Store support device and store support system and store support program

Abstract

To enhance the customer satisfaction by acquiring data related to the store usage.SOLUTION: A store support device according to an embodiment comprises a first acquisition part, a second acquisition part, and an output part. The first acquisition part acquires data related to the store usage. The second acquisition part inputs the content represented by the data and a question instructing the derivation of an improvement proposal related to the store usage into a generative AI trained using at least data related to the store, and acquires a reply text including the improvement proposal derived by the generative AI. The output part outputs the reply text.SELECTED DRAWING: Figure 9

Description

[Technical Field]

[0001] An embodiment of the present invention relates to a store support device, a store support system, and a store support program. [Background technology]

[0002] Traditionally, analyzing customer satisfaction has been considered an important factor in determining store sales and attracting repeat customers. Existing systems involve analyzing customer responses through surveys and telephone conversations. However, these methods are limited in the number of customer responses they can obtain, and they are unable to capture unexpected complaints in stores or unspoken complaints from consumers.

[0003] Furthermore, a technology has been proposed that converts a customer's voice into text data, sequentially performs morphological analysis, syntactic analysis, and semantic analysis on the converted text data to identify the customer's concept, and supports the improvement of service quality to customers based on the analysis results that include information on the service status of store staff (for example, Patent Document 1).

[0004] In recent years, an AI (Artificial Intelligence) model called a Large Language Model (LLM) has also emerged. LLMs are AI models that have been pre-trained on a large corpus in the field of natural language processing, and when a question or instruction is input as a sentence, for example, they are equipped with the function of generating and outputting a response sentence that is in line with the meaning of the input sentence. Summary of the Invention [Problem to be solved by the invention]

[0005] The problem to be solved by the present invention is to provide a store support device, a store support system, and a store support program that can improve customer satisfaction by acquiring data related to store usage. [Means for solving the problem]

[0006] A store support device according to an embodiment includes a first acquisition unit, a second acquisition unit, and an output unit. The first acquisition unit acquires data related to store usage. The second acquisition unit inputs the content represented in the data and a question instructing the user to derive an improvement plan for store usage to a generation AI trained using at least the store-related data, and acquires an answer sentence including the improvement plan derived by the generation AI. The output unit outputs the answer sentence. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 1 is a schematic diagram showing a schematic configuration of a store support system according to an embodiment. [Figure 2] FIG. 2 is a diagram illustrating an example of the installation of a plurality of microphones in a store according to the embodiment. [Figure 3] FIG. 3 is a diagram illustrating an example of a hardware configuration of a store support device according to the embodiment. [Figure 4] FIG. 4 is a diagram illustrating an example of a correspondence DB according to the embodiment. [Figure 5] FIG. 5 is a diagram illustrating an example of a functional configuration of a controller in the store support device according to the embodiment. [Figure 6] FIG. 6 is a block diagram showing a schematic configuration of a store terminal according to the embodiment. [Figure 7] FIG. 7 is a diagram illustrating an example of an outline of the improvement plan proposing process according to the embodiment. [Figure 8] FIG. 8 is a diagram illustrating an example of input and output to the LLM in the improvement plan proposal process according to the embodiment. [Figure 9] FIG. 9 is a flowchart illustrating an example of a procedure for an improvement plan proposing process according to the embodiment. [Figure 10] FIG. 10 is a diagram illustrating an example of input and output to the LLM in the improvement plan proposal process according to the embodiment. [Figure 11] FIG. 11 is a diagram showing an example of a case where the subject of a detected word is unknown according to the embodiment. [Figure 12]FIG. 12 is a diagram illustrating an example of an outline of a learning process for a generation AI before training according to an embodiment. [Figure 13] FIG. 13 is a diagram illustrating an example of an outline of a learning process for a generation AI before training according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0008] (Embodiment) Hereinafter, an embodiment of the store support system 1 will be described with reference to the drawings. Furthermore, the present invention is not limited to the following embodiment, and the components in the following embodiment include those that would be easily conceived by a person skilled in the art, those that are substantially the same, and those that are within the scope of what is called equivalent. Furthermore, various omissions, substitutions, modifications, and combinations of the components can be made without departing from the spirit of the following embodiment.

[0009] 1 is a schematic diagram showing the overall configuration of a store support system 1 according to an embodiment. As shown in FIG. 1, the store support system 1 includes an external storage device 3, a collection device 5, a store control system 7, a store support device 9, a store terminal 11, and a headquarters terminal 13.

[0010] The external storage device 3 stores, for example, text data such as online store reviews, text data such as store-related questionnaires, text data related to stores on social networking services (SNS), and audio data such as responses to complaints about stores. The external storage device 3 is realized by a hard disk drive (HDD), a solid state drive (SSD), various types of memory, etc.

[0011] The collection device 5, store support device 9, store control system 7, and store terminal 11 are communicably connected to each other via a network Na such as a LAN (Local Area Network) or the Internet. The collection device 5, store support device 9, and store terminal 11 are installed in a store such as a supermarket, for example.

[0012] The collection device 5 is installed in the store. The collection device 5 has a microphone for collecting voices of customers in the store, a scanner for collecting text data of questionnaires installed in the store, and the like. The collection device 5 may also collect the operation history of a tablet-type order terminal. FIG. 2 is a diagram showing an example of the installation of multiple microphones 51 in the store 21. As shown in FIG. 2, the multiple microphones 51 are installed around specific product shelves 23 (for example, a display shelf for prepared foods, a vegetable section, a fresh food section, a dairy section, etc.). The locations where the multiple microphones 51 are installed are not limited to those shown in FIG. 2, and they may also be installed at tables used by customers in the store 21, in waiting rooms, at cash registers, etc.

[0013] The store control system 7 controls the store 21 based on output from the store support device 9 or instructions from a store clerk via the store terminal 11. For example, the store control system 7 controls the store environment of the store 21, such as the temperature, lighting, temperature of refrigerated and frozen foods, and air conditioning, based on instructions from a store clerk via the store terminal 11. Note that the objects controlled by the store control system 7 are not limited to the store environment, but may also include inventory management, sales management such as product management and ordering, etc. Furthermore, the store control system 7 controls the temperature and airflow of air conditioning devices, lighting devices, and the temperature of refrigeration devices related to food and beverages, etc., in accordance with control signals output from the store support device 9. The control signals correspond to, for example, electrical signals for adjusting the temperature, air volume, illuminance, etc. of the air conditioning devices, lighting devices, and / or refrigeration devices.

[0014] A store support device 9 is provided in each of the multiple stores 21. The store support device 9 may be connected to the network Nb as a server device. In this case, the store support device 9 functions, for example, as a server device, and the store terminal 11 functions as a client device. The store support device 9 may be configured by multiple computers. In other words, the functions of the store support device 9 may be distributed among multiple computers. The server device may also be realized by a cloud server installed on the cloud. The store support device 9 may also function as a client device.

[0015] The store support device 9 performs a series of processes related to providing an answer to data related to the use of the store 21. For example, the store support device 9 inputs the content expressed in the data related to the use of the store 21 and a question instructing the generation AI to derive an improvement plan related to the use of the store 21, and outputs an answer including the improvement plan derived by the generation AI. The configuration and functions of the store support device 9 will be described below.

[0016] Fig. 3 is a diagram showing an example of the hardware configuration of the store support device 9. As shown in Fig. 3, the store support device 9 includes a controller 91, a storage device 93 that stores various data and control programs, and a network communication interface (IF) 95 for communicating between the collection device 5, the store terminal 11, and the network Nb via the network Nb.

[0017] The controller 91 in the store support device 9 includes a CPU (Central Processing Unit) 911 that controls the entire store support device 9, a ROM (Read Only Memory) 913 that stores control programs and the like for the CPU 911 to operate, and a RAM (Random Access Memory) 915 that temporarily stores various data. The controller 91 controls the operation of various components by the CPU 911 expanding and executing control programs stored in the ROM 913, storage device 93, etc. The controller 91 also controls the operation of sending and receiving various types of data between the network communication IF 95 and various devices connected via the network Na and network Nb.

[0018] The storage device 93 is realized by a hard disk drive (HDD), a solid state drive (SSD), etc. The store support device 9 may further include an operation unit having a keyboard, a mouse, etc., for a user (e.g., a store clerk) using the store support device 9 to perform various operations, and a display unit configured by a liquid crystal display, etc.

[0019] The storage device 93 stores data (hereinafter referred to as store information) related to the store 21 in which the store support device 9 is installed as a store information database (hereinafter referred to as store information DB) 931. The store information DB 931 is a database that has, for example, store map data, product data such as Point of Sale (POS) data, and data indicating the store environment (various temperatures, lighting intensity, airflow intensity, etc.). The storage device 93 also stores learning samples for a generation AI (Artificial Intelligence) 935 (described below) as a correspondence database (hereinafter referred to as correspondence DB) 933. The learning samples correspond to, for example, questions that instruct the derivation of improvement plans related to the use of the store 21. The correspondence DB 933 corresponds to a set of multiple prompts input to the generation AI 935.

[0020] Fig. 4 is a diagram showing an example of the correspondence DB 933. As shown in Fig. 4, in the correspondence DB 933, for example, character strings corresponding to a response policy, a notification destination / system linkage, and an immediate response / permanent response (long-term response) are associated with numbers that distinguish the prompts and stored as learning samples. The correspondence DB 933 also stores evaluation results based on input to the generation AI 935 and output from the generation AI 935.

[0021] The storage device 93 also stores a generation AI 935 used in the second acquisition unit 105. The generation AI 935 is a trained model capable of generating new content such as sentences by learning data patterns and relationships. The generation AI 935 is trained using at least data related to the store 21. For the sake of concreteness, the generation AI 935 used in this embodiment will be described below as a large-scale language model (hereinafter referred to as an LLM (Large Language Model)). The LLM 935 is an AI model (also referred to as a language model) that has been pre-trained using a large-scale corpus in the field of natural language processing.

[0022] LLM935 is built using large amounts of data and deep learning technology. For example, when a question or instruction is input as a sentence, it generates and outputs a response that matches the meaning of the input sentence. The strings of characters input to LLM935 may be collectively referred to as prompts. LLM935 learning will be explained later.

[0023] The network communication IF 95 functions as an interface for transmitting and receiving various data between the various devices connected to the network Na and the network Nb.

[0024] 5 is a diagram showing an example of the functional configuration of the controller 91 in the store support device 9. The controller 91 has, as its functional configuration, a system control unit 101, a first acquisition unit 103, a second acquisition unit 105, a determination unit 107, an analysis unit 109, and an output unit 110.

[0025] The system control unit 101 controls the entire store support device 9. For example, the system control unit 101 controls various processes executed in the store support device 9. The system control unit 101 also controls the transmission and reception of various data between the store support device 9 and various devices.

[0026] The first acquisition unit 103 acquires data related to the use of the store 21. The data related to the use of the store 21 is, for example, voice data related to the store 21, text data related to the store 21, etc. The voice data is collected, for example, from the customer's speech through a collection device (microphone) 5. At this time, the first acquisition unit 103 acquires the customer's voice data from the collection device 5. The first acquisition unit 103 also acquires various text data related to the store 21 from the external storage device 3. The first acquisition unit 103 stores the data related to the use of the store 21 in RAM 915.

[0027] The second acquisition unit 105 reads the generation AI 935 from the storage device 93. The second acquisition unit 105 also reads a learning sample from the storage device 93. That is, the second acquisition unit 105 reads a question that instructs the derivation of an improvement plan related to the use of the store 21 from the correspondence DB 933. The second acquisition unit 105 also reads store information from the store information DB 931 in the storage device 93. The second acquisition unit 105 inputs the content expressed in the data related to the store 21 and the question that instructs the derivation of an improvement plan related to the use of the store 21 to the generation AI (LLM) 935, and acquires an answer sentence including the improvement plan derived by the generation AI 935. The second acquisition unit 105 may also be referred to as an improvement plan answer sentence generation unit. In this case, the first acquisition unit 103 may simply be referred to as an acquisition unit.

[0028] The determination unit 107 determines whether the timing of the improvement of the store 21 is an immediate response or a long-term response (permanent response) based on the response sentence output from the LLM 935. Specifically, the determination unit 107 performs morphological analysis on the response sentence. The morphological analysis and an executable program are stored in advance in the storage device 93. If the morphological analysis of the response sentence shows that the response sentence contains morphemes corresponding to an immediate response (e.g., a statement about the temperature inside the store) as shown in FIG. 4, for example, the determination unit 107 determines that the response sentence is an immediate response. Furthermore, if the morphological analysis of the response sentence shows that the response sentence contains morphemes corresponding to a long-term response as shown in FIG. 4, for example, the determination unit 107 determines that the response sentence is a long-term response.

[0029] The morphemes corresponding to immediate or long-term responses are set in advance and stored in the storage device 93. Note that the method for determining whether a reply sentence is an immediate or long-term response is not limited to the above, and a trained model or a segmentation model that can distinguish between immediate and long-term responses using the reply sentence as input may also be used.

[0030] The analysis unit 109 analyzes the voice data input to the LLM 935 to quantify the emotion of the customer who has spoken the voice data as an emotion value. For example, the analysis unit 109 calculates the emotion value by analyzing the sound pressure of the voice data and the morphemes of the reply sentence. Specifically, if the reply sentence is positive and the sound pressure is high, the analysis unit 109 determines that the customer has a strong positive emotion toward the store 21 and sets an emotion value with a large absolute value. On the other hand, if the reply sentence is positive and the sound pressure is low, the analysis unit 109 determines that the customer has a weak positive emotion toward the store 21 and sets an emotion value with a small absolute value. On the other hand, if the reply sentence is neither positive nor negative, regardless of the sound pressure, the analysis unit 109 determines that the customer's emotion toward the store 21 is neutral and sets an emotion value of zero.

[0031] Furthermore, if the reply sentence is negative and the sound pressure is high, the analysis unit 109 will set an emotion value that is negative and has a large absolute value, assuming that the negative emotion toward the store 21 is strong. Furthermore, if the reply sentence is negative and the sound pressure is low, the analysis unit 109 will set an emotion value that is negative and has a small absolute value, assuming that the negative emotion toward the store 21 is weak.

[0032] The analysis unit 109 attaches an emotional value to the answer sentence output from the LLM 935. Note that the quantification of the emotion is not limited to the above, and may be performed by a trained model that inputs voice data and outputs an emotional value. In this case, the second acquisition unit 105 may further input the emotional value to the generation AI 935 to output an answer sentence that reflects the emotion.

[0033] The output unit 110 outputs the answer sentence acquired by the LLM 935. At this time, the output unit 110 may output the answer sentence together with the value of the emotion value. For example, if the timing of the improvement of the store 21 is an immediate response, the output unit 110 outputs the answer sentence to a store clerk of the store 21 and / or outputs a control signal based on the answer sentence to a control device for the environment of the store 21. Specifically, if the answer sentence is an immediate response, the output unit 110 outputs the answer sentence to the store terminal 11. Furthermore, if the answer sentence is an immediate response related to the environment of the store 21 (for example, a sentence related to the temperature of the store 21), the output unit 110 outputs a control signal based on the answer sentence (a signal to adjust the temperature setting of the air conditioner) to the store control system 7.

[0034] The store terminal 11 is a terminal device installed in each of a plurality of stores. The store terminal 11 may be realized as a tablet-type terminal device, or may be realized as any terminal device as long as it can be operated by a store clerk in each of the plurality of stores.

[0035] Fig. 6 is a block diagram showing a schematic configuration of the store terminal 11. As shown in Fig. 6, the store terminal 11 includes a controller 111 that controls the entire store terminal 11, a display unit 119, an operation unit 121 that allows the user to perform various operations, and a network communication interface (IF) 123 that communicates with various devices via a network Na.

[0036] The controller 111 in the store terminal 11 includes a CPU 113 that controls the entire store terminal 11, a ROM 115 that stores control programs and the like for the CPU 113 to operate, and a RAM 117 that temporarily stores various data. The CPU 113 executes the control program stored in the ROM 115, and the controller 111 receives, via the network communication IF 123, data transmitted from the store support device 9 over the network Na. The controller 111 displays the received data on the display unit 119. The operation unit 121 may be configured with various switches, buttons, and the like, or may be configured integrally with the display unit 119 as a touch panel.

[0037] The display unit 119 displays the response message regarding the immediate response. At this time, the operation unit 121 inputs the response result for the response message and whether the response message is correct or incorrect, in accordance with the store clerk's instructions. The correctness or incorrectness of the response message corresponds, for example, to the evaluation result of the input / output related to the LLM 935. The controller 111 associates the response message, the response result, and the correctness or incorrectness, transmits them to the store support device 9 via the network communication IF 123, and stores them in the correspondence DB 933 in the storage device 93.

[0038] The headquarters terminal 13 is communicably connected to the network Na and the external storage device 3 via a network Nb such as the Internet. The number of devices connected to the networks Na and Nb is not limited to the example shown in FIG. 1. The network Nb may be the same network as the network Na. The general configuration of the headquarters terminal 13 is the same as that of the store terminal 11.

[0039] The headquarters terminal 13 is a terminal device installed at the headquarters of a chain of stores that includes multiple stores. The headquarters terminal 13 may be realized as a tablet-type terminal device, or may be realized as any terminal device that can be operated by an operator (hereinafter referred to as the headquarters operator) who belongs to the headquarters of the chain of stores. The headquarters terminal 13 is realized, for example, by a computer having an operation unit that allows the headquarters operator to input various types of information, a memory that stores various types of information, and a display that can display various types of information.

[0040] The headquarters terminal 13 displays a response message regarding the long-term response (permanent response) on the display. At this time, the operation unit of the headquarters terminal 13 inputs the response result for the response message and the success or failure of the response message at the instruction of the headquarters operator. The headquarters terminal 13 associates the response message, the response result, and the success or failure, transmits them to the store support device 9, and stores them in the response DB 933 in the storage device 93.

[0041] The configuration of this embodiment has been described above. Below, we will explain the process of proposing an improvement plan (hereinafter referred to as the improvement plan proposal process) performed by the store support device 9. To make the explanation more specific, it is assumed that the data acquired by the first acquisition unit 103 is voice data collected from customer utterances.

[0042] The improvement plan proposal process will be described below with reference to Figs. 7 to 11. Fig. 7 is a diagram showing an example of an outline of the improvement plan proposal process. Fig. 8 is a diagram showing an example of input / output to the LLM 935 in the improvement plan proposal process. The external data in Fig. 8 corresponds to, for example, data related to the use of the store 21. That is, the external data in Fig. 8 corresponds to sentences in which the voice data shown in Fig. 7 is expressed as a character string, various text data related to the store 21 from SNS and questionnaires, etc. Furthermore, the store information data in Fig. 8 corresponds to various data stored in the store information DB 931.

[0043] Moreover, the data showing the response policy, notification destination / system linkage, and immediate / permanent response shown in Fig. 8 corresponds to the data included in the learning sample shown in Fig. 7, for example, the correct answer data in the training process of the generation AI 935. The inference results in Fig. 8 show the answer sentence output from the LLM 935 by the document creation shown in Fig. 7, and the evaluation results of the answer sentence by the headquarters operator and the store clerk.

[0044] Fig. 9 is a flowchart showing an example of the procedure for the improvement plan proposal process. Fig. 10 is a diagram showing an example of input / output to the LLM 935 in the improvement plan proposal process. Fig. 11 is a diagram showing an example when the subject of the detected word is unknown.

[0045] (Improvement proposal processing) (Step S1) The first acquisition unit 103 acquires data related to the use of the store 21 from the collection device 5 or the external storage device 3. The data related to the use of the store 21 is, for example, voice data or text data. When the voice data is acquired, the first acquisition unit 103 may acquire, in addition to the voice data, information about the area around the location where the voice data was acquired (store map data) from the store information DB 931. In addition, the first acquisition unit 103 acquires, from the correspondence DB 933, learning samples that can be input to the LLM 935, i.e., questions that instruct the derivation of improvement plans related to the use of the store 21. In this step, the analysis unit 109 may also calculate an emotion value of the customer based on the voice data.

[0046] The first acquisition unit 103 may also perform morphological analysis on a character string or text based on the voice data. At this time, if the character string or text corresponding to the voice data includes a product name or the like, the first acquisition unit 103 acquires product data for the product from the store information DB 931. If the morphemes of the character string or text corresponding to the voice data include the temperature inside the store, the warmth or coldness of the store 21, or the brightness or darkness of the store 21, the first acquisition unit 103 acquires data indicating the store environment from the store information DB 931 and / or the store control system 7.

[0047] (Step S2) The second acquisition unit 105 inputs data related to the use of the store 21 to the encoder of the LLM 935. Specifically, the second acquisition unit 105 inputs a character string, morpheme by morpheme, sequentially to the encoder of the LLM 935. As shown in FIG. 8, for example, in the case of a character string "The seat is cold," the second acquisition unit 105 inputs these morphemes, "seat" and "cold," in that order, to the encoder. Next, the second acquisition unit 105 inputs a seat number (store information data) indicating the position information of the voice data to the encoder. Note that, if an emotion value has been calculated by the analysis unit 109, the second acquisition unit 105 may input the emotion value to the encoder.

[0048] (Step S3) The second acquisition unit 105 inputs the output from the encoder and the learning samples in the correspondence DB 933 to the decoder of the LLM 935. Note that the input to the decoder may be only the output from the encoder. The timing of the input to the decoder is, for example, shifted by one word from the word input to the encoder.

[0049] (Step S4) The second acquisition unit 105 acquires the answer sentence from the decoder. As a result, the second acquisition unit 105 creates a document as an output from the LLM 935. That is, the second acquisition unit 105 acquires the answer sentence including the improvement plan derived by the LLM 935.

[0050] (Step S5) The determination unit 107 performs morphological analysis on the response sentence to determine whether the improvement plan indicated by the response sentence is an immediate response or a long-term response (permanent response). For example, "plus" and "minus" in FIG. 10 correspond to positive and negative satisfaction with the store 21, respectively. As shown in FIG. 10, for example, if a customer in the waiting room utters, "The air conditioner is cold," the response sentence output from the LLM 935 will be a sentence indicating an instruction to increase the temperature in the store 21 and / or an instruction to decrease the air volume of the air conditioner at the position in the store 21 related to the utterance, for example, "Increase the set temperature and decrease the air volume of the air conditioner close to the waiting room." In this case, the determination unit 107 determines that the response sentence is an immediate response.

[0051] The detected words in Fig. 10 indicate, for example, words detected based on audio data. Furthermore, the items in Fig. 10 correspond, for example, to the types to which the detected words belong. Note that although the emotion value column is blank in Fig. 10, if an emotion value has been calculated by analysis unit 109, a numerical value indicating the emotion value will be displayed in the emotion value column.

[0052] 11, if the subject word input to the LLM 935 is unknown, the LLM 935 infers the subject of "sold out" based on the time of acquisition of the voice data, inventory information, etc., and creates a response sentence corresponding to the voice data. In this case, the response sentence output from the LLM 935 may be, for example, a sentence requesting that the product be replenished from the backroom or a sentence requesting that the product be ordered. In this case, the determination unit 107 determines that the response sentence is an immediate response.

[0053] (Step S6) If it is determined that the response text generated by the LLM 935 is an immediate response (Yes in step S6), the process proceeds to step S7. If it is determined that the response text generated by the LLM 935 is an immediate response (No in step S6), the process proceeds to step S9.

[0054] (Step S7) The output unit 110 outputs the answer sentence to the shop terminal 11. As a result, the shop terminal 11 displays the answer sentence on the display unit 119. At this time, if the analysis unit 109 has calculated an emotion value, the output unit 110 attaches the emotion value to the answer sentence and outputs it to the shop terminal 11. As a result, the shop terminal 11 displays the answer sentence and the emotion value on the display unit 119. Note that if the absolute value of the negative emotion value is large, the shop terminal 11 may further display on the display unit 119, in addition to the answer sentence and the emotion value, a warning indicating that immediate action should be taken.

[0055] For example, if a customer makes a comment that the air conditioning in the store is not suitable for them, the response text will suggest adjusting the air conditioning. If a customer makes a comment that they cannot find a product, the response text will provide guidance to the store clerk. By visually checking these texts, the store clerk can take immediate action.

[0056] (Step S8) The operation unit 121 of the store terminal 11 inputs the response result to the answer sentence and the correctness or incorrectness of the answer sentence in response to the instruction of the store clerk. The store terminal 11 transmits the response result and the correctness or incorrectness to the answer sentence, which is the inference result of the LLM 935, via the network communication IF 123 to the store support device 9.

[0057] (Step S9) The output unit 110 outputs the answer sentence to the headquarters terminal 13. As a result, the headquarters terminal 13 displays the answer sentence on the display unit. At this time, if the analysis unit 109 has calculated an emotion value, the output unit 110 attaches the emotion value to the answer sentence and outputs it to the headquarters terminal 13. As a result, the headquarters terminal 13 displays the answer sentence and the emotion value on the display unit. Note that if the absolute value of the negative emotion value is large, the headquarters terminal 13 may further display a warning on the display unit, in addition to the answer sentence and the emotion value, indicating that immediate action should be taken.

[0058] Furthermore, for example, if an impression of a dish is detected, the reply text will be a text indicating the evaluation of the dish. At this time, the headquarters operator can understand the evaluation of the product. Furthermore, if a complaint is detected, the reply text will be a text that, for example, detects a store 21 where complaints are likely to occur and suggests retraining of the staff at that store 21.

[0059] (Step S10) The operation unit of the headquarters terminal 13 inputs the response result for the answer statement and the correctness or incorrectness of the answer statement in response to the instructions of the headquarters operator. The headquarters terminal 13 sends the response result and the correctness or incorrectness to the answer statement, which is the inference result of the LLM 935, via the network communication IF to the store support device 9.

[0060] (Step S11) The store support device 9 causes the system control unit 101 to store the response result and whether the response was successful or not in the correspondence DB 933. At this time, the system control unit 101 may store the response result in the correspondence DB 933 in association with the speech data and learning sample that are the source of the response response output.

[0061] (Step S12) If new data relating to the use of the store 21 is collected (Yes in step S12), the processes from step S1 onwards are repeated. If new data relating to the use of the store 21 is not collected (No in step S12), the improvement plan proposal process ends.

[0062] If no new data related to the use of the store 21 has been collected (No in step S12), the system control unit 101 may re-train the LLM 935 by inputting the answer sentence, the actual response result to the answer sentence, and the correctness or incorrectness of the answer sentence into the decoder, as shown in FIGS. 7 and 8. At this time, the LLM (generative AI) 935 is re-trained using the data related to the use of the store 21 and the response result. By this re-training of the generative AI 935, the generative AI 935 is adjusted according to the store 21. Note that this re-training of the generative AI 935 may be performed by a learning unit separately provided in the controller 91, instead of by the system control unit 101.

[0063] The learning process of the generation AI 935 in this embodiment will be described below. Figures 12 and 13 are diagrams showing an example of an outline of the learning process for the generation AI before training. As shown in Figures 12 and 13, learning data for the generation AI before training includes, for example, data (training data) input to the encoder and decoder of the LLM 935 and correct answer data associated with the training data. As shown in Figures 12 and 13, the training data includes data related to the use of the store 21.

[0064] 12 and 13, the data related to the use of the store 21 includes, for example, voice data in the collection device 5, text data such as SNS and questionnaires stored in the external storage device 3, the store information DB 931, and learning samples included in the correspondence DB 933. Also, the correct answer data corresponds to the input-output evaluation results as shown in FIG.

[0065] The pre-trained generative AI (LLM) is constructed using a large amount of learning data and deep learning technology, etc. The pre-trained generative AI (LLM) is trained to generate and output a response sentence that matches the meaning of the input sentence, for example, when a question or instruction is input as a sentence. Specifically, the pre-trained generative AI (LLM) is trained using at least data related to the store 21, as shown in Figures 12 and 13. The learned generative AI 935 is stored in the storage device 93 in the store support device 9.

[0066] As described above, the store support device 9 according to this embodiment acquires data related to the use of the store 21, inputs the content of the data and a question instructing the generation of an improvement plan for the use of the store 21 to the generation AI 935, which has been trained using at least the data related to the store 21, acquires a response sentence including the improvement plan derived by the generation AI 935, and outputs the acquired response sentence. As a result, the store support device 9 according to this embodiment does not limit the number of customer data items from which an answer can be obtained. Additionally, the store support device 9 can acquire unexpected complaints at the store or complaints that customers have expressed. By inputting these complaints and complaints into the generation AI 935, it is possible to output a response sentence including an improvement plan. Therefore, the store support device 9 according to this embodiment can improve customer satisfaction by presenting the response sentence to a store clerk and / or a headquarters operator and implementing the improvement plan.

[0067] Furthermore, the store support device 9 according to this embodiment determines whether the timing for improving the store 21 is immediate or long-term based on the response statement generated by the generation AI 935, and if the timing is for immediate response, outputs the response statement to a salesperson at the store 21 and / or outputs a control signal based on the response statement to an environmental control device (store control system 7) for the store 21, and if the timing is for long-term response, outputs the response statement to the headquarters of affiliated stores including the store 21. As a result, the store support device 9 according to this embodiment can implement improvement proposals at a timing that meets the customer's wishes.

[0068] Furthermore, the generation AI 935 in the store support device 9 according to this embodiment is re-trained using data related to use of the store 21 and the response results for the response sentences. As a result, the store support device 9 according to this embodiment can adjust the generation AI 935 according to the store 21, and can output an optimal improvement plan according to the store 21 as a response sentence.

[0069] Furthermore, the data related to the use of the store 21 in the store support device 9 according to the present embodiment includes voice data related to the store 21, and the store support device 9 according to the present embodiment analyzes the voice data to quantify the emotion of the customer who spoke the voice data as an emotion value, and further inputs the quantified emotion value to the generation AI 935 to output a reply that reflects the emotion of the customer. As a result, the store support device 9 according to the present embodiment can improve the accuracy of reply sentences that include improvement suggestions.

[0070] As described above, the store support device 9 according to the embodiment can acquire data related to the use of the store 21, and output a response including an improvement proposal at an appropriate timing according to the customer's request, in accordance with the store 21. Therefore, the store support device 9 according to the embodiment can analyze the customer's request and improve customer satisfaction.

[0071] When the technical idea of ​​this embodiment is realized in a store support system 1, the store support system 1 includes a collection device 5 that collects data related to the use of the store 21, and a store support device 9. The store support device 9 includes a first acquisition unit 103 that acquires data related to the use of the store 21, a second acquisition unit 105 that inputs the content expressed in the data and a question that instructs the deriving of an improvement plan related to the use of the store 21 to a generation AI 935 that has been trained using at least data related to the store 21, and acquires an answer sentence including the improvement plan derived by the generation AI 935, and an output unit 110 that outputs the acquired answer sentence. The processing procedure of the improvement plan proposal processing realized by the store support system 1 is the same as in the embodiment. Furthermore, the effects realized by the store support system 1 are the same as in the embodiment. For these reasons, a description of the processing procedure and effects of the improvement plan proposal processing realized by the store support system 1 will be omitted.

[0072] When the technical idea in the embodiment is realized by a store support program, the store support program causes a computer to acquire data related to the use of store 21, input the content expressed in the acquired data and a question instructing the derivation of an improvement plan related to the use of store 21 to generation AI 935 that has been trained using at least the data related to store 21, acquire an answer sentence including the improvement plan derived by generation AI 935, and output the acquired answer sentence.

[0073] For example, the improvement plan proposal process can be realized by installing a store support program on a computer such as the store support device 9 shown in FIG. 1 or a server device on the network Nb and expanding the program in memory. In this case, the program that enables a computer to execute the process can be stored and distributed on a storage medium such as a magnetic disk (such as a hard disk), an optical disk (such as a CD-ROM or DVD), or a semiconductor memory. Distribution of the store support program is not limited to the above media, and it may be distributed using a telecommunications function, such as downloading via the Internet. The processing procedure of the store support program conforms to the improvement plan proposal process. The effects of the store support program are similar to those of the embodiment. For these reasons, a description of the processing procedure and effects of the store support program will be omitted.

[0074] Although several embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be embodied in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, and are also included in the scope of the invention and its equivalents as defined in the claims. [Explanation of symbols]

[0075] 1. Store support system 3 External storage device 5. Collection Device 7. Store Control System 9. Store support equipment 11 Store terminal 13 Headquarters terminal 21 stores 23 Specific product shelves 91 Controller 93 Storage device 95 Network Communication Interface (IF) 101 System control unit 103 First acquisition part 105 Second acquisition part 107 Judgment section 109 Analysis Department 110 Output section 111 Controller 113 CPU 115 ROM 117 RAM 119 Display Unit 121 Operation section 123 Network Communication Interface (IF) 911 CPU 913 ROM 915 RAM 931 store information database 933 compatible database 935 Generation AI [Prior art documents] [Patent documents]

[0076] [Patent Document 1] Japanese Patent Application Laid-Open No. 2000-347686

Claims

1. a first acquisition unit that acquires data related to store usage; a second acquisition unit that inputs the content expressed in the data and a question that instructs the derivation of an improvement plan related to the use of the store into a generation AI that has been trained using at least the data related to the store, and acquires an answer sentence including the improvement plan derived by the generation AI; an output unit that outputs the answer sentence; Equipped with store support equipment.

2. a determination unit that determines whether to take immediate action or long-term action as the timing for improving the store based on the response text; The output unit If the timing is the immediate response, outputting the reply sentence to a store clerk and / or outputting a control signal based on the reply sentence to a control device of the environment of the store; If the timing is the long-term response, output to the headquarters of the affiliated stores including the store. The store support device according to claim 1.

3. The generating AI is re-trained using data related to the use of the store and the response result to the answer sentence. The store support device according to claim 2.

4. the data relating to the use of the store includes voice data relating to the store, An analysis unit analyzes the voice data to quantify the emotion of the customer who has uttered the voice data as an emotion value, The second acquisition unit further inputs the emotion value to the generation AI, thereby outputting the answer sentence that reflects the emotion. The store support device according to any one of claims 1 to 3.

5. a collection device that collects data related to store usage; a first acquisition unit that acquires data related to the use of the store; a second acquisition unit that inputs the content expressed in the data and a question that instructs the derivation of an improvement plan related to the use of the store into a generation AI that has been trained using at least the data related to the store, and acquires an answer sentence including the improvement plan derived by the generation AI; an output unit that outputs the answer sentence; A store support device having A store support system equipped with the above.

6. On the computer, Collect data related to store usage, The content expressed in the data and a question instructing the derivation of an improvement plan related to the use of the store are input to a generating AI that has been trained using at least the data related to the store; Obtaining a response sentence including the improvement plan derived by the generation AI, outputting the answer sentence; A store support program that makes this a reality.

Images