Recording management device and allocation method

Abstract

To allow appropriate data content to be input to appropriate items even if item names are not included.SOLUTION: A recording management terminal 30 receives input of a form name of a recording form having a plurality of data items and data input fields, and a speech including data content to be input to the data input fields, and converts the input speech into a character string to be transmitted to a recording management server 20 (Step 81). The recording management server 20 creates a processing request sentence including the data items in the recording form specified by the recording form name, the data content included in the character string, and an allocation instruction that associates the data content with the data items, and transmits the processing request sentence to a generative AI (Step 85). The generative AI executes allocation processing to allocate the data content included in the speech into data items according to the content (Step 88).SELECTED DRAWING: Figure 11A

Description

[Technical field]

[0001] The present invention relates to a record management device and a distribution method. [Background technology]

[0002] Patent Document 1 discloses a device that automatically inputs values ​​into each item of a form on a nursing care record input screen using text data obtained by voice recognition. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2017 / 010506

[0004] In the device described in Patent Document 1, templates are prepared in advance for extracting numerical values ​​and the like from text data that has been subjected to voice recognition. If there is a portion in the text data that matches the template, it is extracted from the text data and input as the value of the corresponding item on the nursing record input screen. The template includes predetermined terms such as "body temperature" and "blood pressure," and the item is identified by recognizing this predetermined term.

[0005] It is possible to input the desired data content (a numerical value representing body temperature) into the data input field of the desired item (for example, a body temperature input field) by first vocalizing the item name and then vocalizing the data content of that item. However, if a recording form contains multiple items (data input fields), vocalizing the item names and data content for all of them would take a long time to record the data, making on-site operations more difficult. Summary of the Invention [Problem to be solved by the invention]

[0006] An object of the present invention is to enable easy input of appropriate data content into data input fields for appropriate items.

[0007] In particular, the object of this invention is to enable appropriate data contents to be input into appropriate items simply by speaking, without having to take into consideration the order of speaking or the pronunciation of item names.

[0008] The data input support device of the present invention comprises a microphone that accepts input of a speech including a form name of a record form having a plurality of record items (names) and a data input field for each record item, and record contents to be entered into the data input fields, a voice recognition means that converts the speech including the form name and record contents received by the microphone into a character string, and a processing request statement including a plurality of record items (names) and record contents included in the character string (which may be the character string itself) of a record form identified by a record form name included in the character string converted by the voice recognition means, and an allocation command that associates the record contents with the record items, the processing request statement being executed in accordance with the processing request statement in a generation system AI (generation AI) that is given a receiving means that receives a result of an allocation process in which the record contents included in the speech are allocated to record items according to their contents.

[0009] The data input assistance method of the present invention receives, by a microphone, an input of an utterance including the form name of a record form having a plurality of record items and an input field for each record item, and record content to be entered into the input fields, by a voice recognition means, converting the utterance including the record form name and record content received by the microphone into a character string, and by a receiving means receiving a result of an allocation process in which the record content included in the utterance, which is executed in accordance with the processing request statement, is allocated to record items according to its content in a generative AI to which is given a processing request statement including a plurality of record items included in a record form identified by the record form name included in the character string converted by the voice recognition means, the record content included in the character string, and an allocation command that matches the record content to the record items.

[0010] According to this invention, the recording contents (data contents) contained in the utterance are sorted (associated) with the recording items according to their contents, so that the recording contents (data contents) suitable for the recording items can be input into the data input field for each recording item provided in the recording form. Since there is no need to repeatedly select a recording item (or its data input field) using an input device (keyboard, etc.) and input data into the data input field of the selected recording item, the efficiency of the clerical work of recording is improved.

[0011] That is, since the recording form includes multiple recording items and a data input field for each recording item, in order to input the appropriate recording contents into the appropriate recording item, a specific recording item must be selected from the multiple recording items and the recording contents (e.g., numerical values) must be input into the data input field of the selected recording item. When data is input using an input device (mouse, keyboard, etc.), a specific recording item (or its data input field) is selected in advance by, for example, clicking, and the recording contents (e.g., numerical values) are input in that state. Selection of a specific recording item (selection of a data input field) is performed with one click, but if the recording form includes many recording items (data input fields), the number of clicks will increase. In the case of voice input, it is possible to input the desired recording contents into the data input field of the desired recording item by vocalizing the name of the recording item in advance and then vocalizing the recording contents. However, if the recording form includes many recording items (data input fields), it still takes time to vocalize the names of the recording items for all of them, which increases the difficulty of operations at the data input site.

[0012] According to this invention, a generative AI is used to realize the matching (sorting process) between a desired record item and the record contents without necessarily having to vocalize the record item (name). This invention operates the generative AI as a machine that executes the process (sorting process) of matching the record contents with the record items, and for this, ChatGPT can be used, for example.

[0013] A processing request statement is sent to the generative AI, which includes multiple record items (names), record contents (e.g., numerical values ​​to be entered into the data input fields for each record item), and a distribution processing command.

[0014] According to this invention, when a record form name is received by the microphone, a processing request statement including a plurality of record items (names) that the record form specified by the record form name has is provided to the generative AI. Similarly, the record items (names) (candidates) to which the record contents included in the processing request statement should be allocated are accurately conveyed to the generative AI, so that the accuracy of matching (allocation) between the record items and the record contents can be improved.

[0015] When the sorting process was actually performed on Chat GPT, it was confirmed that the desired record contents were matched (sorted) to the desired record items with relatively high accuracy, even when the utterance contained only the record contents and did not contain the corresponding record items (names). The sorting process was performed with high accuracy even for spoken language, significantly improving the administrative efficiency of data entry. It has been proposed that Chat GPT can be used for text correction and proofreading, summarizing texts and concepts, brainstorming, research and identifying issues, proposing ideas, programming, English-to-Japanese translation, Japanese-to-English translation, and composition, but it has not been confirmed that Chat GPT has been proposed for use in sorting to match appropriate record contents to appropriate record items.

[0016] The speech recognition process for converting the speech into a string of characters is performed in a record management terminal equipped with a microphone. For example, a virtual assistant (Siri, Google Now, etc.) included in the operating system of the record management terminal can be used for the string conversion (transcription).

[0017] The above-mentioned record forms preferably include forms for nursing care records or support records for elderly people or disabled people, and forms for personal records for elderly people or disabled people themselves. Nursing care records or support records are created, or personal records are created, using various types of record forms. By improving the efficiency of record-keeping work, time that was previously spent on nursing care records, support records, and personal records can be allocated to nursing care itself or support itself.

[0018] In one embodiment, the apparatus includes a means for displaying a data-entered record form in which record contents are entered into data entry fields of record items of the record form according to the results of the sorting process received by the receiving means. The data-entered record form can be displayed as if the data had been entered using an input device, without necessarily using an input device.

[0019] In one embodiment, the character string based on the utterance does not include a character string related to the record item for at least one of the multiple record items included in the record form, but includes a character string related to the record content corresponding to the record item. In other words, a character string (transcription result) when the name of at least one of the multiple record items included in the record form is not uttered is given to the generative AI. Since the processing request sentence given to the generative AI includes the multiple record items provided in the record form, the generative AI can accurately grasp the record item to which the record content should be allocated, even if it is not included in the utterance. Therefore, even if the utterance includes record content for which the record item is not specified, the generative AI can execute an allocation process that matches the appropriate record content to the appropriate record item. Of course, for some of the multiple record items included in the record form, not only the record content but also the record item (name) may be included in the utterance. For example, instead of uttering "36 degrees," utter "body temperature 36 degrees" ("body temperature" is the record item and "36 degrees" is the record content). This will improve the accuracy of the generative AI processing that distributes (matches) record contents to record items.

[0020] In addition, if a unit (for example, in the above case, "degrees" representing temperature) is included, the accuracy of the generative AI processing that matches the recorded contents to the recorded items may be improved. For some of the multiple recorded items, units may be included in the utterance.

[0021] In one embodiment, the character string based on the utterance is a plurality of record contents arranged in an order different from the order of the plurality of record items included in the record form. It has been confirmed that the generative AI can assign appropriate record contents to appropriate record items even if the record contents are not uttered in the same order as the plurality of record items included in the record form.

[0022] In one embodiment, the generative AI executes an inference process to infer record contents corresponding to the record items included in the record form from the context of the character string based on the utterance, and executes a process to allocate the inferred record contents to record items corresponding to the contents. One example is the number of visitors (e.g., 2 people) when the utterance includes the blood relationship of the visitor (e.g., the eldest son and his wife). It has been confirmed that even if the number of visitors is included as a record item but is not spoken, the generative AI can accurately infer that the record content corresponding to the record item "number of visitors" is "2 people" because the utterance includes the fact that the visitors were the eldest son and his wife. By using the generative AI, the inference results by the generative AI can be allocated to record items.

[0023] The data content voice input system according to the present invention comprises a record management terminal (client) and a record management server, the record management terminal comprises a microphone for receiving an input of a speech including a form name of a record form having a plurality of data items (record items) and a data input field for each data item, and data content (record content) to be input into the data input field, a speech recognition means for converting the speech including the record form name and data content received by the microphone into a character string, and a transmission means for transmitting the character string converted by the speech recognition means to the record management server, and the record management server comprises a receiving means for receiving the character string transmitted by the transmission means of the record management terminal, a process for generating a processing request statement including a plurality of data items provided in a record form specified by the record form name contained in the character string received by the receiving means, the data content contained in the character string (which may be the character string itself), and a distribution command for associating the data content with the data items. The system comprises a request statement creation means, a generation AI utilization means for sending a processing request statement created by the processing request statement creation means, the processing request statement including the data items, data contents and allocation commands, to a generation AI, and receiving a result of an allocation process executed by the generation AI in accordance with the processing request statement, in which the data contents included in the utterance are allocated to data items according to their contents, and a transmission means for transmitting the allocation process result received by the generation AI utilization means and form data representing the record form to the record management terminal, and the record management terminal further comprises a creation means for creating data representing a data-entered record form in which the data contents included in the utterance are allocated and input into data input fields of data items according to their contents in accordance with the allocation process result and the form data representing the record form sent from the transmission means of the record management server, and a display means for displaying the data-entered record form in accordance with the data representing the data-entered record form created by the creation means.

[0024] The data representing the record form with data entered may be created by the record management terminal or by the record management server. When created by the record management terminal, the distribution process result and form data representing the record form are sent from the record management server to the record management terminal. When created by the record management server, data representing the record form with data entered is sent from the record management server to the record management terminal.

[0025] The data content is input into the data input field of the record form according to the speech received by the microphone equipped in the record management terminal. It is not necessary to use an input device (keyboard, etc.), and the efficiency of record clerical work is improved.

[0026] When a record form name that specifies a record form with multiple data items is sent from the record management terminal to the record management server, the record management server includes the multiple data items that the record form specified by the record form name has in the processing request statement. Since the data items (candidates) to which the data content in the utterance should be allocated are accurately conveyed to the generative AI, even if the data item (name) is not included in the utterance, the generative AI can accurately match (allocate) the data items with the appropriate data content.

[0027] The present invention also provides a record management device. The record management device according to the present invention includes a processing request statement creation means for creating a processing request statement including a plurality of data items provided in a record form specified by a record form name included in a character string including a form name of a record form having a plurality of data items (record items) and a data input field for each data item, and a distribution command for associating the data items with the data contents, according to a character string including the form name of a record form provided in accordance with an utterance, the form name including the form name of a record form having a plurality of data items (record items) and a data input field for each data item, and the data contents to be input into the data input fields, and a generation system AI utilization means for transmitting the processing request statement including the data items, data contents, and distribution command created by the processing request statement creation means to a generation system AI, and receiving the result of a distribution process executed by the generation system AI according to the processing request statement, in which the data contents included in the utterance are distributed to data items according to their contents. The result of the distribution process can be used to create data representing a record form with data input. The result of the distribution process may be stored in a storage device as it is, or may be displayed on a screen of a record management terminal before storage, and may be stored in a storage device after confirmation for the presence of errors (after correcting any errors).

[0028] The present invention also provides a data input method. In the data input method according to the present invention, a processing request statement creation means creates a processing request statement including a plurality of data items provided in a record form identified by a record form name included in a character string including a plurality of data items and a data input field for each data item, and data contents to be input into the data input fields, according to the utterance, the plurality of data items provided in the record form identified by the record form name included in the character string, the data contents included in the character string, and a distribution command for associating the data contents with the data items, and a generative AI utilization means transmits the processing request statement created by the processing request statement creation means, including the data items, data contents, and distribution command, to a generative AI, and receives a result of a distribution process executed by the generative AI according to the processing request statement, in which the data contents included in the utterance are distributed to data items according to their contents.

[0029] In one embodiment, the character string is the result of speech-based voice recognition processing.

[0030] Preferably, a record form master database in which the record form name and a plurality of data items (record items) (names) included in the record form specified by the record form name are stored in association with each other is connected to the record management device. As described above, the data items to which the data contents should be allocated can be accurately conveyed to the generation system AI, and the accuracy of the allocation process can be improved.

[0031] In one embodiment, the processing request statement includes a command statement instructing the generative AI to operate as a machine that distributes the data contents according to the data items.

[0032] In another embodiment, the processing request statement includes a constraint statement that constrains the operation of the generative AI.

[0033] In another embodiment, the processing request statement includes a command statement instructing the generative AI to support multiple languages.

[0034] The purpose of using a generative AI is to have it execute a process of allocating (associating) appropriate data contents with appropriate data items without detailed instructions (commands). The accuracy of the allocation process can be improved by including in the processing request statement given to the generative AI a command statement that commands it to act as a machine that allocates data contents for each data item. In addition, by including in the processing request statement a constraint statement that constrains the operation of the generative AI, the processing results of the generative AI can be made to conform to the data format that represents the data-entered record form, which is the final deliverable that uses the processing results of the generative AI. Furthermore, by issuing a command for multilingual support, the allocation process can be executed with high accuracy even if the utterance (character strings given according to the utterance) includes a foreign language or dialect.

[0035] Preferably, the device is provided with a control means for causing the record management device to function as the processing request statement creation means and generative AI utilization means when a given character string contains a specific keyword. The specific keyword can cause the record management device to function as a processing device for another task. For example, if the specific keyword "data save" is included in the utterance (character string), the record management device can be operated to utilize the generative AI described above. If another specific keyword, such as "user change", is included, the record management device can be operated to perform an operation different from the operation using the generative AI, such as a user change process (a process for changing the target user of a record form in which data has been entered).

[0036] The present invention also provides an accident report creation support device. The accident report creation support device according to the present invention is an accident report creation support device that supports data input into each data input field of an accident report record form that has a plurality of record items related to an accident and has a data input field for each accident record item, and includes a microphone that receives an input of an utterance related to the accident content to be input into each data input field of the accident report record form, a voice recognition means that converts the utterance related to the accident content received by the microphone into a character string, and a processing request statement including the accident record items provided in the accident report record form and the accident content (which may be the character string itself) included in the character string, as well as a processing request statement including a sorting command that associates the accident content with the accident record items, the processing request statement being executed according to the processing request statement, and that sorts the accident content included in the utterance to the accident record items according to the content. The accident record items include, for example, the date of the accident, the time of the accident, the location of the accident, the type of the accident, and the circumstances of the accident. An accident report is created when a user falls down the stairs and gets injured in a care facility.

[0037] A processing request text given to the generative AI includes multiple accident record items (names) (e.g., accident date, accident time, accident location, accident type, and accident occurrence circumstances) provided in the accident report record form. Similarly, the accident record item (name) to which the accident content based on the utterance included in the processing request text should be allocated is accurately conveyed to the generative AI, so the accuracy of matching (allocation) between the accident record items and the accident record contents can be improved. Since the accident record contents included in the utterance are allocated (associated) to the accident record items according to their contents, the record contents (data contents) suitable for the record items can be input into the data input field for each accident record item provided in the accident report record form. Since there is no need to repeatedly select an accident record item (or its data input field) using an input device (keyboard, etc.) and then input data into the data input field for the selected accident record item, the administrative work of creating an accident report can be made more efficient.

[0038] The present invention further provides a visit report creation support device. The visit report creation support device according to the present invention supports data input into each data input field of a visit report record form having a plurality of record items related to visits and a data input field for each visit record item, and includes a microphone for receiving an input of an utterance related to the visit to be input into each data input field of the visit report record form, a voice recognition means for converting the utterance related to the visit received by the microphone into a character string, and a generative AI to which a processing request statement including the visit record items provided in the visit report record form, the visit contents included in the character string, and a distribution command for associating the visit contents with the visit record items is given, the processing request statement being executed according to the processing request statement, and which distributes the visit contents included in the utterance to the visit record items according to the contents. For example, a visit report is created when a relative of a user of a nursing facility comes to visit the nursing facility. The visit record items included in the visit report record form include, for example, the visitor, the relationship, the situation, and the number of visitors.

[0039] The processing request text given to the generative AI includes multiple visitation record items (names) (for example, visitor, relationship, situation, and number of visitors) provided in the above-mentioned visitation report record form. Similarly, the visitation record items (names) to which the visitation content should be allocated based on the utterance included in the processing request text are accurately conveyed to the generative AI, so the accuracy of matching (allocation) the visitation record items and the visitation record contents can be improved. There is no need to repeatedly use an input device to select a visitation record item (or its data input field) and enter data into the data input field of the selected visitation record item, so the administrative work of creating a visitation report is made more efficient. [Brief description of the drawings]

[0040] [Figure 1A] 1 is a block diagram showing a hardware configuration of a voice input system for nursing care details. [Figure 1B] FIG. 13 is a block diagram showing a hardware configuration of another aspect of a voice input system for care content. [Diagram 2] 2 is a block diagram showing the hardware configuration of a recording management terminal. FIG. [Diagram 3] 1 shows a health care record form. [Figure 4] Shows the user master database. [Diagram 5] Showing record form master database. [Figure 6] Shows record table database. [Figure 7] 4 is a flowchart showing a process of the voice input system. [Figure 8] 13 shows a screen for inputting care details. [Figure 9A] 13 is a flowchart showing the processing of the voice input system in a user selection phase. [Figure 9B] 13 shows an example of data transmitted and received in the user selection phase. [Figure 10] 13 shows a screen for inputting care details. [Figure 11A] 13 is a flowchart showing the processing of the voice input system in the record creation phase. [Figure 11B] 13 shows an example of data sent and received in the record creation phase. [Figure 12] 13 shows a screen for inputting care details. [Figure 13] An example of the speech content and the result of the sorting process by the item sorting server are shown below. [Figure 14] An example of the speech content and the result of the sorting process by the item sorting server are shown below. [Figure 15] An example of the speech content and the result of the sorting process by the item sorting server are shown below. [Figure 16] An example of the speech content and the result of the sorting process by the item sorting server are shown below. [Figure 17] An example of the speech content and the result of the sorting process by the item sorting server are shown below. [Figure 18] An example of the speech content and the result of the sorting process by the item sorting server are shown below. [Figure 19] An example of the speech content and the result of the sorting process by the item sorting server are shown below. [Figure 20] An example of the speech content and the result of the sorting process by the item sorting server are shown below. [Figure 21] An example of the speech content and the result of the sorting process by the item sorting server are shown below. [Figure 22] An example of the speech content and the result of the sorting process by the item sorting server are shown below. [Diagram 23] An example of the speech content and the result of the sorting process by the item sorting server are shown below. [Figure 24] An example of the speech content and the result of the sorting process by the item sorting server are shown below. [Diagram 25] An example of the speech content and the result of the sorting process by the item sorting server are shown below. [Figure 26] An example of the speech content and the result of the sorting process by the item sorting server are shown below. EXAMPLES

[0041] 1. Hardware Configuration FIG. 1A is a block diagram showing the hardware configuration of a voice input system for nursing care contents (service contents) (hereinafter referred to as the voice input system).

[0042] The voice input system includes a database server 10, a record management server 20, a record management terminal 30, a morphological analysis server 40, and an item allocation server 50. Databases (user master database 11, record form master database 12, record table database 13) that store various data required for the operation of the voice input system are connected to the database server 10. Other databases can also be connected to the database server 10 as necessary.

[0043] The voice input system inputs various types of care (assistance) provided by care facility staff to users (typically elderly people or people with disabilities) and records (saves) the input care content (care records). The care records are used to consider care plans for users of the care facility, to utilize them in high-quality care services and care management, to improve the qualifications of care facility staff, and to share information.

[0044] There are multiple nursing care facilities, each of which generally has multiple users. Care is provided to each of the multiple users at the multiple nursing care facilities on a daily basis, and care records are kept on a daily basis. Care record keeping is an administrative task that is often performed by the staff of the nursing care facility providing the care, and is a separate task from the care itself. The voice input system reduces the administrative burden on the staff of the nursing care facility who must keep care records in addition to providing care.

[0045] In this embodiment, the care contents inputted and recorded using the voice input system include many items (care items), such as the user's vital data (body temperature, blood pressure, transcutaneous arterial oxygen saturation, etc.), dietary contents (amount of food), oral care, height, weight, exercise, visitors, and recreational contents. The care contents are inputted and recorded (saved) for each user. In this embodiment, the care contents (care records) include not only details regarding personal care for the user, but also the user's own personal records (the above-mentioned body temperature, height, weight, etc.), reports, etc.

[0046] Referring to Fig. 1, generally, multiple record management terminals 30 (or just one) are used to input care contents (care records). Details will be described later, but the voice input system of this embodiment can accept input (including selection) of care contents by speech (voice) from staff of a care facility, etc., instead of or in addition to input of care contents by character (text) input from a keyboard (including selection of characters by clicking with a mouse). However, even when care contents are input by voice, the care contents to be recorded (saved) are character (text) data, not voice data. To accommodate voice input, the record management terminal 30 is equipped with a voice recognition library 31. Processing of the voice recognition library 31 will be described later.

[0047] The voice input system can be composed of the above-mentioned record management terminal 30, record management server 20, and database server 10. To enable voice input, a morphological analysis server 40 and an item sorting server 50 are further used. These servers 10, 20, 40, 50 are directly or indirectly connected to the record management terminal 30 via a network such as the Internet, and the record management terminal 30 can receive the processing results. Details of the processing of the record management terminal 30 and the servers 10, 20, 40, 50 will be described later.

[0048] FIG. 1B shows a hardware configuration of a voice input system according to another embodiment.

[0049] The voice input system shown in Fig. 1B integrates the functions of the database server 10 and record management server 20 shown in Fig. 1A into a record management terminal 30. The user master database 11, record form master database 12, and record table database 13 may be stored in the storage device of the record management terminal 30, or may be connected to the record management terminal 30 via a network such as the Internet.

[0050] In the following embodiment, the voice input system shown in Fig. 1A will be described. In the voice input system shown in Fig. 1B, the processes of the database server 10 and the record management server 20 shown in Fig. 1A are executed in the record management terminal 30.

[0051] Fig. 2 is a block diagram showing the hardware configuration of the record management terminal 30. Fig. 3 shows an example of a record form (health care record form 32) for inputting (recording) care contents. Multiple types of record forms are prepared, and care contents are input according to care items (service items) defined in each record form.

[0052] 2, the record management terminal 30 can be used as an operation terminal for inputting care contents by voice. The record management terminal 30 includes a microphone 30G and a display device 30J connected to a data bus 30A. Care contents are input into a record form displayed on the display device 30J according to speech by staff of the care facility, etc., picked up by the microphone 30G. The record form has data input fields for a plurality of care items (service items) that differ for each record form, and care contents (numeric values, characters) are input into each data input field.

[0053] A health care record form 32, which is an example of a record form shown in Fig. 3, includes six care items (service items) of "body temperature," "abnormal / innormal blood pressure," "pulse," "SPO2," "breathing," and "condition observation," and has data input fields 32a-32g for each of the six care items (abnormal / innormal blood pressure). Although details will be described later, the voice input system operates so that appropriate care contents can be entered into the data input field for the appropriate care item based on the utterance even if the utterance by the staff of the care facility includes words unrelated to the care items included in the record form, even if the order in which the care contents (numbers, letters) are spoken is different from the order of the care items in the record form, or even if the name of the care item is not included in the utterance.

[0054] Returning to Fig. 2, in addition to the microphone 30G and display device 30J described above, the data bus 30A is also connected to a CPU 30B that controls the overall operation of the record management terminal 30, a RAM 30C that temporarily stores data and programs, a storage (SSD, HDD, etc.) 30D that stores an operating system, various application programs, etc., an input device (keyboard, etc.) 30E for inputting characters, numbers, etc., a communication device 30F that transmits and receives data via a network, a speaker 30H that outputs sound, and a camera 30I that captures an image of a subject. A portable computer device that can be connected to a network can be used as the record management terminal 30.

[0055] The database server 10, record management server 20, morphological analysis server 40 and item allocation server 50 are also computers equipped with a CPU, RAM, storage, communication devices, etc. However, unlike the record management terminal 30, these servers 10, 20, 40, 50 are not computer devices directly operated by staff of the care facility, etc., and therefore do not necessarily require input devices such as a keyboard, a display device, a microphone, a speaker, or a camera.

[0056] 2. Database Figures 4 to 6 respectively show the contents of the user master database (hereinafter referred to as user master DB) 11, record form master database (hereinafter referred to as record form master DB) 12, and record table database (hereinafter referred to as record table DB) 13 connected to the database server 10.

[0057] 4, the user master DB11 includes data such as a "corporate ID" for identifying the operating corporation of the nursing care facility, a "user ID" for identifying the user of the nursing care facility, the "name" (written in kanji) and "gender" of the user of the nursing care facility specified by the user ID, the "reading" of the name of the user of the nursing care facility (written in katakana), and "reading candidates" (written in katakana) of the name of the user of the nursing care facility. The corporate ID, business operator ID, user ID, name, gender, reading and reading candidates, etc. (records) are stored in the user master DB11 in association with each other.

[0058] "Pronunciation candidates" refers to candidates for the pronunciation of a user's name, assumed from the "name" written in kanji. As described below, in the voice input system, voice is also used when identifying (selecting) the user for whom care details should be recorded. By preparing other pronunciation candidates in addition to the correct pronunciation of a name in the user master DB 11, it is possible to avoid situations where the user cannot be identified if the user's name written in kanji is not pronounced correctly (not converted to kana correctly).

[0059] Referring to Figure 5, the record form master DB12 includes a "corporate ID," a "record form ID" for identifying the record form, the "form name" of the record form (written in kanji), "reading candidates" for the record form name (written in katakana), and the "care item name" (service item name) included in the record form.

[0060] "Reading candidates" are reading candidates for identifying a record form. With reference to FIG. 5, for example, reading candidates for a "health care record form" with form ID "010011" include "kenkou" and "vital." As will be described later, when "kenkou" or "vital" is included in the speech of a staff member or the like at a care facility, the voice input system can be operated so that the "health care record form" with form ID "010011" is identified (selected). The care items included in each of the multiple types of record forms differ for each type of record form, and their names are specified in the "care item name."

[0061] 6, record table DB13 includes a "corporate ID", an "enterprise ID" for identifying a care provider operated by an operating corporation specified by the corporate ID, a "user ID", a "form ID", a "recording date and time", "care content", etc. The care content entered in the data input field for each care item in the above-mentioned record form (e.g., health management record form 32) is stored (saved) in record table DB13 together with the recording date and time, etc.

[0062] 3. Voice input system processing 3-1. Phase selection process The voice input system performs two main operations: a "user selection phase" and a "record creation phase." Below, we will explain in detail the operation of the voice input system in the "user selection phase" and the operation of the voice input system in the "record creation phase."

[0063] 7 is a flow chart showing the flow of a phase selection process for determining whether the voice input system is to operate in the "user selection phase" or the "record creation phase." The phase selection process is executed by the record management terminal 30 that constitutes the voice input system.

[0064] When an application program (client program) for operating a computer device as the record management terminal 30 is started (step 60), a care content input screen (care record screen) is displayed on the display screen of the display device 30J of the record management terminal 30.

[0065] FIG. 8 shows an example of the care content input screen 33 at the time of startup.

[0066] The care content input screen 33 includes a provider selection combo 33a, a user selection list 33b, a record form selection combo 33c, a record write button (save button) 33d, a record form display area 33e, a saved care record list display area 33f, a speech content text display area 33g, a start button 33h and an end button 33i.

[0067] When a business operator (nursing facility) is selected using the business operator selection combo 33a, the names of users who use the business operator are displayed in the user selection list 33b. FIG. 8 shows the state in which the name of the user who uses the business operator "Nursing facility Shizuoka" is displayed in the user selection list 33b as a result of the business operator "Nursing facility Shizuoka" being selected. When another business operator is selected using the business operator selection combo 33a, the names of users who use the selected business operator are displayed in the user selection list 33b. The above-mentioned user master DB 11 (FIG. 4) and the like are used to display the names of the business operators displayed in the business operator selection combo 33a and the user names in the user selection list 33b. That is, when an application program is started in the record management terminal 30, the record management terminal 30 and the record management server 20 are connected by a network (of course, an ID, a password, and the like may be requested). The record management server 20, which has received a request from the record management terminal 30, instructs the database server 10 to read data according to the request. According to the user master DB 11 etc., the business operator names are displayed in the business operator selection combo 33a and the names of users who use the selected business operator are displayed in the user selection list 33b so that they can be selected.

[0068] In the "user selection phase", a user for whom care details should be entered is selected, and in the next "record creation phase", the care details for the user selected in the "user selection phase" are entered. In order to prevent the system from proceeding to the record creation phase without a user being selected, the system may be configured to select a user manually (by clicking with a mouse) rather than by voice input only at the time of startup. Of course, the system may also be configured to select a user for whom care details should be entered by voice input from the time of startup.

[0069] When the start button 33h provided on the care content input screen 33 is clicked, the microphone 30G of the record management terminal 30 receives the speech of the staff of the care facility and inputs it to the record management terminal 30 (step 61).

[0070] The voice data received from the microphone 30G is sent to the voice recognition library 31 (see FIG. 1) included in the record management terminal 30, where it is converted into text (transcribed) (step 62). The voice recognition library 31 can use a virtual assistant (Siri, Google Now, etc.) included in the operating system of the record management terminal 30. The voice recognition library 31 converts the spoken content into a string of characters including Chinese characters.

[0071] The utterances input by voice to the record management terminal 30 include either the keyword "change user" or "save data." In addition, the utterances including the keyword "change user" further include the name (user name), and the utterances including the keyword "save data" include the name of the record form at the beginning of the utterance and also include the care content.

[0072] When the start button 33h is clicked, the record management terminal 30 determines whether the character string converted into text by the voice recognition library 31 contains a specific keyword (the above-mentioned ``user change'' or ``data save'') (step 63).

[0073] For example, if "Shizuoka Taro-san ryosha henkou" (Mr. Shizuoka Ichiro, user change) is spoken and converted into text exactly as it is by the voice recognition library 31, the record management terminal 30 (application program) recognizes the keyword "user change" and proceeds to the user selection phase (user change in step 63, step 64). For example, if "health management, 36 degrees and 5 minutes, data save" (health management, 36 degrees and 5 minutes, data save) is spoken and converted into text exactly as it is, the record management terminal 30 recognizes the keyword "data save" and proceeds to the record creation phase ("data save" in step 63, step 65). The keyword for proceeding to the user selection phase and the keyword for proceeding to the record creation phase can be set arbitrarily and can be programmed in the application program in advance. By recognizing the keyword "user change" or "data save", the record management terminal 30 can switch between the user selection phase and the record creation phase.

[0074] If neither of the two keywords "change user" nor "save data" is included in the utterance (including the case where voice recognition is not possible), the record management terminal 30 may display a message such as "Please say it again" on the display screen of the display device 30J or output it as voice from the speaker 30H to prompt the staff of the care facility to include one of the two keywords in the utterance (to be voice recognized). The same applies when both "change user" and "save data" are included in the utterance.

[0075] 3-2. User selection phase Fig. 9A shows the flow of processing of the voice input system in the user selection phase, and Fig. 9B shows an example of data transmitted and received in the user selection phase.

[0076] When proceeding to the user selection phase, as described above, the name (user name) and the keyword "change user" are spoken and converted into text by the record management terminal 30. The record management terminal 30 transmits the text data of the spoken content to the record management server 20 (step 71, data example 71A).

[0077] The record management server 20 creates an analysis request including the received user name and sends it to the morphological analysis server 40 (step 73, data example 73A). The analysis request text created in the record management server 20 and sent to the morphological analysis server 40 has the following content, for example:

[0078] {“message”:[{“role”:“system”,“content”:“Please read the names of the people below in kana.\r\nShizuoka Ichiro-san, User change\n"}],"max_tokens":600,"temperature":0.1,"top_p":0.1,"frequency_penalty":0,"presence_penalty":0}

[0079] This analysis request is a command to send back the reading (in kana) of the name (user name) contained in the character string “Shizuoka Ichiro-san, user change” that was converted from speech to text by the record management terminal 30 .

[0080] The morphological analysis server 40, which has received the analysis request text including the user name, breaks down the character string (text to be analyzed) "Shizuoka Ichiro-san, user change" contained in the analysis request text into morphemes, and sends back to the record management server 20 the pronunciation of those nouns obtained by the breakdown that are considered to be personal names (step 75, data example 75A). The morphological analysis server 40 has a library of personal names and their pronunciations, and uses this to select (extract) personal names and determine the pronunciation of the selected personal names.

[0081] The record management server 20, which has received the reading data (e.g., "Shizuoka Ichiro") of a user name (e.g., "Shizuoka Ichiro"), instructs the database server 10 to search for a record having the received reading data. The database server 10 refers to the user master DB 11 (Fig. 4) and searches for a record having the received reading data. When the record is found by the search, it transmits the user ID stored in that record in association with the reading data to the record management terminal 30 (step 74, data example 74A).

[0082] A user is specified (selected) by the user ID. When the record management terminal 30 receives the user ID, it displays a care content input screen for the user specified by the received user ID (step 72, data example 72A).

[0083] FIG. 10 shows the care content input screen 33 of the record management terminal 30 after the user ID has been received. It differs from the care content input screen 33 immediately after startup shown in FIG. 8 in that the name of the user identified by the utterance (the user having the received user ID, Shizuoka Ichiro) is highlighted in the user selection list 33b, and the saved care record list display area 33f displays the care records that have been performed in the past (or only the most recent) for the identified user (Shizuoka Ichiro). The care records stored in the record table DB13 (FIG. 6) are used for the past care records displayed in the care record list display area 33f. It can be seen that the care content input screen 33 has become that for the user "Shizuoka Ichiro" by the user name that is highlighted in the user selection list 33b.

[0084] 3-3. Record-making phase Fig. 11A shows the process flow of the voice input system in the record creation phase, and Fig. 11B shows an example of data sent and received in the record creation phase.

[0085] As described above, the record management terminal 30 on which the application program is running proceeds to the record creation phase if the character string converted by the speech recognition library 31 contains the keyword "data storage."

[0086] When proceeding to the record creation phase, as described above, the record form name, care content, and keyword "data storage" at the beginning of the utterance are spoken and converted into text by the record management terminal 30. The record management terminal 30 transmits the text data including the record form name and care content to the record management server 20 together with the user ID acquired in the above-mentioned user selection phase (step 81, data example 81A).

[0087] The record management server 20 creates a processing request including the received record form name and care content (full text of the utterance) and sends it to the item allocation server 50 (step 83, data example 83A). The processing request created by the record management server 20 in step 83 and sent to the item allocation server 50 has the following content, for example:

[0088] {"message":[{"role":"system","content":"Please perform morphological analysis on the following sentence and return only the first word. No punctuation is required.\r\nHealth management, 36 degrees, 98%, complexion is good, and I am in good health. Save data\r\n"}],"max_tokens":800,"temperature":0,"top_p":0.1,"frequency_penalty":0,"presence_penalty":0}

[0089] This processing request is a command to send back the first word contained in the character string "Health management, 36 degrees, 98%, complexion is good, physical condition is good. Data save" which has been converted from speech to text by the record management terminal 30.

[0090] Unlike the analysis request sent to the morphological analysis server 40, the processing request sent to the item sorting server 50 does not include the name of the user of the care facility, etc. By not sending both the user name and the care content to the same server (by not sending the user name to the item sorting server 50), the user's personal information can be protected.

[0091] The item sorting server 50, which has received the processing request text including the received record form name and care content, breaks down the character string contained in the processing request text (text to be analyzed), in the above case "Health management, 36 degrees, 98%, complexion is good, physical condition is good. Data save" into morphemes, selects the first word in the words obtained by the breakdown, and sends the selected word back to the record management server 20 (step 87, data example 87A). If the text to be analyzed is "Health management, 36 degrees, 98%, complexion is good, physical condition is good. Data save", the record form name "Health management" at the beginning of the utterance is sent back to the record management server 20.

[0092] The record management server 20, which has received the record form name, instructs the database server 10 to search for a record having the received record form name. The database server 10 refers to the record form master DB 12 (FIG. 5) and searches for a record having the received record form name. When a record is found, it obtains the form ID contained in that record (step 84). If the received record form name is "health management", the form ID "010011" is obtained (data example 84A).

[0093] The record form master DB12 (FIG. 5) is associated with the form ID and associated with the care item name (service item name) included in the record form identified by the form ID. The record management server 20 creates a new processing request statement including the care item name and care content associated with the acquired form ID, and transmits it to the item allocation server 50 (step 85, data example 85A). The processing request statement created in the record management server 20 in step 85 and transmitted to the item allocation server 50 has, for example, the following content:

[0094] {“messages”:[{“role”:“user”,“content”:“You are a machine that distributes health care records by service item. Based on the following three constraints, transcribe the health care records corresponding to each service item to the right of the colon and output them.\r\n#Constraints on health care records\r\n·Do not rewrite the contents of the care records without permission\r\n·If the yyyy in (yyyy / MM / dd) is missing, assume that it is 2023\r\n·The word “today” should be assumed to be 2023 / 08 / 30\r\n·Convert dialects to standard polite language\r\n·If it is a language other than Japanese, convert it to polite Japanese language (person names can remain in the other language)\r\n#Constraints on service items\r\n·Do not rewrite the contents of the service items without permission\r\n#Constraints on output results\r\n·If it does not apply to any service item, If there is a nursing record that does not contain the keywords "condition", "status", "situation", or "content", transcribe it into a service item that contains the keywords "appearance", "state", "status", or "content".\r\n·Be aware of the unit of service items\r\n·If there is a quotation mark (") to the left of the colon, extract only what you can read from the nursing record and copy it to the right of the colon.\r\n·If the contents of the quotation marks (") cannot be read from the nursing record, enter only "none" to the right of the colon.\r\n·For service items that cannot be read from the nursing record, enter only "none" to the right of the colon.\r\n·Do not arbitrarily rewrite the contents of either the nursing record or the service item and output the allocation results (do not change the contents of the quotation marks "").\r\n·Return all presented service items.\r\n·Be sure to allocate nursing records to service items.\r\n#Nursing record for health management\r\nHealth management, 36 degrees, 98%, complexion is good, and physical condition is good. Data storage\r\n#Health management service items\r\nBody temperature (℃):\r\nBlood pressure_upper (mmHg):\r\nBlood pressure_lower (mmHg):\r\nPulse (beats / min):\r\nSPO2 (%):\r\nRespiration (beats / min):\r\nCondition observation:\r\n"}],"max_tokens":800,"temperature":0,"top_p":0,"frequency_penalty":0,"presence_penalty":0}

[0095] The above processing request sentence includes the following (A) command sentence, (B) constraint condition, (C) care contents (care records), and (D) care items (service items).

[0096] (A) Imperative statement "You are a machine that divides health care records into service items. Based on the following three constraints, please transcribe and print out the health care records that correspond to each service item to the right of the colon." By including the operation purpose in the processing request statement (command statement), a generative AI (specific examples will be described later) can be operated as the item allocation server 50. Because the above command statement includes the record form name "health management", it is transmitted to the item allocation server 50 that a process should be executed to allocate care records (care contents) related to health management by service item (care item).

[0097] (B) Constraints (B-1) Restrictions on health care records Do not rewrite the contents of nursing records without permission. If yyyy is missing from (yyyy / MM / dd), assume the year is 2023. -Assume that the word "today" is 2023 / 08 / 30 -Convert dialects into standard polite language - If the language is other than Japanese, please convert it into polite Japanese (person names can remain in the other language) The "2023" and "2023 / 08 / 30" in the above constraints are based on the machine time of the record management server 20 that creates the processing request. Specific examples of dialects and languages ​​other than Japanese will be described later. (B-2) Restrictions on service items Do not rewrite the contents of the service items without permission. The "service item" means the same thing as the "care item." Whether to use the "service item" or the "care item" in the processing request statement may be selected depending on the processing accuracy of the item allocation server 50, for example. (B-3) Restrictions on output results If there is a nursing record that does not fit into any of the service items, transcribe it into the service item that contains the keywords "condition," "situation," "content," or "situation." · Be aware of the unit of service items If there are quotation marks (") to the left of the colon, extract only what you can read from the nursing record and copy it to the right of the colon. If the contents of the quotation marks (") cannot be read from the nursing record, enter only "none" to the right of the colon. For service items that cannot be read from the nursing records, enter only "none" to the right of the colon. Do not arbitrarily rewrite the contents of the nursing records or service items and output the results of sorting (including the contents of quotation marks " "). -Return all offered services - Be sure to separate nursing records into service items By including the above-mentioned constraints in the processing request statement, the generative AI can be specialized to operate as an item sorting server 50, and inappropriate processing can be prevented.

[0098] (C) Contents of care (care records) Health management, 36 degrees, 98%, complexion is good, and physical condition is good. Data storage This is the full text of the speech. The "data storage" in the speech does not refer to the care content (care records) itself, but is a keyword to proceed to the record creation phase, as mentioned above.

[0099] (D) Nursing care items (service items) Body temperature(℃): Blood Pressure_Upper(mmHg): Blood Pressure_Lower(mmHg): Pulse (times / min): SPO2(%): Breathing (times / min): Condition Observation: In this case, since the processing request message contains the name of the record form "health management", the care items (service items) contained in the health management record form 32 (Fig. 3), which is one of the multiple record forms, are listed. The listed care items (service items) are based on the record form master DB (Fig. 12). The multiple care items (service items) contained in the record form specified in the utterance (in the above example, the health management record form) are included accurately and without omission in the processing request message and transmitted to the item allocation server 50.

[0100] The item allocation server 50, which has received the above-mentioned processing request statement, uses the command statement, constraint conditions, care contents (care records), and care items (service items) contained in the processing request statement to associate care contents with care items (allocation processing), and transmits the result to the record management server 20 (step 88, data example 88A). For example, the following data (allocation result data) is transmitted from the item allocation server 50 to the record management server 20.

[0101] {"id":"chatcmpl-76rGJpqMdVuM87knY7GaP7Sk8e5tv","object":"chat.completion","created":1681869063,"model":"gpt-35-turbo","usage":{"prompt_tokens":856,"completion_tokens":80,"total_tokens":936},"choices":[{"message":{"role":"assistant","content":"Body temperature (℃): 36.0\nBlood pressure_upper (mmHg):\nBlood pressure_lower (mmHg):\nPulse (beats / min):\nSPO2 (%): 98\nRespiration (beats / min):\nCondition observation: Your complexion is good and you are in good health."},"finish_reason":"stop","index":0}]}

[0102] The results of this allocation include the following:

[0103] Body temperature (℃): 36.0 Blood Pressure_Upper(mmHg): Blood Pressure_Lower(mmHg): Pulse (beats / min): SPO2(%):98 Breathing (times / min): Condition observation: His complexion is good and he is in good health.

[0104] The record management server 20 that has received the distribution result transmits the distribution result and the form data of the record form identified by the form ID (data representing a record form with no data entered) to the record management terminal 30 (step 86, data example 86A).

[0105] The record management terminal 30, which receives the allocation results and the form data for the record form, creates a data-entered record form in which care content (numbers, letters) is entered into each input field of the record form according to the allocation results, and displays the care content input screen 33 including the data-entered record form (step 82, data example 82A).

[0106] Fig. 12 shows the care content input screen 33 of the record management terminal 30 including the data-entered record form. It differs from the care content input screen 33 shown in Fig. 10 in that the data-entered record form is displayed in the record form display area 33e, and the transcribed utterance content is displayed in the utterance content text display area 33g.

[0107] If the care contents are correctly entered in the care items of the record form displayed in the record form display area 33e, the save button 33d is clicked, and the care contents entered in the record form are saved in the record table DB13 (FIG. 6). If there is an error in the entered care contents, the error may be corrected using the input device 30E.

[0108] Clicking the end button 33i ends the voice input.

[0109] The record form display area 33e of the care content input screen 33 is used to visually confirm whether the care content has been correctly input by speech (and to correct any errors). Since the voice recognition result of the speech content is also displayed in the speech content text display area 33g, it is also possible to confirm whether the care content has been correctly recognized by speech in the speech content text display area 33g, and therefore the record form display area 33e does not necessarily have to be displayed. In an embodiment in which the record form display area 33e is not included in the care content input screen 33, the transcribed speech content displayed in the speech content text display area 33g can be viewed, and if there is an error, the utterance including the record form name, care content, and the keyword "data save" can be repeated.

[0110] FIG. 13 shows the relationship between the above-mentioned utterance content (excluding the keyword "data storage" for proceeding to the record creation phase) and the result of the sorting process by the item sorting server 50 (care records using the health care record form 32).

[0111] As described above, health management record form 32 is identified (selected) from among multiple types of record forms by the record form name "health management" included in the utterance. By identifying health management record form 32, the care items (service items) included in health management record form 32 ("body temperature," "abnormal / normal blood pressure," "pulse," "SPO2," "breathing," and "condition observation") are correctly identified based on record form master DB 12 (FIG. 5). This care item (service item) (name) is included in the processing request statement created in record management server 20 and sent to item allocation server 50, so that item allocation server 50 can be accurately informed of the care item (service item) to which the care content (care record) should be allocated.

[0112] In the above-mentioned utterance example ("Health care, 36 degrees, 98%, complexion is good, physical condition is good"), the care item (service item) (noun) itself is not uttered at all, and therefore the processing request sentence does not include any care items. In other words, at the stage of utterance and processing request sentence, the care item (service item) and the care content (care record) are not associated at all. However, even if the care item and the care content are not associated at all, the item allocation server 50 that receives the processing request sentence can associate the care item and the care content quite accurately (item allocation). In other words, in the above-mentioned example, "36 degrees" is simply uttered, and the corresponding care item "body temperature" is not uttered either immediately before or after "36 degrees" (anywhere in the utterance), but the item allocation server 50 can correctly associate "36 degrees" with the care item "body temperature". Similarly, "98%" can be correctly associated with the care item "SPO2". Furthermore, the utterance "She looks healthy and is in good health" can also be correctly associated with the care item "Condition Observation". The meaning of the word in the care item "body temperature", the normal range of a person's body temperature, and the unit (degrees) immediately following the number "36" are considered to be used in the inference for allocating (associating) "36 degrees" to the care item "body temperature" among the multiple care items in the health management record form 32. Similarly, the meaning of the word in the care item "SPO2" (meaning oxygen saturation), the normal range of a person's SPO2, and the unit (%) immediately following the number "98" are considered to be used in the inference for allocating "98%" to the care item "SPO2" among the multiple care items in the multiple health management record forms 32.

[0113] The reason why "He looks healthy and is in good physical condition" is assigned to the "Condition Observation" nursing care item among the multiple nursing care items on Health Management Record Form 32 is thought to be that the constraints in the processing request statement (for example, "If there is a nursing care record that does not fit into any of the service items, transcribe it into a service item that contains the keywords 'appearance,' 'condition,' 'situation,' and 'content'") have an effect on the correct assignment result.

[0114] The processing request text given to the item sorting server 50 includes the command "You are a machine that sorts health care records by service item," as described above. This command also contributes to the item sorting server 50 performing the sorting process accurately. Furthermore, as described above, by adding a constraint condition in addition to the command, it is possible to reliably reduce sorting errors (matching errors).

[0115] The item sorting server 50 can use "ChatGPT", a generative AI (artificial intelligence) available via the Internet, etc. In other words, by giving ChatGPT a processing request statement including the above-mentioned command statements, constraint conditions, etc., it is possible to make ChatGPT operate as the item sorting server 50 and appropriately execute the sorting process. Bard, Phind, and other generative AIs may also be used as the item sorting server 50.

[0116] In the above example, the record management terminal 30 receives the allocation results and form data from the record management server 20 and creates a data-entered record form, but it is also possible to create a data-entered record form in the record management server 20 and then send the data-entered record form created in the record management server 20 to the record management terminal 30.

[0117] FIG. 14 shows other examples of the contents of speech (shown in parentheses) and the results of the sorting process by the item sorting server 50.

[0118] As described above, when the item allocation server 50 receives a processing request including a care item name and care content (care record) from the record management server 20, it performs a process of matching them (allocation process). The example shown in Fig. 14 includes three vital data items, namely, systolic blood pressure, diastolic blood pressure, and body temperature, as the care content (care record).

[0119] As described above, the record form name "health management" included in the utterance identifies (selects) health management record form 32 from among multiple types of record forms. By identifying health management record form 32, record form master DB 12 is referenced, and the care item names ("body temperature," "blood pressure systolic / diastolic," "pulse," "SPO2," "respiration," and "condition observation") are correctly identified.

[0120] In the example shown in FIG. 14, unlike FIG. 13, the care items (service items) and the care contents (care records) are already associated at the utterance stage (the corresponding care contents are spoken immediately after the care item name). Of course, it is also possible to associate the care items and the care contents at the utterance stage in this way. However, in the example shown in FIG. 14, the care contents are not spoken in the order of the care items included in the record form. For example, in the health management record form 32, the care item "body temperature" is the first care item, but in the utterance example shown in FIG. 14, the care item "high blood pressure" is spoken before the care item "body temperature"). In this way, regardless of the order of the care contents in the utterance, the item allocation server 50 can allocate appropriate care contents to appropriate care items. In addition, the item allocation server 50 can determine that the process of replacing "36.5 degrees Celsius" with "36.5", pulse rate, SPO2, respiration, and condition observation are not included in the utterance. Furthermore, even if the care item is not spoken accurately, such as "the lower is 75," and no unit is provided, the item allocation server 50 is likely to infer that the diastolic blood pressure is 75 (mmHg) from the context of the speech. As described above, by including the command "You are a machine that allocates health management care records by service item" in the processing request sentence for the generative AI, and by adding further constraints, the generative AI can be made to operate appropriately as the item allocation server 50, and allocation errors (matching errors) can be reliably reduced.

[0121] FIG. 15 shows other example utterances and the results of the sorting process by the item sorting server 50. In FIG.

[0122] The word "saturation" is used in this example utterance. Although the word "saturation" is not included as a care item in the health care record form 32, it is inferred to mean SPO2. Due to the meaning of the word "saturation" and the presence of the unit "%,," it is considered that the care content (care record) "98%" is assigned to "SPO2," which is one of the care items included in the health care record form 32.

[0123] Fig. 16 shows other example utterances and the results of the sorting process by the item sorting server 50. The example utterances shown in Fig. 16 are spoken in English, not Japanese.

[0124] In the case where the staff working in the nursing care service business are foreign nationals, there are cases where the content can be conveyed more accurately by speaking in their native language. The above-mentioned processing request text includes a constraint for multilingual support ("If the language is other than Japanese, convert it into polite Japanese (person's name can remain in the other language)"). This allows the item allocation server 50 to convert the utterance in a foreign language into Japanese and then perform allocation processing (including selection of a record form). In this way, even if the utterance is in a foreign language, the item allocation server 50 can appropriately select a record form and appropriately associate the care items included in the selected record form with the care contents being uttered. In the example of utterance in English shown in FIG. 16, no care items are uttered, but the care contents being uttered can nevertheless be allocated to the appropriate care items.

[0125] Fig. 17 shows an example of an utterance in a foreign language (English) and another example of the result of the allocation process by the item allocation server 50. The example of the utterance shown in Fig. 17 differs from Fig. 16 in that the care item and the care content have already been associated with each other during the utterance. In this case as well, the care content can be allocated to an appropriate care item.

[0126] 18 shows other speech examples and the results of the allocation process by the item allocation server 50 according to the speech examples. The processing request text created in the record management server 20 and sent to the item allocation server 50 is, for example, as follows:

[0127] {“message”:[{“role”:“user”,“content”:“You are a machine that distributes recreational nursing records by service item. Based on the following three constraints, transcribe and output the recreational nursing records that correspond to each service item to the right of the colon.\r\n#Constraints for recreational nursing records\r\n·Do not rewrite the contents of the nursing records\r\n·If yyyy is missing from (yyyy / MM / dd), assume that it is 2023\r\n·The word “today” is assumed to be 2023 / 08 / 30\r\n·Dialects are standard polite language. · In the case of languages ​​other than Japanese, convert them into polite Japanese (person's names can remain in the other language)\r\n#Constraints on service items\r\n·Do not rewrite the contents of the service items without permission\r\n#Constraints on output results\r\n·If there is a care record that does not fit into any of the service items, transcribe it into a service item that contains the keywords "condition", "status", "situation", and "contents".\r\n·Be aware of the units of service items\r\n·If there are quotation marks (") to the left of the colon, extract only what can be read from the care record and copy it to the right of the colon. \r\n·If the contents of the quotation marks ("") cannot be read from the nursing record, enter only "none" to the right of the colon.\r\n·For service items that cannot be read from the nursing record, enter only "none" to the right of the colon.\r\n·Do not arbitrarily rewrite the contents of the nursing record or service items and output the allocation results (do not change the contents of the quotation marks "").\r\n·Return all presented service items.\r\n·Always allocate nursing records to service items.\r\n#Nursing record for recreation\r\nI did recreational exercises. I was very enthusiastic, but I got out of breath halfway through.\r\nFrom next time, please do light exercise and be sure to take your vital signs.Data storage\r\n#Recreation service items\r\nRecreation "Executed, cancelled or not performed":\r\nRecreation / Reason for cancelling:\r\nRecreation / Program:\r\nStatus:\r\nEquipment / Supplies:\r\nInstructions / Contact:\r\n / Evaluation:\r\n"}],"max_tokens}.

[0128] The recreation record form 92 is specified (selected) by the word "recreation" included in the utterance, and the care items included in the recreation record form 92 ("recreation", "reason for discontinuing", "program", "condition", "equipment and supplies", "instructions and contact", and "evaluation") are included in the processing request. The care item to which the care content should be allocated can be accurately conveyed to the item allocation server 50.

[0129] In the above-mentioned utterance example, for example, the care item "program" is not uttered, but "exercises" in the utterance is correctly assigned to the care item "program." It is considered that the item assignment server 50 is able to make the inference to assign the word "exercises" to the care item "program" based on the meaning of the word. Also, the utterance "shortness of breath halfway through" is assigned to the care item "condition." It is considered that the context of the utterance is also used in the assignment process.

[0130] FIG. 19 shows further example utterances and the results of sorting processing by the item sorting server 50 according to the example utterances.

[0131] As the utterance includes the record form name "moisture", moisture record form 91 is specified (selected) from among multiple types of record forms, and the care item names included in moisture record form 91 ("name of beverage" and "condition") are specified. Item allocation server 50 allocates the following information: "tea" in the utterance is the name of the beverage, "200cc" immediately after "tea" indicates that the amount of tea drunk was "200cc", and the care content corresponding to "condition" is not included in the utterance.

[0132] FIG. 20 shows other example utterances and the results of the sorting process by the item sorting server 50. In FIG.

[0133] The Dinner Record Form 93 is specified (selected) by the word "dinner" included in the utterance, and the care item names included in the Dinner Record Form 93 ("Swallowing rehabilitation," "Amount of food," "Oral nutrition," "Tube feeding," "Medication," "Supplementary foods," "Oral care," "Contents of oral care," "Condition (meals)," and "Missed meals") are identified. The fact that oral care was performed and the contents of oral care are appropriately determined and assigned from the utterance "I brushed my teeth after the meal."

[0134] 21 shows another example of utterance and the result of the allocation process by the item allocation server 50. The processing request sentence created in the record management server 20 and sent to the item allocation server 50 is, for example, as follows.

[0135] {“message”:[{“role”:“user”,“content”:“You are a machine that distributes physical measurement nursing records by service item. Based on the following three constraints, transcribe and output the physical measurement nursing records corresponding to each service item to the right of the colon.\r\n#Constraints for physical measurement nursing records\r\n·Do not rewrite the contents of the nursing records\r\n·If yyyy is missing from (yyyy / MM / dd), assume it is 2023\r\n·The word “today” is assumed to be 2023 / 08 / 30\r\n·Convert dialects to standard polite language · In the case of a language other than Japanese, convert it into polite Japanese (person's names can remain in the other language)\r\n#Constraints on service items\r\n·Do not rewrite the contents of the service items without permission\r\n#Constraints on output results\r\n·If there is a care record that does not fit into any of the service items, transcribe it into a service item that contains the keywords "condition", "status", "situation", or "contents"\r\n·Be aware of the units of service items\r\n·If there are quotation marks (") to the left of the colon, extract only what can be read from the care record and copy it to the right of the colon. \r\n·If the contents of the brackets ("") cannot be read from the nursing record, enter only "None" to the right of the colon.\r\n·For service items that cannot be read from the nursing record, enter only "None" to the right of the colon.\r\n·Do not arbitrarily rewrite the contents of the nursing record or service items and output the allocation results (do not change the contents of the brackets "").\r\n·Return all presented service items.\r\n·Always allocate nursing records to service items.\r\n#Physical measurement nursing record\r\nPhysical measurement, 48,160, I had gained a little weight. Save data\r\n#Physical measurement service itemsr\nHeight (cm):\r\nWeight (kg):\r\nCondition:\r\n"}],"max_tokens":800,"temperature":0,"top_p":0,"frequency_penalty":0,"presence_penalty":0}

[0136] The physical measurement record form 94 is identified (selected) by the word “physical measurement” included in the utterance, and the names of the care items included in the physical measurement record form 94 (“height”, “weight”, and “condition”) are correctly transmitted to the item allocation server 50 .

[0137] Although only "48" and "160" are spoken, "48" is assigned to the care item "weight" and "160" is assigned to the care item "height." It is believed that the meaning of the word for the care item "weight" and the normal range of a person's weight are used in the inference for assigning "48" to the care item "weight" among the multiple care items contained in the physical measurement record form 94. Similarly, the meaning of the word for the care item "height" and the normal range of a person's height are used in the inference for assigning "160" to the care item "height" among the multiple care items contained in the physical measurement record form 94. Even if the care contents are not spoken in the order of the care items contained in the physical measurement record form 94, the appropriate care contents are assigned to the appropriate care items.

[0138] FIG. 22 shows other example utterances and the results of the sorting process by the item sorting server 50. In FIG.

[0139] The care items on the Physical Measurement Record Form 94 are listed in the order of height, weight, and condition, but even if the care contents are not spoken in that order, it is clear that the care contents for "160 cm" are appropriately assigned to the "height" care item, and the care contents for "48.3 kg" are appropriately assigned to the "weight" care item.

[0140] 23 shows another example of speech and the result of the allocation process by the item allocation server 50. The processing request sentence created in the record management server 20 and sent to the item allocation server 50 is, for example, as follows:

[0141] {“message”:[{“role”:“user”,“content”:“You are a machine that divides visiting care records into service items. Based on the following three constraints, transcribe and output the visiting care records corresponding to each service item to the right of the colon.\r\n#Constraints for visiting care records\r\n·Do not rewrite the contents of the care records\r\n·If yyyy is missing from (yyyy / MM / dd), assume it is 2023\r\n·Assume the word “today” is 2023 / 08 / 30\r\n·Convert dialects to standard polite language\ ·If the language is other than Japanese, convert it into polite Japanese (person's names can remain in the other language)\r\n#Constraints on service items\r\n·Do not rewrite the contents of the service items without permission\r\n#Constraints on output results\r\n·If there is a care record that does not fit into any of the service items, transcribe it into a service item that contains the keywords "condition", "status", "situation", or "contents"\r\n·Be aware of the units of service items\r\n·If there are quotation marks (") to the left of the colon, extract only what can be read from the care record and copy it to the right of the colon. \r\n·If the contents of the quotation marks ("") cannot be read from the nursing record, enter only "none" to the right of the colon.\r\n·For service items that cannot be read from the nursing record, enter only "none" to the right of the colon.\r\n·Do not arbitrarily rewrite the contents of the nursing record or service items and output the allocation results (do not change the contents of the quotation marks "").\r\n·Return all the presented service items.\r\n·Always allocate the nursing record to the service items.\r\n#Nursing record of visitation\r\nAt the visitation, my eldest son Taro, his wife, and their two grandchildren came to visit. They were excited talking about their school trip. Data storage\r\n#Visit service items\r\nVisitor:\r\nRelationship:\r\nStatus:\r\nNumber of visitors (people):\r\n"}],"max_tokens":800,"temperature":0,"top_p":0,"frequency_penalty":0,"presence_penalty":0}

[0142] The word "visit" contained in the utterance identifies the visit record form 95, and the item names contained in the visit record form 95 ("visitor," "relationship," "situation," and "number of visitors") are included in the processing request message and are thus correctly conveyed to the item allocation server 50.

[0143] The above example utterance does not include any of the item names included in Visitation Record Form 95, but the content is correctly assigned to the items "visitor," "relationship," "situation," and "number of visitors."

[0144] Furthermore, the example utterance does not include a single word about the number of visitors being four. The item allocation server 50 correctly determines that the number of visitors is four. This is thought to be the result of correctly inferring that the number of visitors is four in total, consisting of "the eldest son, Taro," his "wife," and "two grandchildren," based on the portion of the utterance that says, "the eldest son, Taro, his wife, and two grandchildren."

[0145] Furthermore, the above example utterance also includes "um," an utterance that does not have any particular meaning, and the item allocation server 50 also performs processing to delete this part of the utterance that does not have any particular meaning (by not including it in the "How it is" item).

[0146] FIG. 24 shows other example utterances and the results of sorting by the item sorting server 50 according to the example utterances.

[0147] In the example utterance shown in Figure 24, the number of visitors is correctly inferred to be two from the part of the utterance that mentions "the eldest son and his wife," and "two people" is correctly assigned to the item "number of visitors," even though it is not included in the utterance at all.

[0148] Because the visitors are the "eldest son and his wife," the correct relationship of the visitors would be the eldest son and his spouse, but only "eldest son" is assigned to the care item "relationship" in the visitation record form 95 shown in Figure 24. In such a case, the input device can be used to input (add) "eldest son's spouse," "eldest son's wife," "daughter-in-law," etc., into the data input field for the care item "relationship."

[0149] FIG. 25 shows another example of utterance and the result of the sorting process by the item sorting server 50. In FIG.

[0150] The example utterance shown in Fig. 25 includes a dialect. As described above, the processing request statement includes a constraint statement that "the dialect should be converted into standard polite language," which allows the item sorting server 50 to convert the dialect into polite language (standard language) before sorting. Even if the utterance includes a dialect, it is possible to appropriately match the care item with the care content.

[0151] 26 shows other speech examples and the results of the allocation process by the item allocation server 50 according to the speech examples. The processing request text created in the record management server 20 and sent to the item allocation server 50 is, for example, as follows:

[0152] {“messages”:[{“role”:“user”,“content”:“You are a machine that distributes accident report records by service item. Based on the following three constraints, transcribe and output the accident report records corresponding to each service item to the right of the colon.\r\n#Constraints for accident report records\r\n·Do not rewrite the contents of the accident report records\r\n·If the yyyy in (yyyy / MM / dd) is missing, assume that it is 2023\r\n·The word “today” is assumed to be 2023 / 08 / 30\r\n·Convert dialects to standard polite language · In the case of a language other than Japanese, convert it into polite Japanese (person's names can remain in the other language)\r\n#Constraints on service items\r\n·Do not rewrite the contents of the service items without permission\r\n#Constraints on output results\r\n·If there is a record that does not fit into any service item, transcribe it into a service item that contains the keywords "condition", "status", "situation" or "contents"\r\n·Be aware of the unit of service items\r\n·If there are quotation marks (") to the left of the colon, extract only what can be read from the record and copy it to the right of the colon. \r\n·If the contents of the quotation marks ("") cannot be read from the record, enter only "none" to the right of the colon.\r\n·For service items that cannot be read from the record, enter only "none" to the right of the colon.\r\n·Do not arbitrarily rewrite the contents of the record or service items and output the allocation results (do not change the contents of the quotation marks "").\r\n·Return all presented service items.\r\n·Always allocate records to service items.\r\n#Record of accident report\r\nAccident report: Today around 3 p.m., I fell down the stairs and was injured. Umm, it was caused by my careless stepping. Ah, to prevent this from happening again in the future, I need to accompany you as you go down the stairs.Data storage\r\n#Service item of accident report\r\nReporter:r\nDate of occurrence (yyyy / MM / dd):\r\nTime of occurrence "AM or PM or none":\r\nTime of occurrence (HHmm):\r\nLocation of occurrence "Room or living room or toilet or dressing room or bathroom or hallway or stairs or lounge or outside the facility or none":\r\nLocation of occurrence (outside the facility):\r\nLocation of occurrence (other):\r\nType of accident "Fall or fall or medication error or burn or accidental ingestion or drowning":\r\nAccident occurrence status:\r\n"}],"max_tokens":800,"temperature":0,"top_p":0,"frequency_penalty":0,"presence_penalty":0}.

[0153] Accident report record form 96 is identified by the "accident report" included in the utterance. Although the above utterance example does not include an utterance that specifies a date, "August 30, 2023" is entered (assigned) into the "Date of occurrence" item from "today" included in the utterance. In addition, it is determined that it is "PM" from "around 3 p.m." Furthermore, the location of occurrence "stairs" and accident type "fall" are selected from "fell down the stairs" in the utterance. Even if there is no utterance for each of the items "Date of occurrence," "Time of occurrence," "Location of occurrence," "Type of accident," and "Circumstances of accident occurrence," the appropriate content is assigned to the appropriate item.

[0154] The item allocation server 50 using generative AI (ChatGPT) can infer the correspondence (allocation) between care items (names) and care contents quite appropriately, even if the care item name (service item name) and care content (service content) are not spoken as a set. For example, there is no need to select the data entry field in which the care content should be entered by clicking on the data entry field prepared for each care item on the display screen, and then input letters, numbers, etc. using an input device into the selected data entry field. In addition, there is no need to speak the care content in the order in which the multiple care items included in the record form are arranged. It is possible to input the care content in a short time, which significantly reduces the administrative burden on care staff, etc.

[0155] It should be noted that, although there may be data entry fields for more appropriate care items, care contents may end up being entered into other data entry fields. In such cases, the accuracy of allocation can be improved by increasing the constraints mentioned above. [Explanation of symbols]

[0156] 10. Database Server 11 User Master Database 12 Record Form Master Database 13 Record Tables and Databases 20 Record management server (record management device) 30 Record management terminal (data entry support device) 30G Microphone 30J display device 31 Speech Recognition Library 32 Health Care Record Form 40 Morphological Analysis Server 50 Item sorting server (generative AI) 91 Moisture Record Form 92 Recreation Record Form 93 Dinner Record Form 94 Physical Measurement Record Form 95 Visitation Record Form 96 Accident Report Record Form

Claims

1. a processing request statement creation means for creating a processing request statement including a plurality of data items provided in a recording form having a data input field for each data item, the plurality of data items provided in a recording form identified by the recording form name included in the character string, the data contents included in the character string, a distribution command for associating the data contents with the data items, and a constraint condition for constraining the operation of the generation system AI, in accordance with a character string including a form name of a recording form having a plurality of data items and a data input field for each data item, and data contents to be input into the data input fields, the plurality of data items provided in a recording form identified by the recording form name included in the character string, the data contents included in the character string, a distribution command for associating the data contents with the data items, and a constraint condition for constraining the operation of the generation system AI; a generation system AI utilization means for transmitting a processing request statement, which is created by the processing request statement creation means and includes the data items, data contents, allocation commands, and constraint conditions, to a generation system AI, and receiving a result of an allocation process, which is executed by the generation system AI according to the processing request statement and in which the data contents included in the utterance are allocated to data items according to their contents; The utterance includes data content that is not uttered in a set with the data item, and a data item corresponding to the data content that is not uttered in a set with the data item is inferred from the plurality of data items by the generative AI. Record keeping devices.

2. A creation means is provided for creating data representing a data-entered record form in which the data content is assigned and input to a data input field of a data item corresponding to the content according to the assignment process result received by the generation system AI utilization means.

2. The record management device according to claim 1.

3. The above string is the result of speech recognition processing based on the utterance.

2. The record management device according to claim 1.

4. a record form master database in which the record form names and a plurality of data items included in the record forms identified by the record form names are stored in association with each other is connected to the record management device; 2. The record management device according to claim 1.

5. The processing request statement includes a command statement instructing the generating system AI to operate as a machine that distributes the data contents by data item, 2. The record management device according to claim 1.

6. The processing request statement includes a command statement instructing the generating AI to support multiple languages, 2. The record management device according to claim 1.

7. A control means for causing the record management device to function as the processing request statement creation means and the generation system AI utilization means when the given character string includes a specific keyword, 2. The record management device according to claim 1.

8. A processing request statement creation means creates a processing request statement according to a character string including a form name of a record form provided in accordance with the utterance, the form name having a plurality of data items and a data input field for each data item, and data content to be input into the data input field, the processing request statement including a plurality of data items provided in a record form identified by the record form name included in the character string, the data content included in the character string, a distribution command for associating the data content with the data items, and a constraint condition for constraining the operation of the generation system AI; A generation system AI utilization means transmits a processing request statement, which includes the data items, data contents, allocation commands, and constraint conditions, created by the processing request statement creation means, to the generation system AI, and receives a result of the allocation process, which is executed by the generation system AI according to the processing request statement, and allocates the data contents included in the utterance to data items according to their contents; The utterance includes data content that is not uttered in a set with the data item, and a data item corresponding to the data content that is not uttered in a set with the data item is inferred from the plurality of data items by the generative AI. Allocation method.

Images