Disaster risk generation method, information processing device, and disaster risk generation program

Abstract

Disclosed is a disaster risk generation method for generating a disaster risk, wherein a computer inputs, into a generative AI model, a work site image and an imperative sentence indicating a generation instruction to generate disaster risk information pertaining to work on the basis of the work site image. Further, the computer outputs the disaster risk information generated by the generative AI model.

Description

Disaster Risk Generation Method, Information Processing Apparatus, and Disaster Risk Generation Program

[0001] The present disclosure relates to a disaster risk generation method, an information processing apparatus, and a disaster risk generation program.

[0002] Conventionally, in a work support apparatus that supports work at a work site, there is an apparatus that outputs caution information indicating a caution corresponding to the work content and the site situation at the work site. Thereby, the user can work safely and reduce the disaster risk in advance by grasping the caution information.

[0003] Japanese Unexamined Patent Application Publication No. 2023-183217

[0004] However, the caution information output by the conventional apparatus is limited, for example, to the type of building and the weather of the site situation, or the caution information is limited to accident cases, in-house safety standards, precautions for work content, etc. Therefore, there is a need for an apparatus that can be applied to situations other than specific site situations and output cautions that the user has not noticed.

[0005] An object of the present disclosure is to provide a disaster risk generation method, an information processing apparatus, and a disaster risk generation program that can output a disaster risk assumed from the site situation as compared with the prior art.

[0006] To achieve the above object, the disaster risk generation method of the present disclosure is a disaster risk generation method for generating a disaster risk, in which a computer inputs a work site image and a command sentence indicating a generation instruction for generating disaster risk information related to work based on the work site image into a generation AI model. Further, the computer outputs the disaster risk information generated by the generation AI model.

[0007] Figure 1 is a schematic diagram showing an example of an information processing system according to the embodiment. Figure 2 is a functional block diagram showing an example of the functional configuration of a user terminal according to the embodiment. Figure 3 is a functional block diagram showing an example of the functional configuration of a server according to the embodiment. Figure 4 is a schematic diagram showing an example of a work site image according to the embodiment. Figure 5 is a schematic diagram showing an example of disaster risk output by a server according to a comparative example. Figure 6 is a schematic diagram for explaining the first command statement and text information according to the embodiment. Figure 7 is a schematic diagram for explaining the second command statement and disaster risk information according to the embodiment. Figure 8 is a flowchart showing an example of the processing flow executed by a server according to the embodiment. Figure 9 is a schematic diagram for explaining the second command statement and disaster risk information according to the first modified example. Figure 10 is a schematic diagram for explaining the second command statement and disaster risk information according to the second modified example. Figure 11 is a schematic diagram for explaining the first command statement according to the third modified example. Figure 12 is a block diagram showing an example of the hardware configuration of an information processing system according to an embodiment and a modified example.

[0008] The disaster risk generation method, information processing device, and disaster risk generation program according to the embodiments of this disclosure will be described in detail below with reference to the attached drawings.

[0009] (Embodiment) Figure 1 is a schematic diagram showing an example of an information processing system according to an embodiment. As shown in Figure 1, the information processing system 10 includes a user terminal 1 and a server 2. The user terminal 1 and the server 2 are connected to each other so as to be able to communicate via, for example, a network.

[0010] User terminal 1 is a terminal used by users such as administrators, work managers, and workers who manage server 2. User terminal 1 can be various mobile devices such as notebook PCs, tablet devices, and smartphones as appropriate. Alternatively, a stationary terminal such as a regular personal computer may be used as user terminal 1. User terminal 1 transmits work site images and disaster case information related to the company's accident cases, which are entered by the user, to server 2. For example, workers use user terminal 1 to take pictures of their work environment and equipment to understand risks in advance. A detailed explanation of work site images and disaster case information related to the company's accident cases will be provided later.

[0011] Server 2 outputs disaster risks related to the work based on the work site images transmitted by user terminal 1. Server 2 is an example of an information processing device. Details of these processes will be described later.

[0012] Figure 2 is a functional block diagram showing an example of the functional configuration of a user terminal 1 according to an embodiment. The user terminal 1 comprises a communication unit 11, an input unit 12, a display unit 13, a storage unit 14, and a control unit 15. The communication unit 11, input unit 12, display unit 13, storage unit 14, and control unit 15 are connected to each other via a bus or the like so as to be able to communicate.

[0013] The communication unit 11 is a communication interface that communicates with the server 2. For example, the communication unit 11 transmits work site images input by the input unit 12 to the server 2. For example, the communication unit 11 receives disaster risk output by the server 2. The communication unit 11 also communicates various information with external devices.

[0014] The input unit 12 accepts various operations from the user. The input unit 12 may be, for example, a keyboard and a pointing device such as a mouse, a microphone, etc. The display unit 13 displays or outputs various information. The display unit 13 may be a display that shows various information, a speaker that outputs various sounds, etc. The input unit 12 and the display unit 13 may be integrated into a touch panel. Alternatively, the input unit 12 and the display unit 13 may be configured separately from the user terminal 1. In this case, the input unit 12 and the display unit 13 and the user terminal 1 may be connected in a way that allows them to communicate with each other.

[0015] The storage unit 14 stores various types of data. The storage unit 14 may be, for example, a semiconductor memory element such as RAM (Random Access Memory) or flash memory, a hard disk, or an optical disc. The storage unit 14 may also be a storage device located outside the user terminal 1. Furthermore, the storage unit 14 may be a storage medium on which programs and various types of information are downloaded and stored or temporarily stored via a LAN (Local Area Network) or the Internet.

[0016] The control unit 15 is a computer that performs information processing on the user terminal 1. The control unit 15 may be implemented by having a processor such as a CPU (Central Processing Unit) execute a program, i.e., by software. It may also be implemented by a dedicated IC or other processor, i.e., by hardware. The storage unit 14 may be implemented using a combination of software and hardware. At least one of the one or more functional units included in the control unit 15 may be mounted on an external information processing device that is connected to the user terminal 1 via a network or the like in a way that allows communication.

[0017] Figure 3 is a functional block diagram showing an example of the functional configuration of Server 2 according to an embodiment. Server 2 is an example of an information processing device. Server 2 comprises a communication unit 21, a storage unit 22, a first acquisition unit 23, a first generation unit 24, a second acquisition unit 25, a second generation unit 26, a third acquisition unit 27, and an output unit 28. The communication unit 21, storage unit 22, first acquisition unit 23, first generation unit 24, second acquisition unit 25, second generation unit 26, third acquisition unit 27, and output unit 28 are connected to each other via a bus or the like so as to be able to communicate.

[0018] The communication unit 21 is a communication interface that communicates with the user terminal 1. The communication unit 21 also communicates various information with external devices.

[0019] The storage unit 22 stores various types of information. The storage unit 22 may be, for example, a semiconductor memory element such as RAM or flash memory, a hard disk, or an optical disk. The storage unit 22 may also be a storage device located outside the server 2. Alternatively, the storage unit 22 may be a storage medium that stores or temporarily stores programs and various types of information downloaded via a LAN or the internet.

[0020] Furthermore, the memory unit 22 stores accident case information related to the company's accident cases. Accident case information is, for example, a report compiled regarding an accident that occurred within the company. The report includes items such as the details of the accident, the causative agent, and the contributing factors. The details of the accident are, for example, being caught in something, being entangled, falling, slipping, or tripping. The causative agent is, for example, equipment, stairs, or the environment. The contributing factors include, for example, material factors, human factors, and management factors. Material factors are, for example, defects in the equipment or the work environment itself. Human factors are, for example, psychological factors or physiological factors. Management factors are, for example, failure to implement safety measures. Note that the items included in the report are not limited to these.

[0021] The first acquisition unit 23, the first generation unit 24, the second acquisition unit 25, the second generation unit 26, the third acquisition unit 27, and the output unit 28 are computers that perform information processing on the server 2. The first acquisition unit 23, the first generation unit 24, the second acquisition unit 25, the second generation unit 26, the third acquisition unit 27, and the output unit 28 may be implemented by having a processor such as a CPU execute a program, i.e., by software. They may also be implemented by a dedicated processor such as an IC, i.e., by hardware. The storage unit 22 may be implemented using a combination of software and hardware. At least one of the first acquisition unit 23, the first generation unit 24, the second acquisition unit 25, the second generation unit 26, the third acquisition unit 27, and the output unit 28 may be mounted on an external information processing device that is connected to the server 2 via a network or the like so as to be able to communicate.

[0022] The first acquisition unit 23 acquires images of the work site. Specifically, the first acquisition unit 23 acquires images of the work site from the user terminal 1. The work site images will be explained using Figure 4.

[0023] Figure 4 is a schematic diagram showing an example of a work site image according to the embodiment. The work site image 40 shown in Figure 4 is, for example, an image showing a worker carrying objects in both hands descending a staircase. Here, the disaster risk output by the server according to the comparative example will be explained using Figure 5.

[0024] Figure 5 is a schematic diagram showing an example of disaster risk output by the server in the comparative example. The server in the comparative example outputs disaster risk using, for example, an LLM (Large-Scale Language Model) that has not been trained on past disaster cases. For example, in the schematic diagram 50 shown in Figure 5, the server in the comparative example outputs disaster risk information 52 based on the work site image 40 shown in Figure 4 and the command statement 51.

[0025] Command statement 51 displays text that reads, "Please tell us about the disaster risks." Disaster risk information 52 displays text that reads, "Especially when the stairs are wet or when workers are carrying heavy or unstable objects, it is important for workers to keep a firm grip on the stair railing to prevent slipping and falling."

[0026] The disaster risk information 52 shows general disaster risks. However, if a user wants to identify disaster risks related to the causes of industrial accidents from the work site image 40, they may find the content of the disaster risk information 52 insufficient.

[0027] Therefore, the server 2 according to this embodiment will now be described in a form that outputs disaster risks requested by the user from the work site image 40. The server 2 includes at least a processor. The processor inputs the work site image 40 and a command statement indicating a generation instruction to generate disaster risk information related to the work based on the work site image 40 to a generation AI model, and outputs the disaster risk information generated by the generation AI model. Here, the generation AI model is a visual language model and a language model.

[0028] Returning to Figure 3, the explanation continues. The first generation unit 24 generates a first command statement indicating a first generation instruction to be generated by a Vision-Language Model (VLM) capable of providing text information related to the work site image 40, based on the work site image 40. Specifically, the first generation unit 24 generates a first command statement indicating a first generation instruction to be generated by a Vision-Language Model capable of providing text information related to the work site image 40, based on the work site image 40 acquired by the first acquisition unit 23. The first command statement is an instruction or input statement to the Vision-Language Model, and is also called a prompt.

[0029] The second acquisition unit 25 inputs a first command statement to the visual language model indicating a first generation instruction to cause the visual language model, which is capable of providing text information related to the work site image 40, to generate text information related to the work site image 40, based on the work site image 40, and acquires the text information output by the visual language model. Specifically, the second acquisition unit 25 inputs a first command statement to the visual language model indicating a first generation instruction to cause the visual language model, which is capable of providing text information related to the work site image 40, to generate text information based on the work site image 40 generated by the first generation unit 24, and acquires the text information output by the visual language model. The first command statement and text information will be explained using Figure 6.

[0030] Figure 6 is a schematic diagram illustrating the first command statement and text information according to the embodiment. The schematic diagram 60 shown in Figure 6 shows the first command statement 61 input by the second acquisition unit 25 to the visual language model and the text information 62 acquired by the second acquisition unit 25 from the visual language model. The first command statement 61 shows text that reads, "Please tell us the situation that can be seen from the image in 50 words or less. If anything related to the following keywords is visible, please output the information preferentially: Wearing gloves for arm and leg movement on stairs." Here, "Wearing gloves for arm and leg movement on stairs" shown in the first command statement 61 is a keyword that indicates a point of focus regarding the cause of industrial accidents in the work, as entered by the user.

[0031] Text information 62 displays the following text: "• The worker is on the stairs, holding a bottle in one hand and something like a cloth or rag in the other. • As the cleaning supplies indicate, the worker may be cleaning or performing maintenance." This allows server 2 to identify key points regarding the factors contributing to workplace accidents that can be read from the work site image 40.

[0032] Returning to Figure 3, the explanation continues. The second generation unit 26 generates a second command statement indicating a second generation instruction to cause a language model capable of providing accident risk information related to the work to generate it, based on the text information 62. Specifically, the second generation unit 26 generates a second command statement indicating a second generation instruction to cause a language model capable of providing accident risk information related to the work to generate it, based on the text information 62 acquired by the second acquisition unit 25. The second command statement includes points of focus regarding the factors of industrial accidents in the work.

[0033] A language model is a model that has been trained using past disaster cases as training data. Language models are, for example, generative AI that can perform question answering and text generation by computer programs and systems used in the field of natural language processing (NLP). Examples of language models include large language models (LLM), generative language models (for example, GPT (Generative Pretrained Transformer)), representational language models (for example, Bidirectional Encoder Representations from Transformers (BERT)), and / or any other type of language model.

[0034] The generative AI has a language model that has been pre-trained using a large amount of text data from various fields. Here, the generative AI may undergo so-called fine-tuning, in which it retrains a dataset from a specific field to generate a model specialized for that field in order to handle a particular task. In the embodiments of this application, the specific task is, for example, outputting the disaster risk that can be expected from the work content and site conditions at a work site. Also, in the embodiments of this application, the specific field is, for example, identifying the work content at a work site from images of the work site taken at the work site.

[0035] The third acquisition unit 27 inputs a second command statement to the language model indicating a second generation instruction to cause a language model capable of providing disaster risk information related to work to generate it based on the text information 62, and acquires the disaster risk information output by the language model. Specifically, the third acquisition unit 27 inputs a second command statement to the language model indicating a second generation instruction to cause a language model capable of providing disaster risk information related to work to generate it based on the text information 62 generated by the second generation unit 26, and acquires the disaster risk information output by the language model.

[0036] The output unit 28 outputs disaster risk information. Specifically, the output unit 28 outputs the disaster risk information acquired by the third acquisition unit 27 to the user terminal 1. The second command statement and disaster risk information will be explained using Figure 7.

[0037] Figure 7 is a schematic diagram illustrating the second command statement and disaster risk information according to the embodiment. The schematic diagram 70 shown in Figure 7 shows the second command statement 71 input by the third acquisition unit 27 to the language model and the disaster risk information 72 acquired by the third acquisition unit 27 from the language model. The second command statement 71 shows text that reads: "• A worker is on the stairs, holding a bottle in one hand and a cloth or rag in the other. • As indicated by the cleaning supplies, the worker may be cleaning or performing maintenance. • Please tell me the disaster risk corresponding to the above." The disaster risk information 72 shows text that reads: "• Especially if the stairs are wet or if the worker is carrying heavy or unstable objects, it is important for the worker to keep a firm grip on the stair railing to prevent slipping or falling. • Also, the worker needs to pay attention to the condition of their feet and the stairs to avoid tripping. • Furthermore, if the worker is carrying cleaning supplies, care must be taken not to spill liquid on the stairs, as this could create a risk of slipping."

[0038] Here, we compare disaster risk information 72 with disaster risk information 52 shown in Figure 5. Disaster risk information 72 contains more information about disaster risks than disaster risk information 52. Specifically, disaster risk information 72 indicates that measures must be taken to prevent workers from tripping and that there is a possibility of slipping. In other words, disaster risk information 72 is information that includes unknown disaster risks. As a result, server 2 can output disaster risks that can be expected from the site conditions, compared to conventional methods.

[0039] Figure 8 is a flowchart showing an example of the processing flow executed by the server 2 according to this embodiment.

[0040] The first acquisition unit 23 acquires a work site image 40 from the user terminal 1 (step S81). Subsequently, the first generation unit 24 generates a first command statement 61 indicating a first generation instruction to cause a visual language model capable of providing text information related to the work site image 40 to generate text information related to the work site image 40, based on the work site image 40 acquired by the first acquisition unit 23 (step S82). Subsequently, the second acquisition unit 25 inputs the first command statement 61 generated by the first generation unit 24, which indicates a first generation instruction to cause a visual language model capable of providing text information related to the work site image 40 to generate text information related to the work site image 40, to the visual language model, and acquires the text information 62 output by the visual language model (step S83).

[0041] Next, the second generation unit 26 generates a second command statement 71 indicating a second generation instruction to cause a language model capable of providing disaster risk information related to work to generate it, based on the text information 62 acquired by the second acquisition unit 25 (step S84). Subsequently, the third acquisition unit 27 inputs the second command statement 71, which indicates a second generation instruction to cause a language model capable of providing disaster risk information related to work to generate it, based on the text information 62, generated by the second generation unit 26, into the language model and acquires the disaster risk information 72 output by the language model (step S85). Subsequently, the output unit 28 outputs the disaster risk information 72 acquired by the third acquisition unit 27 to the user terminal 1 (step S86). When the processing in step S86 is completed, this process executed by the server 2 is completed.

[0042] As described above, the disaster risk generation method of the embodiment is a disaster risk generation method for generating disaster risks, wherein a computer inputs a work site image 40 and a command statement indicating a generation instruction to generate disaster risk information related to the work based on the work site image 40 to a generation AI model, and outputs the disaster risk information generated by the generation AI model.

[0043] Further, the generative AI model is a vision-language model and a language model. In the disaster risk generation method of the embodiment, the computer inputs a first command sentence 61 indicating a first generation instruction for causing a vision-language model capable of providing text information related to the work site image 40 to be generated based on the work site image 40 into the vision-language model. Also, in the disaster risk generation method of the embodiment, the computer inputs a second command sentence indicating a second generation instruction for causing a language model capable of providing disaster risk information related to the work to be generated based on the text information output by the vision-language model into the language model. Furthermore, the disaster risk generation method of the embodiment outputs the disaster risk information output by the language model.

[0044] For example, the warning information output by conventional devices is limited to the type of building or the weather of the site situation, or the warning information is limited to accident cases, internal safety standards, precautions for work content, etc. Therefore, there is a need for a device that can be applied to situations other than specific site situations and output warnings that users have not noticed.

[0045] In contrast, the disaster risk generation method of the embodiment can be applied to situations other than specific site situations by causing a vision-language model capable of providing text information related to the work site image 40 to be generated based on the work site image 40. Also, since the disaster risk generation method of the embodiment outputs disaster risk information 72 including unknown disaster risks, it includes warnings that users have not noticed. As a result, the disaster risk generation method of the embodiment can output disaster risks assumed from the site situation as compared with the prior art.

[0046] Note that the above-described embodiment can also be appropriately modified and implemented by changing a part of the configuration or function of each of the above-described devices. Therefore, some modification examples according to the above-described embodiment will be described as other embodiments below. In the following, the points different from the above-described embodiment will be mainly described, and detailed descriptions of the points common to the already described content will be omitted.

[0047] (First Modified Example) In the above-described embodiment, the second instruction sentence 71 input by the server 2 to the language model described the content related to the disaster risk. In the first modified example, the second instruction sentence with the disaster risk more concretized and the content of the disaster risk information will be described.

[0048] FIG. 9 is a schematic diagram for explaining the second instruction sentence and the disaster risk information according to the first modified example. The schematic diagram 90 shown in FIG. 9 shows the second instruction sentence 91 input by the third acquisition unit 27 to the language model and the disaster risk information 92 acquired by the third acquisition unit 27 from the language model. The second instruction sentence 91 is generated by the second generation unit 26. The second instruction sentence 91 shows the text with the content of "・Please tell me the risk of falling." The disaster risk information 92 shows the text with the content of "・Do not go down the stairs while holding something in your hand. There is a risk of slipping and falling."

[0049] Here, the second instruction sentence 91 is compared with the second instruction sentence 71 shown in FIG. 7. The content of the disaster risk in the second instruction sentence 91 is more concretized than that in the second instruction sentence 71. That is, the second instruction sentence 91 includes related terms in past disaster cases. Also, the disaster risk information 92 is compared with the disaster risk information 72 shown in FIG. 7. Since the second instruction sentence 91 is more concretized, the content related to the risk of falling is described in the disaster risk information 92 compared to the disaster risk information 72. Thereby, the server 2 can output the disaster risk more demanded by the user.

[0050] (Second Modified Example) For example, the second instruction sentence 71 input by the server 2 to the language model may include, together with the disaster risk, the countermeasures corresponding to the disaster risk. In the second modified example, the second instruction sentence including the disaster risk and the countermeasures and the content of the disaster risk information will be described.

[0051] Figure 10 is a schematic diagram illustrating the second instruction and disaster risk information related to the second modification. The schematic diagram 100 shown in Figure 10 shows the second instruction 101 input by the third acquisition unit 27 to the language model and the disaster risk information 102 acquired by the third acquisition unit 27 from the language model. The second instruction 101 is generated by the second generation unit 26. The second instruction 101 is the text, "Please tell me about the risk of falling and countermeasures." The disaster risk information 102 is the text, "Do not go down the stairs with something in your hands. You may slip and fall. Do not carry anything in both hands; put what you are carrying in a bag, etc., and go down the stairs with one hand free."

[0052] Here, we compare the second instruction statement 101 with the second instruction statement 91 shown in Figure 9. The second instruction statement 101 includes the content of the second instruction statement 91 plus the content of countermeasures to address the risk of tipping over. We also compare the disaster risk information 102 with the disaster risk information 92 shown in Figure 9. The disaster risk information 102 includes content related to countermeasures for tipping over, as the content of countermeasures to address the risk of tipping over has been added to the second instruction statement 101. As a result, the server 2 can output disaster risks and countermeasures to address those disaster risks that are more relevant to the user.

[0053] (Third variation) For example, the first instruction may include points of focus regarding the factors of disaster risk in order to better identify the disaster risk.

[0054] Figure 11 is a schematic diagram illustrating the first command statement relating to the third modified example. The schematic diagram shown in Figure 11 shows the first command statement 110 input to the visual language model by the second acquisition unit 25. The first command statement 110 is generated by the first generation unit 24. The first command statement 110 shows text that reads: "First, count the number of people in the image. Then, focus on the people and describe what they are about to do, their posture and where they are looking. Next, state whether these people are holding anything in one hand or both hands. Finally, describe any potential hazards in the surroundings, such as objects on the ground, stairs, or objects near the head. If there are no people, describe the condition of the ground and any objects there. Answer each question in a short sentence."

[0055] Here, we compare the first instruction 110 with the first instruction 61 shown in Figure 6. The first instruction 110 includes more points of focus regarding factors of disaster risk than the first instruction 61. Points of focus regarding factors of disaster risk include, for example, the situation of the workers (e.g., number of people, posture, direction of gaze), the equipment and objects that the workers are handling, and environmental information captured in the image (e.g., the surrounding environment above, at hands, and at feet). Points of focus may also include what the workers are wearing (e.g., whether they are wearing work clothes or a suit). In other words, the first instruction 110 includes a generation instruction to analyze points of focus regarding factors of industrial accidents in the work. As a result, the server 2 can identify points of focus regarding the work content, work environment, and factors of disaster risk that can be read from the work site image 40.

[0056] (Fourth Modification) For example, Server 2 may output first disaster risk information output by a model trained using past disaster cases as training data, and second disaster risk information output by an untrained model that has not trained on past disaster cases. Also, Server 2 may merge or omit the first and second disaster risk information depending on the output priority or the similarity between the first and second disaster risk information.

[0057] Here, the first disaster risk information is, for example, the disaster risk information 52 shown in Figure 5. The second disaster risk information is, for example, the disaster risk information 72 shown in Figure 7, the disaster risk information 92 shown in Figure 9, and the disaster risk information 102 shown in Figure 10. For example, if the output priority is high, server 2 outputs the first disaster risk information and the second disaster risk information to user terminal 1. Also, for example, if the output priority is low, server 2 outputs at least one of the first disaster risk information or the second disaster risk information to user terminal 1.

[0058] For example, the output priority could be set so that users can specify the number and percentage of output items for the first disaster information and the second disaster information (for example, displaying 5 items for the first disaster information, 3 items for the second disaster information, or only the first disaster information), making it easier for users to check the disaster information they consider more important.

[0059] Furthermore, for example, if the text contained in the first disaster risk information and the second disaster risk information is similar, server 2 outputs at least one of the first or second disaster risk information to user terminal 1. For example, if the text contained in the first disaster risk information and the second disaster risk information is not similar, server 2 outputs both the first and second disaster risk information to user terminal 1. This allows the user to understand the disaster risk information before and after learning from past disaster cases. Therefore, server 2 can output disaster risks that can be expected from the on-site situation compared to conventional methods.

[0060] (Fifth Modification) For example, disaster risk information may include past disaster case information showing past disaster cases. This allows users to compare the disaster risk information with past disaster case information, thereby verifying the validity of the disaster risk information.

[0061] (Sixth Modification) In the above-described embodiment, the generation AI model has been described as being provided by Server 2, but is not limited thereto. For example, the generation AI model may have a generation AI server different from Server 2 that is communicated with Server 2 via a network or the like. In this case, Server 2 sends a work site image and a command statement indicating a generation instruction to generate disaster risk information related to the work based on the work site image to the generation AI server, and receives the disaster risk information generated by the generation AI server.

[0062] (Hardware Configuration) Figure 12 is a block diagram showing an example of the hardware configuration of the information processing system 10 according to the embodiment and modified example. The user terminal 1 and server 2 of the above embodiment and modified example have a processor 121, main memory 122, auxiliary storage 123, and device I / F 124 etc. interconnected by a bus 125, and have a hardware configuration that uses a normal computer.

[0063] The processor 121 is, for example, a CPU, and is a computing device that controls the user terminal 1 and server 2 in the above embodiment and its modified form. The main memory 122 is, for example, RAM, and stores programs and the like that realize information processing by the processor 121. The auxiliary storage device 123 is, for example, ROM, and stores data necessary for various processes by the processor 121. The device I / F 124 is an interface connected to the communication unit, storage unit, etc., for sending and receiving data. In the user terminal 1 and server 2 in the above embodiment and its modified form, the processor 121 reads programs from the main memory 122 onto the auxiliary storage device 123 and executes them, thereby realizing each of the above-mentioned functional units on the computer.

[0064] Furthermore, the programs for executing the above-mentioned processes performed on the user terminal 1 and server 2 in the above-described embodiment and modified version may be stored in an HDD (hard disk drive). Alternatively, the programs for executing the above-mentioned processes performed on the user terminal 1 and server 2 in the above-described embodiment and modified version may be pre-installed and provided in the main memory 122.

[0065] Furthermore, the program for executing the above-described process performed on the user terminal 1 and server 2 in the above-described embodiment and modified version may be provided as a computer program product by being stored in an installable or executable file format on a computer-readable storage medium such as a CD-ROM, CD-R, memory card, DVD (Digital Versatile Disk), or flexible disk (FD).

[0066] Furthermore, the program for executing the information processing performed on the user terminal 1 and server 2 of the above embodiment and its modifications may be stored on a computer connected to a network such as the Internet and provided by allowing users to download it via the network. Alternatively, the program for executing the information processing performed on the user terminal 1 and server 2 of the above embodiment and its modifications may be provided or distributed via a network such as the Internet.

[0067] According to at least one embodiment described above, it is possible to output disaster risks that can be anticipated from the site conditions, compared to conventional methods.

[0068] Although embodiments have been described above, these embodiments are presented as examples only and are not intended to limit the scope of the invention. This novel embodiment can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. This embodiment and its variations are included in the scope and spirit of the invention, as well as in the claims of the invention and its equivalents.

[0069] (Note) The various aspects of this disclosure are described below as a summary. (1) A disaster risk generation method for generating disaster risks, wherein a computer inputs a work site image and an instruction statement indicating a generation instruction to generate disaster risk information relating to work based on the work site image to a generation AI model, and outputs the disaster risk information generated by the generation AI model. (2) The disaster risk generation method according to (1) above, wherein the generation AI model is a visual language model and a language model, and the computer inputs a first instruction statement indicating a first generation instruction to cause the visual language model, which is capable of providing text information relating to the work site image, to generate based on the work site image, inputs a second instruction statement indicating a second generation instruction to cause the language model, which is capable of providing disaster risk information relating to work, to generate based on the text information output by the visual language model, and outputs the disaster risk information output by the language model. (3) The disaster risk generation method according to (2) above, wherein the first instruction includes a point of focus regarding the factors of industrial accidents in the work, input by the user. (4) The disaster risk generation method according to (2) above, wherein the first instruction includes a generation instruction to analyze the point of focus regarding the factors of industrial accidents in the work. (5) The disaster risk generation method according to (3) or (4) above, wherein the second instruction includes the point of focus. (6) The disaster risk generation method according to (2) above, wherein the second instruction includes relevant terms from past disaster cases. (7) The disaster risk generation method according to (1) or (2) above, wherein the disaster risk information includes unknown disaster risks. (8) The disaster risk generation method according to (2) above, wherein the language model is a model that has learned past disaster cases as training data.(9) The disaster risk generation method according to (1) or (2) above, wherein the computer outputs first disaster risk information output by a model that has learned past disaster cases as training data, and second disaster risk information output by an untrained model that has not learned the past disaster cases, and outputs the first disaster risk information and the second disaster risk information merged or omitted according to the priority of the output or the similarity between the first disaster risk information and the second disaster risk information. (10) The disaster risk generation method according to (1) or (2) above, wherein the disaster risk information includes past disaster case information showing past disaster cases. (11) An information processing device comprising at least a processor, wherein the processor transmits a work site image and a command statement indicating a generation instruction to generate disaster risk information related to work based on the work site image to a generating AI model, and receives the disaster risk information generated by the generating AI model. (12) A disaster risk generation program that causes a computer to perform the following actions: sending an image of a work site and a command statement indicating a generation instruction to generate disaster risk information related to the work based on the image of the work site to a generation AI server, and receiving the disaster risk information generated by the generation AI server.

[0070] 1 User terminal 2 Server 10 Information processing system 21 Communication unit 22 Storage unit 23 First acquisition unit 24 First generation unit 25 Second acquisition unit 26 Second generation unit 27 Third acquisition unit 28 Output unit

Claims

1. A disaster risk generation method for generating disaster risks, wherein a computer inputs a work site image and a command statement indicating a generation instruction to generate disaster risk information related to the work based on the work site image into a generation AI model, and outputs the disaster risk information generated by the generation AI model.

2. The method for generating disaster risk according to claim 1, wherein the generating AI model comprises a visual language model and a language model, the computer inputs a first command statement to the visual language model indicating a first generation instruction to cause the visual language model, which is capable of providing text information related to the work site image, to generate the text information related to the work site image, based on the work site image; the computer inputs a second command statement to the language model indicating a second generation instruction to cause the language model, which is capable of providing disaster risk information related to the work, to generate the text information output by the visual language model; and the computer outputs the disaster risk information output by the language model.

3. The accident risk generation method according to claim 2, wherein the first instruction statement includes a point of focus regarding the factors of industrial accidents in the work, which is entered by the user.

4. The accident risk generation method according to claim 2, wherein the first instruction statement includes a generation instruction for analyzing points of focus regarding the factors of industrial accidents in the work.

5. The method for generating disaster risk according to claim 3 or 4, wherein the second instruction includes the aforementioned point of focus.

6. The disaster risk generation method according to claim 2, wherein the second instruction includes relevant terminology from past disaster cases.

7. The disaster risk generation method according to claim 1 or 2, wherein the disaster risk information includes unknown disaster risks.

8. The method for generating disaster risks according to claim 2, wherein the language model is a model that has been trained using past disaster cases as training data.

9. The method for generating disaster risks according to claim 1 or 2, wherein the computer outputs first disaster risk information output by a model that has been trained using past disaster cases as training data, and second disaster risk information output by an untrained model that has not been trained on past disaster cases, and depending on the priority of the output, or the similarity between the first disaster risk information and the second disaster risk information, the computer merges or omits the first disaster risk information and outputs the second disaster risk information.

10. The disaster risk generation method according to claim 1 or 2, wherein the disaster risk information includes past disaster case information showing past disaster cases.

11. An information processing device comprising at least a processor, wherein the processor transmits a work site image and a command statement indicating a generation instruction to generate disaster risk information related to the work based on the work site image to a generation AI model, and receives the disaster risk information generated by the generation AI model.

12. A disaster risk generation program that causes a computer to perform the following actions: send an image of a work site and a command statement indicating a generation instruction to generate disaster risk information related to the work based on the image of the work site to a generation AI server, and receive the disaster risk information generated by the generation AI server.

Images