Information processing device, information processing method, and program

Abstract

This information processing device comprises a generation means for generating medical information from input data by using a first language model, and an acquisition means for acquiring a medical document generated by a second language model for which the medical information serves as at least a part of an input.

Description

Information processing device, information processing method, and program

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

[0002]

[0003] Technologies for supporting the creation of medical documents (also called medical documents) or other documents are known. For example, Patent Literature 1 discloses an information processing system including a document creation support device 100 that supports the creation of documents used in medical care.

[0003] Japanese Patent Application Laid-Open No. 2020-166864

[0004] Generally, high accuracy is required for generating medical documents or other documents. However, the technique described in Patent Document 1 has a problem in terms of accuracy.

[0005] The present disclosure has been made in consideration of the above problems, and has an object to provide a document generation technology with high accuracy.

[0006] An information processing device according to one aspect of the present disclosure includes a generation means for generating medical information from input data using a first language model, and an acquisition means for acquiring medical documents generated by a second language model that uses the medical information as at least a portion of its input.

[0007] An information processing method according to one aspect of the present disclosure includes generating medical information from input data using a first language model, and obtaining a medical document generated by a second language model that uses the medical information as at least a portion of its input.

[0008] A program according to one aspect of the present disclosure causes a computer to perform a generation process for generating medical information from input data using a first language model, and an acquisition process for acquiring medical documents generated by a second language model that uses the medical information as at least a portion of its input.

[0009] According to the present disclosure, a highly accurate document generation technology can be provided.

[0010] FIG. 1 is a block diagram showing a configuration of an information processing device according to the present disclosure. FIG. 2 is a flow diagram showing a flow of an information processing method according to the present disclosure. FIG. 3 is a block diagram showing a configuration of an information processing system according to the present disclosure. FIG. 4 is a flow diagram showing a flow of an information processing method according to the present disclosure. FIG. 5 is a diagram for explaining information processing according to the present disclosure. FIG. 6 is a diagram for explaining information processing according to the present disclosure. FIG. 7 is a diagram for explaining information processing according to the present disclosure. FIG. 8 is a block diagram showing a hardware configuration of an information processing device according to the present disclosure.

[0011] The following are examples of embodiments of the present invention. However, the present invention is not limited to the following exemplary embodiments, and various modifications are possible within the scope of the claims. For example, embodiments obtained by appropriately combining the technical means employed in the following exemplary embodiments may also be included in the scope of the present invention. Alternatively, embodiments obtained by appropriately omitting some of the technical means employed in the following exemplary embodiments may also be included in the scope of the present invention. Furthermore, the effects described in the following exemplary embodiments are examples of effects expected in the exemplary embodiments, and do not define the scope of the present invention. In other words, embodiments that do not exhibit the effects described in the following exemplary embodiments may also be included in the scope of the present invention.

[0012] [First Embodiment] A first exemplary embodiment, which is an example of an embodiment of the present invention, will be described in detail with reference to the drawings. This exemplary embodiment is a basic form for each of the exemplary embodiments described below. Note that the scope of application of each technical means employed in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technical means employed in this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical obstacles arise. Furthermore, each technical means shown in the drawings referenced to explain this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical obstacles arise.

[0013] (Configuration of information processing device 1) The configuration of the information processing device 1 according to this exemplary embodiment will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the configuration of the information processing device 1. As shown in Fig. 1, the information processing device 1 includes a generation unit 11 and an acquisition unit 12.

[0014] (Generation Unit 11) The generation unit 11 generates medical information from input data using a first language model. Here, the input data referenced by the generation unit 11 includes, as an example, unstructured data such as text. As another example, the input data includes unstructured data such as text and structured data such as a table. Furthermore, as an example, the input data is data related to the medical field, and a specific example is data included in an electronic medical record (also referred to as electronic medical record data). However, this example does not limit the present exemplary embodiment, and the input data may also be data outside the medical field.

[0015] The medical information is, for example, data referenced by a second language model described below. For example, the medical information includes a plurality of entity data constituting elements of the data, and relationship data defining the relationships between the entity data. Here, for example, the entity data is a named entity extracted from the input data described above. More specifically, the entity data is a medical-related named entity (also referred to as a medical named entity) extracted from the input data described above.

[0016] The first language model is a trained model that extracts the medical information from the input data. The first language model is, for example, a model trained by machine learning using training data including text data in the medical field. Specific examples of the first language model include BERT (Bidirectional Encoder Representations from Transformers) or other natural language processing models (natural language processing algorithms), but this does not limit the present exemplary embodiment.

[0017] (Acquisition unit 12) The acquisition unit 12 acquires a medical document generated by a second language model that uses the medical information as at least a part of its input. As an example, the acquisition unit 12: generates a prompt including the medical information, the prompt instructing the second language model to create a medical document by referring to the medical information; provides the generated prompt to a second language model included in the information processing device 1 or a second language model included in a device other than the information processing device 1; and acquires text output by the second language model to which the prompt has been input as a medical document or material for a medical document.

[0018] Here, the text acquired by the acquisition unit 12 may be output as is as a medical document, or the text may be corrected and the corrected text may be output as a medical document.

[0019] The second language model is, for example, a language model provided in an external device of the information processing device 1, and is a model trained using not only texts in the medical field but also texts related to multiple other fields, but this does not limit the present exemplary embodiment.

[0020] (Effects of information processing device 1) As described above, the information processing device 1 is configured to generate medical information from input data using a first language model, and to obtain medical documents generated by a second language model that uses the medical information as at least a portion of the input.

[0021] In this way, the information processing device 1 employs a configuration for generating medical documents in multiple stages using multiple language models, making it possible to provide a highly accurate document generation technology.

[0022] (Flow of Information Processing Method S1) Next, the flow of information processing method S1 according to this exemplary embodiment will be described with reference to Fig. 2. Fig. 2 is a flow diagram showing the flow of information processing method S1. As shown in Fig. 2, information processing method S1 includes a step (process) S11 of generating medical information from input data and a step (process) S12 of acquiring medical documents.

[0023] (Step S11) In step S11, the generator 11 generates medical information from input data using the first language model. A more specific description of the generator 11 has been given above, so a description thereof will be omitted here.

[0024] (Step S12) In step S12, the acquisition unit 12 acquires a medical document generated by a second language model that uses the medical information as at least a part of an input. Since the specific processing performed by the acquisition unit 12 has been described above, a detailed description thereof will be omitted here.

[0025] (Effects of information processing method S1) As described above, information processing method S1 employs a configuration in which medical information is generated from input data using a first language model, and a medical document generated by a second language model that uses the medical information as at least a portion of the input is obtained.

[0026] In this way, the information processing method S1 employs a configuration for generating medical documents in multiple stages using multiple language models, and therefore the information processing device 1 can provide a highly accurate document generation technology.

[0027] [Second Embodiment] A second exemplary embodiment, which is one example of an embodiment of the present invention, will be described in detail with reference to the drawings. Components having the same functions as those described in the above exemplary embodiment will be assigned the same reference numerals, and their description will be omitted as appropriate. The scope of application of each technical means employed in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technical means employed in this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical obstacles arise. Furthermore, each technical means shown in each drawing referenced to describe this exemplary embodiment can also be employed in other exemplary embodiments included in the present disclosure, to the extent that no particular technical obstacles arise.

[0028] (Configuration of Information Processing System 1A) The configuration of the information processing system 1A according to this exemplary embodiment will be described with reference to Fig. 3. Fig. 3 is a block diagram showing the configuration of the information processing system 1A. As shown in Fig. 3, the information processing system 1A includes an information processing device 100 and a generating device 50 connected to the information processing device 100 via a network N. Here, the specific configuration of the network N does not limit this exemplary embodiment, but as an example, a wireless local area network (LAN), a wired LAN, a wide area network (WAN), a public line network, a mobile data communication network, or a combination of these networks can be used.

[0029] (Generation device 50) As shown in FIG. 3 , the generation device 50 includes a control unit 51, a storage unit 52, and a communication unit 53. The communication unit 53 communicates with devices external to the generation device 50. As an example, the communication unit 53 communicates with the information processing device 100 included in the information processing system 1A. The communication unit 53 transmits data supplied from the control unit 51 to the information processing device 100, and supplies data received from the information processing device 100 to the control unit 51. Note that the data received by the communication unit 53 from the information processing device 100 may include a prompt generated by the information processing device 100. Furthermore, the data provided by the communication unit 53 to the information processing device 100 may include text data generated by a second language model LM2 (described later) based on the prompt.

[0030] A second language model LM2 is stored in the storage unit 52. As an example, a plurality of parameters defining the second language model LM2 are stored in the storage unit 52. As an example, these parameters are parameters that have been learned in advance by machine learning (parameters that have undergone an update process by machine learning), but this does not limit the present exemplary embodiment.

[0031] The second language model LM2 is, for example, a model trained using not only texts in the medical field but also texts related to multiple other fields, although this is not intended to limit the present exemplary embodiment.

[0032] The control unit 51 executes the second language model LM2 to obtain an output result from the second language model LM2. As an example, the control unit 51 inputs a prompt received from the information processing device 100 into the second language model LM2 and obtains text data generated by the second language model LM2. The control unit 51 also provides the text data to the information processing device 100 via the communication unit 53.

[0033] Note that, in the present exemplary embodiment, the generation device 50 is illustrated as a device separate from the information processing device 100, but this does not limit the present exemplary embodiment. The control unit 51 included in the generation device 50 or the function of the language model execution unit in the control unit 51 may be configured to be included in the control unit of the information processing device 100. Similarly, the second language model LM2 stored in the storage unit 52 included in the generation device 50 may be stored in the storage unit of the information processing device 100, and the information processing device 100 itself may be configured to be able to execute the second language model LM2.

[0034] (Configuration of Information Processing Device 100) Next, the configuration of the information processing device 100 according to this exemplary embodiment will be described with reference to Fig. 3. As shown in Fig. 3, the information processing device 100 includes a control unit 10, a storage unit 20, a communication unit 30, and an input / output unit 40.

[0035] (Communication unit 30) The communication unit 30 communicates with devices external to the information processing device 100. As an example, the communication unit 30 communicates with the generation device 50. The communication unit 53 transmits data supplied from the control unit 10 to the generation device 50, and supplies data received from the generation device 50 to the control unit 10. Note that the data transmitted by the communication unit 30 to the generation device 50 may include a prompt generated by the control unit 10. Furthermore, the data received by the communication unit 30 from the generation device 50 may include text data generated by the generation device 50 based on the prompt.

[0036] (Input / Output Unit 40) The input / output unit 40 is configured to include at least one of input / output devices such as a keyboard, a mouse, a display, a printer, a touch panel, etc. Alternatively, the input / output unit 40 may be configured to be connected to input / output devices such as a keyboard, a mouse, a display, a printer, a touch panel, etc. In this configuration, the input / output unit 40 accepts various types of information input to the information processing device 100 from the connected input devices. Furthermore, the input / output unit 40 outputs various types of information to connected output devices under the control of the control unit 10. An example of the input / output unit 40 is an interface such as a USB (Universal Serial Bus).

[0037] (Storage Unit 20) The storage unit 20 stores various types of data referenced by the control unit 10 and various types of data generated by the control unit 10. As an example, the storage unit 20 stores the following: Electronic medical record data EMR as an example of input data Structured data SD included in the electronic medical record data EMR Unstructured data USD included in the electronic medical record data EMR De-identified data DID obtained by applying de-identification processing to at least one of the structured data SD and the unstructured data USD Event data ED generated from the de-identified data DID Entity group EG included in the event data ED Association data RD included in the event data ED Prompt PR generated by the generation unit 11 Medical document MD generated by the second language model LM2 based on the prompt PR Medical document RMD corrected by the correction unit 14 First language model LM1 used by the control unit 10, etc.

[0038] Here, the electronic medical record data EMR is, for example, a set of data included in the electronic medical record of a subject (subject patient), and includes details of medical treatment, progress of medical treatment, nursing records, test results, test images, etc. The medical treatment includes examination, diagnosis, and treatment of the subject. The treatment may also include administration and prescription of medication to the subject.

[0039] Structured data SD refers to structured data included in electronic medical record data EMR. Here, structured data generally refers to data whose structure is predefined, for example, data structured by at least one of the concepts of "rows" and "columns." In this exemplary embodiment, the term "structured data SD" with the symbol SD refers to structured data included in electronic medical record data EMR. In addition, in this exemplary embodiment, the term "structured data" without the symbol SD is used to generally refer to data whose structure is predefined.

[0040] The data format of the structured data SD included in the electronic medical record data EMR is not limited to this exemplary embodiment, but examples include a table format, a CSV (Comma Separated Values) format, and a relational database format. Prescription history, examination history, rehabilitation history, medical receipts, etc. included in the electronic medical record data EMR are expressed as the structured data SD. In addition, the prescription history, examination history, rehabilitation history, etc. included in the electronic medical record data EMR may be stored as ordering records for the subject.

[0041] Unstructured data USD refers to unstructured data included in electronic medical record data EMR. Here, unstructured data generally refers to data that does not have a predefined structure. In this exemplary embodiment, the term "unstructured data USD" with the symbol USD refers to unstructured data included in electronic medical record data EMR. Also, in this exemplary embodiment, the term "unstructured data without the symbol USD" is used to generally refer to data that does not have a predefined structure.

[0042] A specific example of unstructured data USD included in the electronic medical record data EMR is text data created by medical professionals. Examples of unstructured data USD include text data, progress notes, memos, etc., related to symptoms, medical examination details, and the progress of a medical condition, which are included in the electronic medical record data EMR.

[0043] The de-identified data DID is data obtained by applying de-identification (anonymization) to at least one of the structured data SD and the unstructured data USD by the generating unit 11 described later. Here, the de-identification refers to, for example, a process of de-identifying personally identifiable information included in the above-mentioned data. Specific examples of the de-identification process will be described later.

[0044] The event data ED is data generated from the de-identified data DID by the generator 11 (described later). The event data ED includes, as an example, an entity group EG and relationship data RD representing the relationships between the entities included in the entity group EG. In this exemplary embodiment, the event data ED is an example of medical information referenced by the generator 11 (described later) to generate a prompt PR. Furthermore, since the event data ED includes the relationship data RD as described above, it is classified as structured data. Specific examples of the event data ED will be described later.

[0045] The prompt PR is data generated by the generation unit 11 (described later) and input to the second language model LM2. The prompt PR is data including instructions for causing the second language model LM2 to generate a medical document, and includes, for example, an instruction sentence and medical information. Specific examples of the prompt PR will be described later.

[0046] The medical document MD is a medical document generated by the second language model LM2 based on a prompt PR generated by the generation unit 11, which will be described later. Specific examples of the medical document MD will be described later. The corrected medical document RMD is a medical document obtained by applying a correction process to the medical document MD by the correction unit 14. Specific examples of the corrected medical document RMD will be described later.

[0047] The first language model LM1 executes various language processes performed by the control unit 10. As an example, the first language model LM1 executes: a non-identification process executed by the generation unit 11; an event data generation process executed by the generation unit 11; and a prompt generation process executed by the generation unit 11, but this does not limit the present exemplary embodiment.

[0048] As described in exemplary embodiment 1, the first language model LM1 is, for example, a model trained by machine learning using training data including text data in the medical field. Specific examples of the first language model include BERT (Bidirectional Encoder Representations from Transformers) or other natural language processing models (natural language processing algorithms), but this does not limit the exemplary embodiment.

[0049] 3, the control unit 10 includes a generating unit 11 and an acquiring unit 12, similar to the first exemplary embodiment. The control unit 10 also includes a first presenting unit 13, a correcting unit 14, and a second presenting unit 15.

[0050] (Acquisition unit 12) The acquisition unit 12 acquires input data. As an example, the acquisition unit 12 acquires input data input via the communication unit 30 or the input / output unit 40, or input data stored in the storage unit 20, and provides the acquired input data to each unit included in the control unit 10. An example of input data acquired by the acquisition unit 12 is the electronic medical record data EMR described above.

[0051] The acquisition unit 12 also acquires a user's correction instruction for at least one of the event data ED (medical information), the visualized data of the event data ED, and the prompt PR generated by the generation unit 11. The generation unit 11 refers to the correction instruction to correct at least one of the event data ED (medical information), the visualized data of the event data ED, and the prompt PR.

[0052] The acquisition unit 12 also acquires a medical document MD generated by a second language model LM2 based on a prompt PR generated by the generation unit 11 (described later). The prompt PR includes the event data ED (medical information) described above. Therefore, the acquisition unit 12 may be described as being configured to acquire a medical document MD generated by the second language model LM2 that uses medical information as at least a portion of its input. The acquisition unit 12 stores the acquired medical document MD in the memory unit 20, for example. The stored medical document MD is referenced by the control unit 10.

[0053] (Generation unit 11) The generation unit 11 generates event data ED (medical information) from electronic medical record data EMR (input data) using a first language model LM1. Here, the generation process of the event data ED by the generation unit 11 includes a process of generating the event data ED, which is structured data, by referring to unstructured data USD included in the electronic medical record data EMR.

[0054] The process of generating the event data ED by the generator 11 also includes a process of de-identifying the personally identifiable information contained in the electronic medical record data EMR. Here, personally identifiable information refers to information that can identify a subject by referencing the information, and includes, for example, personal information such as the subject's name, date of birth, address, and facial photograph. Furthermore, the personally identifiable information in this exemplary embodiment may include information other than these personal information.

[0055] Furthermore, the above-mentioned de-identification processing may include, as an example, at least one of the following: - A process of deleting personally identifiable information contained in the electronic medical record data EMR, in other words, a process of extracting named entities other than personally identifiable information from one or more named entities contained in the electronic medical record data EMR; - A process of changing personally identifiable information contained in the electronic medical record data EMR based on predetermined rules or replacing it with an identifier; etc., but these examples do not limit this exemplary embodiment.

[0056] The generator 11 also generates a prompt PR to be input to the second language model LM2 using the generated event data ED (medical information). Specific examples of the prompt PR generated by the generator 11 will be described later.

[0057] The generator 11 may be configured to generate visualized data of the event data ED (medical information) in order to present the event data ED to a user. As an example, the visualized data may be configured to visibly include a timeline related to the treatment of the subject.

[0058] The generation unit 11 may also be configured to modify at least one of the event data ED (medical information), the visualized data of the event data ED, and the prompt PR by referring to a user's instruction to modify at least one of the event data ED (medical information), the visualized data of the event data ED, and the prompt PR acquired by the acquisition unit 12. Specific processing by the generation unit 11 will be described later. The first language model LM1 can be used for each of the above-described processing by the generation unit 11.

[0059] (First Presentation Unit 13) The first presentation unit 13 presents the medical document MD generated by the second language model LM2 based on the prompt PR generated by the generation unit 11. As an example, the first presentation unit 13 displays the medical document MD on a display included in the input / output unit 40 or a display connected to the input / output unit 40.

[0060] The first presenting unit 13 may also be configured to present the medical document RMD corrected by the correcting unit 14, which will be described later. As an example, the first presenting unit 13 displays the corrected medical document RMD on a display included in the input / output unit 40 or a display connected to the input / output unit 40.

[0061] (Modification Unit 14) The modification unit 14 modifies the medical document MD generated by the second language model LM2 based on the prompt PR generated by the generation unit 11. The modification process by the modification unit 14 may or may not be based on a modification instruction and modification content from a user. For example, the modification process by the modification unit 14 includes replacing one or more words or phrases included in the medical document MD with a hypernym, a hyponym, a synonym, or a similar word. The modification unit 14 may also change the content of the modification process depending on the type of medical document MD. For example, if the medical document MD is primarily intended for medical professionals, the name of the illness or injury included in the medical document MD may not be changed, but if the medical document MD is primarily intended for non-medical professionals, the name of the illness or injury included in the medical document MD may be replaced with a hypernym or a synonym. Performing such a process by the modification unit 14 has the advantage of generating a medical document that is more suitable for those who refer to the medical document. However, these examples do not limit the present exemplary embodiment. Specific correction processing by the correction unit 14 will be described later.

[0062] The correction unit 14 can use the first language model LM1 in the above-described processes.

[0063] (Second Presentation Unit 15) The second presentation unit 15 presents at least one of the visualized data of the event data ED (medical information) generated by the generation unit 11 and the prompt PR generated by the generation unit 11. As an example, the second presentation unit 15 displays at least one of the visualized data of the event data ED (medical information) and the prompt PR on a display included in the input / output unit 40 or a display connected to the input / output unit 40.

[0064] Furthermore, the second presenter 15 may present data including a timeline relating to the treatment of the subject as visualization data of the event data ED (medical information). The data including the timeline is generated by the generator 11 with reference to the event data ED.

[0065] In the above description, the first presentation unit 13 and the second presentation unit 15 are described as separate blocks, but this does not limit this exemplary embodiment, and these two presentation units may be combined into one presentation unit.

[0066] (Processing example by information processing device 100) Next, a processing example by the information processing device 100 according to this exemplary embodiment will be described with reference to Figures 4 and 5. Figure 4 is a flow diagram showing the flow of a processing example (also referred to as information processing method S1A) by the information processing device 100. Figure 5 is a diagram for explaining the processing example by the information processing device 100, and is a diagram showing examples of various processes executed in the processing example and data referenced in or generated by each process.

[0067] (Step S101) In step S101, the acquisition unit 12 acquires input data. As described above, an example of the input data is electronic medical record data EMR. Furthermore, as shown in FIG. 5, the electronic medical record data EMR includes various tables as structured data and text information as unstructured data. Here, the tables are an example of the structured data SD described above, and the text information is an example of the unstructured data USD described above. Furthermore, examples of the tables include ordering records related to the subject, such as prescription history, examination history, and rehabilitation history. However, this does not limit the present exemplary embodiment.

[0068] (Step S102) Subsequently, in step S102, the generation unit 11 performs a process of de-identifying the personally identifiable information included in the electronic medical record data EMR (input data) and generating event data ED (medical information) by referring to the de-identified data. Here, the de-identification process may include: a process of deleting the personally identifiable information included in the electronic medical record data EMR, in other words, a process of extracting named entities other than the personally identifiable information from one or more named entities included in the electronic medical record data EMR; and a process of changing the personally identifiable information included in the electronic medical record data EMR based on a predetermined rule or replacing it with an identifier.

[0069] More specifically, an example of the processing in step S102 is a process of generating anonymized text information DID-1 by anonymizing personally identifiable information included in the text information of the electronic medical record data EMR, as shown in step S102-1 of Figure 5. Here, an example of the anonymization process is a process of changing personally identifiable information included in the text information of the electronic medical record data EMR based on a predetermined rule or replacing it with an identifier. Furthermore, the anonymized text information DID-1 is an example of the above-mentioned de-identified data DID.

[0070] Another example of the processing in step S102 is processing for extracting medical named entities from the anonymized text information DID-1 and various tables of the electronic medical record data EMR, as shown in step S102-2 of Figure 5. More specifically, this example includes processing for extracting medical named entities that are not personally identifiable information and are medical-related named entities from one or more named entities included in various tables of the electronic medical record data EMR. Here, the medical named entities extracted by the generation unit 11 are also referred to as entity data EG.

[0071] Furthermore, as shown in step S102-2 of Figure 5, the generation unit 11: - applies the above-mentioned medical named entity extraction process to the text information DID-1 after the anonymization process and various tables of the electronic medical record data EMR, and then - executes a relationship extraction process, which is a process of generating (extracting) relationship data RD, which is data that specifies the relationship between each of the extracted named entities.

[0072] By performing the above processing, the generation unit 11 can generate structured event data ED (medical information) including medical unique expressions (entity data EG) and relationship data RD from the structured data SD included in the electronic medical record data EMR and the unstructured data USD included in the electronic medical record data EMR.

[0073] (Step S103) Subsequently, in step S103, the generation unit 11 uses the event data ED (medical information) to generate a prompt PR to be input into the second language model LM2. As an example, the generation unit 11 generates the prompt PR including an instruction statement indicating an instruction to generate a medical document and the event data ED (medical information) generated in step S102. Here, the instruction statement may be one input by the user, a predetermined instruction statement, or one selected by the user from a plurality of instruction statement candidates.

[0074] 6 is a diagram showing prompt PR1, which is an example of a prompt PR generated by the generation unit 11. Prompt PR1 is, as an example, a prompt for causing the second language model LM2 to create a discharge summary (also called a discharge summary) as a medical document. As shown in FIG. 6 , prompt PR1 includes: A tag indicating a command statement (#command statement); command statement content (PRE11); a tag indicating an identifier (#identifier); identifier content (PRE12); a tag indicating event data (medical information) (#event data); event data content (PRE13); a tag indicating a prescription history (#prescription history); and prescription history content (PRE14).

[0075] The content of the command (PRE11) includes the content corresponding to the instruction described above. In this example, the instruction includes the content "Please prepare a discharge summary as a medical document based on the following event data and prescription history."

[0076] The content of the identifier (PRE12) indicates an identifier assigned to the subject of the medical document. As an example, the generation unit 11 may use, as the identifier (PRE12), an identifier assigned in the anonymization process in step S102 (step S102-1 in FIG. 5 ) as a substitute for the subject's personal identification information. Alternatively, the generation unit 11 may be configured to retain correspondence information that associates the subject with the identifier, and may be configured to include the identifier (PRE12) in the prompt PR1 based on the correspondence information.

[0077] The content of the event data (PRE13) includes the content corresponding to the event data ED generated in step S102. In this example, the event data ED includes each event from "Test AAA performed on 2023 / 01 / 30" to "Discharged from hospital on 2023 / 02 / 28."

[0078] The prescription history contents (PRE14) include the prescription history as structured data SD included in the input data (electronic medical record data EMR) acquired in step S101.

[0079] The prompt PR generated in this step is provided to the generation device 50 via the communication unit 30. The generation unit 11 may be configured to provide the medical document template to the generation device 50 (second language model LM2) together with the above-mentioned prompt PR or by including the medical document template in the prompt PR.

[0080] (Step S12) Subsequently, in step S12, the generation device 50 acquires the prompt PR generated in step S103 and inputs the acquired prompt PR into the second language model LM2 to generate a medical document MD. The generated medical document MD is provided to the information processing device 100 and acquired by the acquisition unit 12.

[0081] 7 is a diagram showing a medical document MD1, which is an example of a medical document MD generated by the second language model LM2 based on the prompt PR1 and acquired by the acquisition unit 12. As shown in Fig. 7, the medical document MD1 includes, for example, data attributes (MDE11), admission and discharge information (MDE12), diagnosis information (MDE13), surgery information (MDE14), and prescription information (MDE15).

[0082] The data attribute (MDE11) includes, for example, an identifier included in the prompt PR1. By referencing the identifier, the information processing device 100 can identify the intended subject of the medical document MD1.

[0083] The admission / discharge information (MDE12), diagnosis information (MDE13), and surgery information (MDE14) are, for example, information generated by the second language model LM2 based on the event data (PRE13) included in the prompt PR1. As shown in Figure 7, the admission / discharge information (MDE12), diagnosis information (MDE13), and surgery information (MDE14) reflect the contents of the event data (PRE13).

[0084] The prescription information (MDE15) is, for example, information generated by the second language model LM2 based on the prescription history (PRE14) included in the prompt PR1. As shown in Fig. 7, the prescription information (MDE15) reflects the contents of the prescription history (PRE14).

[0085] For example, the medical document MD acquired in this step may be presented to the user by the first presenting unit 13 or the second presenting unit 15 via a display provided in the input / output unit 40. Alternatively, the medical document MD acquired in this step may be subjected to a correction process in step S104, which will be described later, and the corrected medical document RMD may be presented.

[0086] (Step S104) In step S104, the correction unit 14 corrects the medical document MD acquired in step S12 to generate a corrected medical document RMD. The corrected medical document RMD is presented to the user by the first presenting unit 13 or the second presenting unit 15, for example, via a display provided in the input / output unit 40. In this case, the first presenting unit 13 or the second presenting unit 15 can prompt the user to decide whether or not to make further corrections to the corrected medical document RMD. Here, the correction process by the correction unit 14 may or may not be performed in accordance with a correction instruction and the content of the correction from the user.

[0087] Fig. 8 is a diagram showing RMD1, which is an example of a corrected medical document RMD generated by the correction unit 14. The correction unit 14 generates the corrected medical document RMD1 by correcting the medical document MD1 shown in Fig. 7. As shown in Fig. 8, the corrected medical document RMD includes the following data items, similar to the medical document MD1: Data attributes (MDE11R), Admission and discharge information (MDE12), Diagnosis information (MDE13R), Surgery information (MDE14R), and Prescription information (MDE15).

[0088] The data attribute (MDE11R) is data obtained by correcting the data attribute (MDE11) included in the pre-correction medical document MD by the correcting unit 14. In the example shown in Fig. 8, the correcting unit 14 references the identifier "ID_0011" included in the data attribute (MDE11) shown in Fig. 7 and identifies the name, gender, and date of birth of the subject associated with the identifier. Then, the correcting unit 14 includes the identified name, gender, and date of birth of the subject in the corrected data attribute (MDE11R).

[0089] In the example shown in Figure 8, the correction unit 14 does not perform any correction processing on the admission / discharge information (MDE12) and prescription information (MDE15), and the admission / discharge information (MDE12) and prescription information (MDE15) before correction are included in the corrected medical document RMD.

[0090] The diagnostic information (MDE13R) includes the disease name "bbb" corrected by the correction unit 14. As an example, the correction unit 14 generates the corrected disease name "bbb" by replacing the disease name "BBBBBB" in the pre-correction diagnostic information (MDE13) shown in Fig. 7 with at least one of a hypernym, a hyponym, a synonym, and a similar word. As an example, if the medical document MD is presented mainly to non-medical professionals, the correction unit 14 may be configured to replace the disease name included in the medical document MD with a hypernym or a synonym.

[0091] The surgery information (MDE14R) includes the surgery name "ccc" corrected by the correction unit 14. As an example, the correction unit 14 generates the corrected surgery name "ccc" by replacing the surgery name "CCCCCC" in the pre-correction surgery information (MDE14) shown in Figure 7 with at least one of a hypernym, a hyponym, a synonym, and a similar word. As an example, if the medical document MD is presented primarily to non-medical professionals, the correction unit 14 may be configured to replace the surgery name included in the medical document MD with a hypernym or a synonym.

[0092] (Step S105) In step S105, the first presenting unit 13 or the second presenting unit 15 presents the medical document RMD corrected in step S104 to the user, for example, via a display provided in the input / output unit 40. In addition, the control unit 10 may output the corrected medical document RMD to the outside of the information processing device 100 via the input / output unit 40 or the communication unit 30.

[0093] (Additional Note 1) In step S103 described above, the information processing device 100 may further perform the following processing. Note that, as an example, the following processing is performed before the prompt PR is provided to the generation device 50. The generation unit 11 generates visualization data (visualization data of the event data ED) from the event data ED generated in step S102 to visually present the event data ED to the user. The first presentation unit 13 or the second presentation unit 15 presents at least one of the event data ED, the visualization data of the event data ED, and the prompt PR to the user via, as an example, a display provided in the input / output unit 40. Here, the presented visualization data of the event data ED may be configured to include personal identification information (such as name and date of birth) of the target person to make it easier to identify the target person.

[0094] 9 is a diagram showing timeline data TLD, which is an example of visualization data generated by the generation unit 11 in step S103. As shown in FIG. 9 , the timeline data TLD includes, for example, the subject's name, date of birth, age, and gender; disease name; underlying disease name; treatment period; examination history; symptom and lesion history; treatment and intervention history; medication history; and advanced medical treatment application history. In addition, in the timeline data TLD, the above-mentioned examination history, symptom and lesion history, treatment and intervention history, medication history, and advanced medical treatment application history are represented in chronological order to make them easier for the user to visually recognize. In this way, the timeline data TLD, which is visualization data of the event data ED (treatment information), includes a timeline related to the subject's treatment.

[0095] This allows users such as medical professionals to easily recognize the subject's treatment history, etc. Furthermore, users such as medical professionals can easily check whether the subject's treatment history, etc. contains any incorrect information.

[0096] Furthermore, in the above-described step S103, the information processing device 100 may further perform the following processes: The acquisition unit 12 acquires a user's instruction to modify at least one of the event data ED (medical information), the visualized data of the event data ED, and the prompt PR, and the generation unit 11 modifies at least one of the event data ED (medical information), the visualized data of the event data ED, and the prompt PR in accordance with the user's instruction to modify. The modified prompt PR is provided to the generation device 50.

[0097] This allows the information processing device 100 to instruct the second language model LM2 to generate a medical document MD based on the event data ED or prompt PR that has been confirmed and corrected as necessary by a user such as a medical professional, thereby enabling a more accurate medical document MD to be obtained.

[0098] (Additional Note 2) In each of the above embodiments, specific examples of medical documents acquired by the information processing device 1 or 100 are not limited to the above embodiments, but include the following documents as examples: The information processing device 1 or 100 can acquire a medical document MD in any format by providing the generation device 50 with information such as the items and format to be included in the medical document MD, together with or as part of the prompt PR as necessary.

[0099] (Related work category: surgery) ・Explanation (surgery, anesthesia, examination, procedure, treatment method) (Related work category: explanation of condition) ・Receipt (detailed description of symptoms) ・Explanation of medical treatment details (Related work category: at the time of decision to admit patient) ・Admission / discharge center contact form ・Request for medication instructions (Related work category: at the time of admission / preoperative) ・Inpatient treatment plan ・Pre-operative confirmation checklist (Related work category: at the time of discharge) ・Discharge summary (Related work category: hospital liaison, medical certificate, etc.) ・First visit thank-you letter ・Out-of-hospital referral letter ・In-hospital referral letter ・Medical certificate (specified by the hospital) ・Medical certificate (for recipients of medical insurance, allowances, welfare benefits, etc.) ・Medical certificate (for physically disabled patients) (Related work category: department-specific) ・Consent form for chemotherapy ・Consent form for gastrostomy (PEG) insertion ・Gastrostomy checklist (required document for gastroenterology) ・Radiation therapy request document ・Explanation of medical treatment details In this manner, the medical document MD generated by the second language model LM2 in this exemplary embodiment, or the medical document RMD after correction by the correction unit 14, includes at least one of a surgery description, a medical condition description, an inpatient treatment plan, a discharge summary, a referral letter, a medical certificate, a consent form, and a request form.

[0100] Furthermore, documents that can be acquired by the information processing device 1 or 100 performing the processes described in the exemplary embodiments above are not limited to medical documents. The information processing device 1 or 100 may perform the following processes: referencing the structured data SD and unstructured data USD included in the input data, structuring the unstructured data USD by using, for example, a first language model LM1 to generate event data ED, the event data ED including the structured data SD included in the input data, generating a prompt PR by referring to the event data ED and an instruction sentence, generating a document by providing the generated prompt PR to a second language model LM2 included in the information processing device 1 or 100 or external to the information processing device 1 or 100, and acquiring the generated document, and generating a corrected document by modifying the acquired document using a modification unit 14 as necessary.

[0101] (Additional Note 3) The control unit 10 of the information processing device 100 may be configured to optimize the first language model LM1 described above. As an example, the correction unit 14 of the control unit 10 may be configured to optimize the first language model LM1, which is a machine learning model, by referencing the medical document MD generated by the second language model LM2. As an example, the correction unit 14 may be configured to retain the contents of past correction processes on multiple medical documents MD and to train (optimize) the first language model LM1 so as to reduce the amount of correction content on the medical document MD. With this configuration, the information processing device 100 can generate (acquire) more suitable medical documents.

[0102] (Effects of Exemplary Embodiment 2) The information processing device 100 having the above configuration achieves the same effects as the information processing device 1 according to exemplary embodiment 1. Furthermore, according to the information processing device 100, the process of generating the event data ED (medical information) in step S102 includes a process of de-identifying the personally identifiable information included in the input data, so that the prompt PR generated with reference to the event data ED does not include the personally identifiable information of the subject.

[0103] Therefore, even if the second language model LM2 is provided in a device (e.g., the generation device 50) external to the information processing device 100, there is no risk of personal identification information being leaked. Therefore, according to the information processing device 100 having the above configuration, it is possible to preferably generate a medical document MD using the second language model LM2 while avoiding the risk of personal identification information being leaked.

[0104] Furthermore, according to the information processing device 100, before providing the prompt PR to the second language model LM2, the following steps are performed: - Present at least one of the prompt PR, the event data ED, and the visualized data of the event data ED to the user; - Obtain correction instructions from the user, and correct at least one of the prompt PR and the event data ED based on the correction instructions; - Obtain a medical document based on the corrected prompt PR, thereby making it possible to obtain a highly accurate medical document MD.

[0105] Furthermore, the information processing device 100 includes the correction unit 14 that corrects the acquired medical document MD, so that a medical document suitable for the user or subject can be generated.

[0106] As described above, the first language model LM1 is, for example, a model trained by machine learning using training data including text data in the medical field. As described above, the second language model LM2 is, for example, a language model provided in a device external to the information processing device 1, and is a model trained using not only text in the medical field but also text related to multiple other fields. As an example, the second language model LM2 is a language model that is larger in scale (has a larger number of parameters) than the first language model LM1.

[0107] The second language model LM2 described above is trained using a wider range of text than the first language model LM1, and therefore has high versatility and a wide range of knowledge. However, the second language model LM2 requires more time and text for training than the first language model LM1. Furthermore, when it comes to information extraction in a highly specialized field, such as the medical field, the first language model LM1, which is dedicated to that field, is superior even if it is small in scale, and also has the advantage of being easier to train.

[0108] The information processing device 100 according to this exemplary embodiment can enjoy the advantages of both the versatility of the second language model LM2, which has been trained using a wider range of text, and the specialization of the first language model LM1, which has been trained specifically for the medical field, thereby providing a document generation technology with higher accuracy.

[0109] [Example of Software Implementation] Some or all of the functions of the information processing devices 1, 100 (hereinafter also referred to as "each of the above devices") may be implemented by hardware such as an integrated circuit (IC chip), or by software.

[0110] In the latter case, each of the above devices is realized by, for example, a computer that executes instructions of a program, which is software that realizes each function. An example of such a computer (hereinafter referred to as computer C) is shown in Figure 10. Figure 10 is a block diagram showing the hardware configuration of computer C that functions as each of the above devices.

[0111] The computer C includes at least one processor C1 and at least one memory C2. The memory C2 stores a program P for causing the computer C to function as each of the above-mentioned devices. In the computer C, the processor C1 reads and executes the program P from the memory C2, thereby realizing the functions of each of the above-mentioned devices.

[0112] The processor C1 may be, for example, a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), a micro processing unit (MPU), a floating point number processing unit (FPU), a physics processing unit (PPU), a tensor processing unit (TPU), a quantum processor, a microcontroller, or a combination thereof. The memory C2 may be, for example, a flash memory, a hard disk drive (HDD), a solid state drive (SSD), or a combination thereof.

[0113] The computer C may further include a RAM (Random Access Memory) for expanding the program P during execution and for temporarily storing various data. The computer C may also include a communication interface for transmitting and receiving data to and from other devices. The computer C may also include an input / output interface for connecting input / output devices such as a keyboard, a mouse, a display, and a printer.

[0114] The program P can also be recorded on a non-transitory, tangible recording medium M that can be read by the computer C. Such a recording medium M can be, for example, a tape, a disk, a card, a semiconductor memory, or a programmable logic circuit. The computer C can acquire the program P via such a recording medium M. The program P can also be transmitted via a transmission medium. Such a transmission medium can be, for example, a communication network or broadcast waves. The computer C can also acquire the program P via such a transmission medium.

[0115] [Addendum 1 related to this specification] This disclosure includes the technologies described in the following appendices. However, the present invention is not limited to the technologies described in the following appendices, and various modifications are possible within the scope of the claims.

[0116] (Appendix A1) An information processing device comprising: a generation means for generating medical information from input data using a first language model; and an acquisition means for acquiring a medical document generated by a second language model that uses the medical information as at least a part of its input.

[0117] (Appendix A2) The information processing device according to Appendix A1, further comprising a first presentation means for presenting a medical document generated by the second language model.

[0118] (Supplementary Note A3) The information processing device according to Supplementary Note A2, further comprising: a correction unit that corrects the medical document; and the first presentation unit that presents the medical document after the correction.

[0119] (Appendix A4) The information processing device described in Appendix A3, wherein the medical document generated by the second language model or the medical document corrected by the correction means includes at least one of a surgery description, a medical condition description, an inpatient treatment plan, a discharge summary, a referral letter, a medical certificate, a consent form, and a request form.

[0120] (Supplementary Note A5) The information processing device according to Supplementary Note A3 or A4, wherein the correction means optimizes the first language model, which is a machine learning model, by referring to medical documents generated by the second language model.

[0121] (Supplementary Note A6) The information processing device according to any one of Supplementary Notes A1 to A5, wherein the generating means generates a prompt to be input to the second language model using the medical information.

[0122] (Supplementary Note A7) The information processing device according to Supplementary Note A6, further comprising a second presentation means for presenting at least one of the medical information, visualized data of the medical information, and the prompt.

[0123] (Supplementary Note A8) The information processing device according to Supplementary Note A7, wherein the second presentation means presents data including a timeline regarding treatment of the subject as the visualization data of the medical information.

[0124] (Supplementary Note A9) The information processing device described in Supplementary Note A7 or A8, wherein the acquisition means acquires a user's instruction to modify at least one of the medical information, the visualization data of the medical information, and the prompt, and the generation means modifies at least one of the medical information, the visualization data of the medical information, and the prompt in accordance with the user's instruction to modify.

[0125] (Supplementary Note A10) The information processing device according to any one of Supplementary Notes A1 to A9, wherein the process of generating the medical information by the generating means includes a process of generating structured data by referring to unstructured data included in the input data.

[0126] (Supplementary Note A11) The information processing device according to any one of Supplementary Notes A1 to A10, wherein the process of generating the medical information by the generating means includes a process of de-identifying individual identifying information included in the input data.

[0127] (Supplementary Note A12) The information processing device according to Supplementary Note A11, wherein the process of de-identifying the personally identifiable information includes a process of extracting named entities other than the personally identifiable information from among one or more named entities included in the input data.

[0128] (Supplementary Note A13) The information processing device according to Supplementary Note A11, wherein the process of de-identifying the personally identifiable information includes a process of changing the personally identifiable information included in the input data or a process of replacing the personally identifiable information with an identifier.

[0129] (Appendix B1) An information processing method including: generating medical information from input data using a first language model; and obtaining a medical document generated by a second language model that uses the medical information as at least a portion of its input.

[0130] (Supplementary Note B2) The information processing method according to Supplementary Note B1, further comprising presenting medical documents generated by the second language model.

[0131] (Appendix B3) The information processing method according to Appendix B2, further comprising: correcting the medical document; and presenting the medical document after the correction.

[0132] (Appendix B4) The information processing method described in Appendix B3, wherein the medical document generated by the second language model or the corrected medical document includes at least one of a surgery description, a medical condition description, an inpatient treatment plan, a discharge summary, a referral letter, a medical certificate, a consent form, and a request form.

[0133] (Supplementary Note B5) The information processing method according to Supplementary Note B3 or B4, wherein the modifying step optimizes the first language model, which is a machine learning model, by referring to medical documents generated by the second language model.

[0134] (Supplementary Note B6) The information processing method according to any one of Supplementary Notes B1 to B5, wherein the generating of the medical information comprises generating a prompt to be input to the second language model using the medical information.

[0135] (Supplementary Note B7) The information processing method according to Supplementary Note B6, further comprising presenting at least one of the medical information, visualized data of the medical information, and the prompt.

[0136] (Supplementary Note B8) The information processing method according to Supplementary Note B7, wherein data including a timeline regarding treatment of the subject is presented as the visualized data of the medical information.

[0137] (Appendix B9) An information processing method as described in Appendix B7 or B8, wherein the acquiring step acquires a user's correction instruction for at least one of the medical information, the visualization data of the medical information, and the prompt, and the generating step modifies at least one of the medical information, the visualization data of the medical information, and the prompt in accordance with the user's correction instruction.

[0138] (Supplementary Note B10) The information processing method according to any one of Supplementary Notes B1 to B9, wherein the step of generating the medical information includes generating structured data by referring to unstructured data included in the input data.

[0139] (Supplementary Note B11) The information processing method according to any one of Supplementary Notes B1 to B10, wherein the step of generating the medical information includes de-identifying individual identifying information included in the input data.

[0140] (Supplementary Note B12) The information processing method according to Supplementary Note B11, wherein the step of de-identifying the personally identifiable information includes extracting named entities other than the personally identifiable information from among one or more named entities included in the input data.

[0141] (Supplementary Note B13) The information processing method according to Supplementary Note B11, wherein the step of de-identifying the personally identifiable information includes a process of changing the personally identifiable information included in the input data or replacing the personally identifiable information with an identifier.

[0142] (Appendix C1) A program that causes a computer to execute a generation process that generates medical information from input data using a first language model, and an acquisition process that acquires medical documents generated by a second language model that uses the medical information as at least a portion of its input.

[0143] (Supplementary Note C2) The program according to Supplementary Note C1, further comprising: a first presentation process for presenting a medical document generated by the second language model.

[0144] (Supplementary Note C3) The program according to Supplementary Note C2, further comprising: executing a correction process for correcting the medical document; and the first presentation process presents the medical document after the correction.

[0145] (Appendix C4) The medical document generated by the second language model or the medical document corrected by the correction process includes at least one of a surgery description, a medical condition description, an inpatient treatment plan, a discharge summary, a referral letter, a medical certificate, a consent form, and a request form. (Appendix C4) The program described in Appendix C3.

[0146] (Supplementary Note C5) The program according to Supplementary Note C3 or C4, wherein the correction process optimizes the first language model, which is a machine learning model, by referring to medical documents generated by the second language model.

[0147] (Supplementary Note C6) The program according to any one of Supplementary Notes C1 to C5, wherein the generation process generates a prompt to be input to the second language model using the medical information.

[0148] (Supplementary Note C7) The program according to Supplementary Note C6, further executing a second presentation process of presenting at least one of the medical information, visualized data of the medical information, and the prompt.

[0149] (Supplementary Note C8) The program according to Supplementary Note C7, wherein the second presentation process presents data including a timeline regarding treatment of the subject as the visualization data of the medical information.

[0150] (Appendix C9) The program described in Appendix C7 or C8, wherein the acquisition process acquires a user's modification instruction for at least one of the medical information, the visualization data of the medical information, and the prompt, and the generation process modifies at least one of the medical information, the visualization data of the medical information, and the prompt in accordance with the user's modification instruction.

[0151] (Supplementary Note C10) The program according to any one of Supplementary Notes C1 to C9, wherein the process of generating the medical information includes a process of generating structured data by referring to unstructured data included in the input data.

[0152] (Supplementary Note C11) The program according to any one of Supplementary Notes C1 to C10, wherein the process of generating the medical information includes a process of de-identifying personal identification information included in the input data.

[0153] (Supplementary Note C12) The program according to Supplementary Note C11, wherein the process of de-identifying the personally identifiable information includes a process of extracting named entities other than the personally identifiable information from among one or more named entities included in the input data.

[0154] (Appendix A13) The program according to Appendix C11, wherein the process of de-identifying the personally identifiable information includes a process of changing the personally identifiable information included in the input data or a process of replacing the personally identifiable information with an identifier.

[0155] [Addendum 2 related to this specification] This disclosure includes the technologies described in the following appendices. However, the present invention is not limited to the technologies described in the following appendices, and various modifications are possible within the scope of the claims.

[0156] (Appendix D1) An information processing device comprising at least one processor, the at least one processor executing: a generation process for generating medical information from input data using a first language model; and an acquisition process for acquiring medical documents generated by a second language model using the medical information as at least a portion of an input.

[0157] (Supplementary Note D2) The information processing device according to Supplementary Note D1, further comprising: a first presentation process for presenting a medical document generated by the second language model.

[0158] (Supplementary Note D3) The information processing device according to Supplementary Note D2, further comprising: executing a correction process for correcting the medical document; and presenting the medical document after correction in the first presentation process.

[0159] (Appendix D4) The information processing device described in Appendix D3, wherein the medical document generated by the second language model or the medical document corrected by the correction process includes at least one of a surgery description, a medical condition description, an inpatient treatment plan, a discharge summary, a referral letter, a medical certificate, a consent form, and a request form.

[0160] (Supplementary Note D5) The information processing device according to Supplementary Note D3 or D4, wherein the correction process optimizes the first language model, which is a machine learning model, by referring to a medical document generated by the second language model.

[0161] (Supplementary Note D6) The information processing device according to any one of Supplementary Notes D1 to D5, wherein the generation process generates a prompt to be input to the second language model using the medical information.

[0162] (Supplementary Note D7) The information processing device according to Supplementary Note D6, further executing a second presentation process of presenting at least one of the medical information, visualization data of the medical information, and the prompt.

[0163] (Supplementary Note D8) The information processing device according to Supplementary Note D7, wherein the second presentation process presents data including a timeline related to treatment of the subject as the visualization data of the medical information.

[0164] (Appendix D9) The information processing device described in Appendix D7 or D8, wherein the acquisition process acquires a user's correction instruction for at least one of the medical information, the visualization data of the medical information, and the prompt, and the generation process corrects at least one of the medical information, the visualization data of the medical information, and the prompt in accordance with the user's correction instruction.

[0165] (Supplementary Note D10) The information processing device according to any one of Supplementary Notes D1 to D9, wherein the process of generating the medical information includes a process of generating structured data by referring to unstructured data included in the input data.

[0166] (Supplementary Note D11) The information processing device according to any one of Supplementary Notes D1 to D10, wherein the process of generating the medical information includes a process of de-identifying individual identifying information included in the input data.

[0167] (Supplementary Note D12) The information processing device according to Supplementary Note D11, wherein the process of de-identifying the personally identifiable information includes a process of extracting named entities other than the personally identifiable information from among one or more named entities included in the input data.

[0168] REFERENCE SIGNS LIST 1,100... Information processing device 11... Generation unit (generation means) 12... Acquisition unit (acquisition means) 13... First presentation unit (first presentation means) 14... Correction unit (correction means) 15... Second presentation unit (second presentation means)

Claims

1. A generation means that generates medical information from input data using a first language model, Acquisition means for acquiring a medical document generated by a second language model that takes at least some of the aforementioned medical information as input, An information processing device equipped with the following features.

2. The system further comprises a first presentation means for presenting medical documents generated by the second language model. The information processing apparatus according to claim 1.

3. The medical document is provided with means for making corrections, The first presentation means presents the revised medical document. The information processing apparatus according to claim 2.

4. The medical document generated by the second language model, or the medical document modified by the modification means, includes at least one of the following: a surgical explanation document, a medical condition explanation document, an inpatient treatment plan, a discharge summary, a referral letter, a medical certificate, a consent form, or a request form. The information processing apparatus according to claim 3.

5. The modification means optimizes the first language model, which is a machine learning model, by referring to the medical documents generated by the second language model. The information processing apparatus according to claim 3 or 4.

6. The generating means is Using the aforementioned medical information, generate prompts for input to the second language model. The information processing apparatus according to any one of claims 1 to 4.

7. The system further comprises a second presentation means for presenting at least one of the aforementioned medical information, visualization data of the aforementioned medical information, and the aforementioned prompt. The information processing apparatus according to claim 6.

8. The second presentation means presents data including a timeline of the subject's treatment as visualization data of the medical information. The information processing apparatus according to claim 7.

9. Using the first language model, medical information is generated from the input data, To obtain a medical document generated by a second language model that takes the aforementioned medical information as at least some of its inputs, and An information processing method that includes this.

10. On the computer, Using the first language model, a generation process is performed to generate medical information from the input data, An acquisition process for acquiring a medical document generated by a second language model that takes the aforementioned medical information as at least some of the input, and A program that executes the command.