibits the first behavior that is different from the second behavior, the detection unit detects an emotion change before and after the user exhibits the first behavior based on the emotion at the time the user exhibits the first behavior. The information processing device according to claim 2 .

4. the determination unit determines whether the user accidentally exhibited the first behavior or whether the user intentionally exhibited the first behavior based on environmental data corresponding to the history of the user data; The management unit assigning a weight value to the personal feature amount data according to whether the user accidentally exhibits the first behavior or whether the user intentionally exhibits the first behavior; The information processing device according to claim 3 .

5. the determination unit determines whether the first situation is a difference deviating from a general situation based on a comparison between information indicating the first situation or a history of the user data used to estimate the first situation and information indicating the inference result, as information regarding the first situation; The information processing device according to claim 1 .

6. the acquisition unit acquires, as the user data, sensor data detected by a sensor provided in the terminal device of the user; The sensor data includes data on the user's position, posture, movement, vital signs, vision, hearing, or food, clothing, and shelter. The information processing device according to claim 1 .

7. the acquiring unit acquires, as the user data, response data of the user to questions about a situation. The information processing device according to claim 1 .

8. An information processing method executed by an information processing device, an acquisition step of acquiring user data relating to a user; an estimation step of estimating, based on the history of the user data, a first situation which is the situation of the user when the user data included in the history was acquired from the user; a detection step of detecting an emotional change before and after the user exhibits the first situation based on the user data; a determination step of determining, when the emotion change is detected, whether or not the first situation includes a difference that deviates from a general situation based on a comparison between information about the first situation and information indicating an inference result output by a predetermined generative model in response to an input based on the first situation; a management step of managing, when the first situation includes the difference that deviates from a general situation, difference information indicating the difference as personal feature data representing personal information of the user, so that the personal feature data is used as learning data for a machine learning model corresponding to the user; An information processing method including:

9. the acquisition procedure for acquiring User Data relating to the User; an estimation step of estimating, based on the history of the user data, a first situation that is the situation of the user when the user data included in the history was acquired from the user; a detection step of detecting an emotional change before and after the user exhibits the first situation based on the user data; a determination step of, when the emotion change is detected, determining whether or not the first situation includes a difference that deviates from a general situation based on a comparison between information about the first situation and information indicating an inference result output by a predetermined generative model in response to an input based on the first situation; a management procedure of managing, when the first situation includes the difference that deviates from a general situation, difference information indicating the difference as personal feature data representing personal information of the user, so that the personal feature data is used as learning data for a machine learning model corresponding to the user; An information processing program that causes a computer to execute the above.

Images