Service support system, service support method, data management program, and search program

Abstract

Provided is a structure that supports a reply service for a current query. The business support system is provided with: a division unit that generates a plurality of divided file data by dividing, for each of the query items, a plurality of query items included in file data of a current query; a first acquisition unit that acquires a plurality of pieces of input-completed divided file data in which reply content for the question content of each of the plurality of question items is input; a second acquisition unit that acquires input-completed file data, which is file data in a single file format, generated on the basis of the acquired plurality of input-completed divided file data; and a storage unit that associates each of the plurality of pieces of input-completed divided file data acquired with the input-completed file data and stores the associated data in a database.

Description

Technical Field

[0001] The present disclosure relates to a business support system, a business support method, a data management program, and a retrieval program. Background Art

[0002] When manufacturing and selling drugs, etc., it is necessary to apply for approval of manufacturing and selling drugs, etc. to the authorities in advance and obtain approval from the Minister of Health, Labour and Welfare. Abroad, it is also necessary to obtain approval from authorities such as the Food and Drug Administration in the United States and the European Medicines Agency (European Medicines Agency). In addition, in drug reviews, etc., when there is an inquiry (authority inquiry) from the authorities regarding the application content, the applicant is required to appropriately respond to the authority inquiry. In addition, outside of drug reviews, etc., when there is an inquiry from the authorities (hereinafter referred to as a pharmaceutical affairs inquiry, etc.) to a drug development company, etc., the drug development company, etc. is also required to appropriately respond to the pharmaceutical affairs inquiry, etc. In the following description, it is assumed that the authority inquiry and its response also include the pharmaceutical affairs inquiry and its response.

[0003] Prior Art Documents

[0004] Patent Documents

[0005] Patent Document 1: International Publication No. 2016 / 157467

[0006] Non-Patent Documents

[0007] Non-Patent Document 1: Independent Administrative Institution , [online], [searched on October 21, Reiwa 4], Internet <URL: https: / / www.pmda.go.jp / review-services / drug-reviews / 0001.html> Summary of the Invention

[0008] Problems to be Solved by the Invention

[0009] Here, until now, the response operation to the authority inquiry has been carried out based on the experience of the staff (the person who actually formulates the response for each inquiry content). Specifically, the staff, for example, relies on their own memory, searches for the response content to past similar authority inquiries, and formulates the response for each inquiry content while referring to them.

[0010] On the other hand, for such business processes, for example, if a structure (framework) that can simply retrieve past similar cases of inquiries by the authorities can be provided, the business efficiency of the staff will be improved, which helps support the staff's response operations.

[0011] An object of the present disclosure is to provide a structure (framework) that supports response operations to inquiries by the authorities.

[0012] Means for Solving the Problem

[0013] According to one aspect, a business support system includes:

[0014] A division unit that generates a plurality of divided file data by dividing a plurality of inquiry items included in the document data of the authority inquiry for each inquiry item;

[0015] A first acquisition unit that acquires a plurality of divided file data with input response contents for each of the plurality of inquiry items;

[0016] A second acquisition unit that acquires input-completed file data that is file data in a single file form generated from the plurality of acquired divided file data with input response contents; and

[0017] A storage unit that associates the plurality of acquired divided file data with input response contents with the input-completed file data and stores them in a database.

[0018] Advantageous Effects of the Invention

[0019] According to the present disclosure, a structure (framework) that supports response operations to inquiries by the authorities can be provided. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 FIG. 1 is an example of a system configuration diagram of a business support system.

[0021] Figure 2 FIG. 2 is an example of a system configuration diagram of a business support system.

[0022] Figure 3 FIG. is an example of a hardware configuration diagram of a data management device and a server device.

[0023] Figure 4 FIG. is an example of an authority inquiry document data.

[0024] Figure 5 FIG. 1 is an example of a functional configuration diagram of a management unit of a data management device.

[0025] Figure 6 This is a diagram showing a specific example of the processing of the division unit included in the management unit of the data management device.

[0026] Figure 7 This is a diagram showing a specific example of the processing of the distribution unit and the collection unit included in the management unit of the data management device.

[0027] Figure 8 This is a diagram showing a specific example of the processing of the file data generation unit included in the management unit of the data management device.

[0028] Figure 9 This is a diagram showing a specific example of the processing of the attribute information extraction unit included in the management unit of the data management device.

[0029] Figure 10 This is a diagram showing an example of registration data.

[0030] Figure 11 This is the first diagram showing an example of the functional structure of the retrieval unit of the server device.

[0031] Figure 12 This is the first diagram showing an example of the retrieval screen displayed on the staff terminal.

[0032] Figure 13 This is the first timing diagram showing the process flow in the accumulation phase of the business support system.

[0033] Figure 14 This is the first timing diagram showing the process flow in the search phase of the business support system.

[0034] Figure 15 This is the third diagram showing an example of the system structure of the business support system.

[0035] Figure 16 This is the fourth diagram showing an example of the system structure of the business support system.

[0036] Figure 17 This is a diagram showing an example of the functional structure of the learning unit of the learning device.

[0037] Figure 18 This is a diagram showing an example of the functional structure of the management unit of the data management device and the functional structure of the prediction unit of the prediction device.

[0038] Figure 19It is a timing diagram showing the process of processing in the training phase of the business support system.

[0039] Figure 20 It is the second timing diagram showing the process of processing in the accumulation phase of the business support system.

[0040] Figure 21 It is the fifth figure showing an example of the system structure of the business support system.

[0041] Figure 22 It is the second figure showing an example of the functional structure of the retrieval unit of the server device.

[0042] Figure 23 It is the second figure showing an example of the retrieval screen displayed on the staff terminal.

[0043] Figure 24 It is the first figure showing a specific example of the processing of the first vectorization unit, separation unit, and second vectorization unit included in the retrieval unit of the server device.

[0044] Figure 25 It is the second timing diagram showing the process of processing in the retrieval phase of the business support system.

[0045] Figure 26 It is the third figure showing an example of the functional structure of the retrieval unit of the server device.

[0046] Figure 27 It is the third figure showing an example of the retrieval screen displayed on the staff terminal.

[0047] Figure 28 It is the sixth figure showing an example of the system structure of the business support system.

[0048] Figure 29 It is a figure for explaining the outline of the processing in the feedback phase of the business support system.

[0049] Figure 30 It is a figure showing an example of the functional structure of the analysis unit of the analysis device.

[0050] Figure 31 It is a figure showing a specific example of the processing of the learning unit included in the analysis unit of the analysis device.

[0051] Figure 32 It is the second figure showing a specific example of the processing of the first vectorization unit, separation unit, and second vectorization unit included in the retrieval unit of the server device.

[0052] Figure 33 It is a timing diagram showing the process of processing in the feedback phase of the business support system.

[0053] Figure 34 Figure 7 showing an example of the system structure of the business support system.

[0054] Figure 35 Figure 1 showing an example of the LM learning dataset.

[0055] Figure 36 The first timing diagram showing the process flow in the LM learning stage of the business support system.

[0056] Figure 37A Figure 8 showing an example of the system structure of the business support system.

[0057] Figure 37B Figure 9 showing an example of the system structure of the business support system.

[0058] Figure 38 Figure 2 showing an example of the LM learning dataset.

[0059] Figure 39A The second timing diagram showing the process flow in the LM learning stage of the business support system.

[0060] Figure 39B The third timing diagram showing the process flow in the LM learning stage of the business support system.

[0061] Figure 40 Figure 10 showing an example of the system structure of the business support system.

[0062] Figure 41 The first timing diagram showing the process flow in the generation stage of the business support system.

[0063] Figure 42 Figure 11 showing an example of the system structure of the business support system.

[0064] Figure 43 Figure 3 showing an example of the LM learning dataset.

[0065] Figure 44 The fourth timing diagram showing the process flow in the LM learning stage of the business support system.

[0066] Figure 45 Figure 12 showing an example of the system structure of the business support system.

[0067] Figure 46 The timing diagram showing the process flow in the pharmacy application stage of the business support system.

[0068] Figure 47 Figure 13 showing an example of the system structure of the business support system.

[0069] Figure 48A It is the first timing chart showing the process of processing in the pharmaceutical affairs inquiry stage of the business support system.

[0070] Figure 48B It is the second timing chart showing the process of processing in the pharmaceutical affairs inquiry stage of the business support system.

[0071] Figure 49A It is the 14th figure showing an example of the system structure of the business support system.

[0072] Figure 49B It is the 15th figure showing an example of the system structure of the business support system.

[0073] Figure 50A It is the first timing chart showing the process of processing in the retrieval stage and the generation stage of the business support system.

[0074] Figure 50B It is the second timing chart showing the process of processing in the retrieval stage and the generation stage of the business support system. Detailed implementation manners

[0075] Hereinafter, each embodiment will be described with reference to the accompanying drawings. In addition, in this specification and the accompanying drawings, for structural elements having substantially the same functional structure, duplicate explanations are omitted by assigning the same reference numerals.

[0076] [First Embodiment]

[0077] <System Structure of Business Support System>

[0078] First, the system structure of the business support system according to the first embodiment will be described. In the first embodiment, the business support system executes the processing in the accumulation stage and the processing in the retrieval stage. Therefore, the system structure will be described for each stage here.

[0079] (1) Business Support System in Accumulation Stage

[0080] Figure 1 It is the first figure showing an example of the system structure of the business support system, showing the system structure in the accumulation stage. The accumulation stage refers to the stage that is executed in parallel with the response service to the authority inquiry from the authority, and is the stage of constructing a database for enabling simple retrieval of past similar cases of the authority inquiry.

[0081] As Figure 1As shown, in the accumulation stage, the business support system 100 includes an authority terminal 110, a data management device 120, staff terminals 130, 140, and a server device 150. In the business support system 100, the authority terminal 110 and the data management device 120 are communicably connected via an external network 160. In addition, in the business support system 100, the data management device 120, the staff terminals 130, 140, and the server device 150 are communicably connected via an internal network 170.

[0082] The authority terminal 110 is a terminal operated by the person in charge 111 who has received an application for approval of manufacturing and sales of drugs, etc., when making inquiries (authority inquiries) to the applicant 122 regarding the application content. By operating the authority terminal 110 by the person in charge 111, for example, authority inquiry file data (file data sent from the authority during authority inquiries) is sent to the data management device 120 via the external network 160. In addition, in the present embodiment, it is described that the authority inquiry conducted by the person in charge 111 is performed by sending authority inquiry file data as electronic data, but the method of the authority inquiry conducted by the person in charge 111 is not limited to this. For example, there may be a case where the authority inquiry is conducted by fax transmission. However, in this case, the applicant 122 will digitalize the authority inquiry file, and the data management device 120 will obtain the digitalized authority inquiry file data.

[0083] In addition, the authority terminal 110 receives input completion file data (file data for which a reply content has been input in response to the authority inquiry file data) input from the data management device 120 via the external network 160 in response to the sending of the authority inquiry file data to the data management device 120.

[0084] A data management program is installed in the data management device 120, and by executing this program in the accumulation stage, the data management device 120 functions as a management unit 121.

[0085] The management unit 121 operates based on the operation instructions of the applicant 122. Specifically, the management unit 121 receives the authority inquiry file data sent by the authority terminal 110, divides the received authority inquiry file data for each inquiry item, and thus generates a plurality of divided file data. In addition, the management unit 121 sends the generated plurality of divided file data to the corresponding staff terminals 130, 140, etc.

[0086] In addition, the management unit 121 collects input completion divided file data for which reply contents for each of the plurality of inquiry items have been input from the staff terminals 130, 140, etc. in response to the sending of the generated plurality of divided file data to the corresponding staff terminals 130, 140, etc.

[0087] In addition, the management department 121 generates the input-completed file data by setting the collected and divided file data after input as a file format, and sends the generated input-completed file data to the authority terminal 110 via the external network 160. Additionally, in the present embodiment, it is described that the response to the authority inquiry is made by sending the input-completed file data as electronic data, but the response method of the applicant 122 is not limited to this. For example, there may be a case where the response to the authority inquiry is sent by fax.

[0088] Furthermore, the management department 121 stores the registration data including the divided file data, the input-completed divided file data, the input-completed file data, etc. in the database 151 of the server device 150. Additionally, when storing the registration data, the management department 121 associates and stores the keywords given by the applicant 122 for easy retrieval. The keyword is a word or phrase recalled based on the inquiry content and the response content included in the input-completed divided file data.

[0089] The staff terminals 130 and 140 receive the divided file data sent from the data management device 120 and display the inquiry content. In addition, the staff terminals 130 and 140 accept the response content input by the staff 131 and 141 in response to the display of the inquiry content, and generate the input-completed divided file data. Furthermore, the staff terminals 130 and 140 send the generated input-completed divided file data to the data management device 120.

[0090] The server device 150 receives the registration data sent from the data management device 120 and stores it in the database 151. In the accumulation stage, whenever the data management device 120 sends the input-completed file data to the authority terminal 110, the server device 150 stores the registration data in the database 151.

[0091] (2) Business support system in the retrieval stage

[0092] Figure 2 Fig. 2 shows an example of the system structure of the business support system, showing the system structure in the retrieval stage. The retrieval stage refers to the stage that is executed in parallel with the response operation to the authority inquiry from the authority, and is the stage where the staff retrieves past similar cases of the authority inquiry from the database to formulate a response to the inquiry content.

[0093] The difference from the system structure in the accumulation stage is that in the case of the business support system 200 in the retrieval stage, the server device 150 functions as the retrieval unit 201.

[0094] A retrieval program is installed in the server device 150, and by executing this program in the retrieval phase, the server device 150 functions as the retrieval unit 201.

[0095] The retrieval unit 201 provides a retrieval screen to the staff terminals 130 and 140. In addition, when the retrieval unit 201 receives a retrieval request from the staff terminals 130 and 140 in response to the provision of the retrieval screen, it retrieves the database 151 and obtains the registered data that matches the retrieval conditions included in the retrieval request. In addition, the obtained registered data is displayed on the retrieval screen as the retrieval result.

[0096] In addition, in Figure 2 the difference in the system structure in the accumulation phase is that in the case of the business support system 200 in the retrieval phase, the staff terminals 130 and 140 function as retrieval terminals.

[0097] Specifically, the staff terminals 130 and 140 access the server device 150 based on the operation instructions of the staff 131 and 141, thereby displaying the retrieval screen provided by the server device 150. In addition, the staff terminals 130 and 140 accept the input of the retrieval conditions by the staff 131 and 141, and when the retrieval instruction is input by the staff 131 and 141, they send the retrieval request including the retrieval conditions to the server device 150. In addition, the staff terminals 130 and 140 display the retrieval screen including the retrieval result provided by the server device 150 in response to the sending of the retrieval request.

[0098] <Hardware of the Data Management Device and the Server Device>

[0099] Next, the hardware structures of the data management device 120 and the server device 150 in each device constituting the business support systems 100 and 200 will be described. Figure 3 It is a diagram showing an example of the hardware structures of the data management device and the server device.

[0100] As Figure 3 shown, the data management device 120 has a processor 301, a memory 302, an auxiliary storage device 303, an operation device 304, a display device 305, a communication device 306, and a drive device 307. Each hardware included in the data management device 120 is interconnected via a bus 308.

[0101] The processor 301 has various computing devices such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). The processor 301 reads out various programs (such as a data management program, etc.) from the memory 302 and executes them.

[0102] The memory 302 has main storage devices such as a ROM (Read Only Memory) and a RAM (Random Access Memory). The processor 301 and the memory 302 form a so-called computer, and by the processor 301 executing various programs read out from the memory 302, this computer realizes various functions.

[0103] The auxiliary storage device 303 stores various programs and various information used when the processor 301 executes various programs.

[0104] The operation device 304 is used when the applicant 122 inputs various operation instructions to the data management device 120. The display device 305 displays the results of the processing in the data management device 120.

[0105] The communication device 306 is a communication device for communicating with the authority terminal 110 via an external network 160 or for communicating with the staff terminals 130, 140, and the server device 150 via an internal network 170.

[0106] The drive device 307 is a device for setting the storage medium 310. The storage medium 310 mentioned here includes media that record information optically, electrically, or magnetically, such as a CD-ROM, a floppy disk, and a magneto-optical disk. In addition, the storage medium 310 may also include semiconductor memories that record information electrically, such as a ROM and a flash memory.

[0107] In addition, various programs installed in the auxiliary storage device 303 are, for example, set in the drive device 307 through the distributed storage medium 310, and are installed by the drive device 307 reading out various programs recorded in the storage medium 310. Or, various programs installed in the auxiliary storage device 303 may also be installed by downloading from the external network 160 via the communication device 306.

[0108] In addition, as Figure 3 shown, the server device 150 has a processor 321, a memory 322, an auxiliary storage device 323, a connection device 324, a communication device 325, and a drive device 326. Each hardware included in the server device 150 is interconnected via a bus 327.

[0109] In addition, each piece of hardware included in the server device 150 is substantially the same as each piece of hardware included in the data management device 120. Further, the storage medium 333 of the drive device 326 provided in the server device 150 is also the same as the storage medium 310 of the drive device 307 provided in the data management device 120. The difference in the hardware configuration from that of the data management device 120 is that, in the case of the server device 150, the operation device 331 and the display device 332 are connected to the server device 150 via the connection device 324. In addition, in the case of the server device 150, the processor 321 reads out and executes the retrieval program on the memory 322. Further, in the case of the server device 150, the database 151 is implemented in the auxiliary storage device 323.

[0110] <Explanation of the government inquiry document data>

[0111] Next, a specific example of the government inquiry document data sent from the government terminal 110 will be described. Figure 4 is a diagram showing an example of the government inquiry document data. As Figure 4 shown, the government inquiry document data 400 includes a plurality of pairs of inquiry items and inquiry contents from the government regarding the application contents of the manufacturing and sales approval applications for drugs, etc. of the applicant 122. Figure 4 The example of shows a case including three pairs of inquiry items and inquiry contents (the pairs of the three inquiry items 1 to 3 and the inquiry contents indicated by the symbols 410, 420, and 430). In addition, the number of pairs of inquiry items and inquiry contents included in the government inquiry document data 400 is arbitrary, and in the present embodiment, it is described as including a plurality of pairs.

[0112] <Functional structure of the management unit of the data management device>

[0113] Next, the functional structure of the management unit 121 of the data management device 120 will be described. Figure 5 is the first diagram showing an example of the functional structure of the management unit of the data management device. As described above, in the accumulation stage, the data management device 120 functions as the management unit 121. In addition, as Figure 5 shown, the management unit 121 further includes:

[0114] · A document data acquisition unit 501,

[0115] · A user interface unit 502 (an example of the third acquisition unit),

[0116] · A division unit 503,

[0117] · An allocation unit 504,

[0118] · Attribute information extraction unit 505,

[0119] · Collection unit 506 (an example of the first acquisition unit),

[0120] · Document data generation unit 507 (an example of the second acquisition unit),

[0121] · Document data sending unit 508,

[0122] · Storage unit 509.

[0123] The document data acquisition unit 501 acquires the authority inquiry document data sent from the authority terminal 110 (in the case where the authority inquiry is made by fax, it is the electronic authority inquiry document data) and notifies it to the division unit 503.

[0124] The division unit 503 divides the authority inquiry document data notified by the document data acquisition unit 501 in units of inquiry items to generate a plurality of divided document data. In addition, the division unit 503 notifies the plurality of generated divided document data to the distribution unit 504.

[0125] The distribution unit 504 sends the plurality of divided document data notified by the division unit 503 to the staff terminals 130, 140, etc. The sending destinations when the distribution unit 504 sends the plurality of divided document data can be determined either according to pre-determined rules or based on the operation instructions of the applicant 122.

[0126] In response to the situation where the distribution unit 504 has sent a plurality of divided document data, the collection unit 506 collects the plurality of completed input divided document data sent from the staff terminals 130, 140, etc. In addition, the collection unit 506 notifies the plurality of collected completed input divided document data to the document data generation unit 507 and the storage unit 509.

[0127] The document data generation unit 507 generates the completed input document data by setting the plurality of completed input divided document data notified from the collection unit 506 in a single file format. Specifically, the document data generation unit 507, for example, after aggregating the plurality of completed input divided document data in text format into a single document data, converts it into a specified file format such as pdf to generate the completed input document data. In addition, the document data generation unit 507 notifies the generated completed input document data to the document data sending unit 508 and the storage unit 509.

[0128] The document data sending unit 508 sends the completed input document data notified from the document data generation unit 507 to the authority terminal 110 (or sends a fax to the authority).

[0129] The attribute information extraction unit 505 extracts the attribute information included in the authority inquiry document data acquired by the document data acquisition unit 501. In addition, the attribute information extraction unit 505 extracts the attribute information included in the input-completed document data generated by the document data generation unit 507. Further, the attribute information extraction unit 505 notifies the extracted attribute information to the storage unit 509. In addition, the details of the attribute information extracted by the attribute information extraction unit 505 will be described later.

[0130] The user interface unit 502 displays the input-completed divided document data notified from the collection unit 506 to the storage unit 509 to the applicant 122. In addition, the user interface unit 502 acquires the keywords input by the applicant 122 in response to the display of the input-completed divided document data. In addition, the user interface unit 502 notifies the acquired keywords to the storage unit 509. In addition, the keywords input by the applicant 122 are used as part of the search conditions in the search stage.

[0131] The storage unit 509 stores the registration data in the database 151 of the server device 150. The registration data includes:

[0132] · The input-completed divided document data notified from the collection unit 506,

[0133] · The input-completed document data notified from the document data generation unit 507,

[0134] · The attribute information notified from the attribute information extraction unit 505,

[0135] · The keywords notified from the user interface unit 502,...

[0136] etc.

[0137] <Specific examples of the processing of each part of the data management device>

[0138] Next, specific examples of the processing of each part (here, the division unit 503, the allocation unit 504, the collection unit 506, the document data generation unit 507, and the attribute information extraction unit 505) included in the management unit 121 of the data management device 120 will be described.

[0139] (1) Specific example of the processing of the division unit included in the management unit of the data management device

[0140] First, a specific example of the processing of the division unit 503 included in the management unit 121 of the data management device 120 will be described. Figure 6 is a diagram showing a specific example of the processing of the division unit included in the management unit of the data management device.

[0141] As Figure 6As shown, in the case of the authority inquiry document data 400 including a pair of three inquiry items and inquiry contents (the pair of three inquiry items 1 to 3 and the inquiry contents indicated by symbols 410, 420, and 430), it is divided by the division unit 503 into three divided document data 610, 620,.... In addition, in Figure 6 the example of

[0142] as Figure 6 shown, the divided document data 610 includes one inquiry item, the corresponding inquiry content, and one reply content (blank). Similarly, the divided document data 620 includes one inquiry item, the corresponding inquiry content, and one reply content (blank).

[0143] (2) Specific examples of the processing of the distribution unit and the collection unit included in the management unit of the data management device

[0144] Next, specific examples of the processing of the distribution unit 504 and the collection unit 506 included in the management unit 121 of the data management device 120 will be described. Figure 7 is a diagram showing specific examples of the processing of the distribution unit and the collection unit included in the management unit of the data management device.

[0145] As Figure 7 shown, the divided document data 610, 620 sent by the distribution unit 504 to the staff terminals 130, 140, etc. respectively include one inquiry item, the corresponding inquiry content, and one reply content (blank).

[0146] In addition, as Figure 7 shown, the input-completed divided document data 710, 720 collected by the collection unit 506 from the staff terminals 130, 140, etc. respectively include one inquiry item, the corresponding inquiry content, and one reply content. Here, the so-called one reply content is the reply content input by the staff 131, 141, etc.

[0147] (3) Specific examples of the processing of the document data generation unit included in the management unit of the data management device

[0148] Next, specific examples of the processing of the document data generation unit 507 included in the management unit 121 of the data management device 120 will be described. Figure 8 is a diagram showing specific examples of the processing of the document data generation unit included in the management unit of the data management device.

[0149] As Figure 8As shown in the figure, the input completed divided file data 710 and 720 collected by the collection unit 506 are set in a single file format by the file data generation unit 507 to generate the input completed file data 800.

[0150] In addition, Figure 8 The example of

[0151] · The input completed divided file data 710 (including a query item, the corresponding query content, and a reply content therein),

[0152] · The input completed divided file data 720 (including a query item, the corresponding query content, and a reply content therein)

[0153] are continuously arranged within a single page. Among them, the arrangement method of the input completed divided file data in the input completed file data 800 is arbitrary. Multiple input completed divided file data can be arranged within one page, or only one input completed divided file data can be arranged within one page.

[0154] (4) Specific examples of the processing of the attribute information extraction unit included in the management unit of the data management device

[0155] Next, specific examples of the processing of the attribute information extraction unit 505 included in the management unit 121 of the data management device 120 will be described. Figure 9 It is a figure showing specific examples of the processing of the attribute information extraction unit included in the management unit of the data management device.

[0156] As Figure 9 shown, the attribute information extraction unit 505 extracts the attribute information 900 from the authority query file data 400 obtained by the file data acquisition unit 501 and the input completed file data 800 generated by the file data generation unit 507.

[0157] Figure 9 The example of

[0158] In addition, the so-called inquiry period refers to the year, month, and day when the authority inquiry document data 400 is received from the authority terminal 110. Furthermore, the so-called disease field of the inquiry content refers to the disease field to which the inquiry content belongs. The disease fields are predefined. In the present embodiment, the disease fields include, for example, digestive organs, circulatory organs, respiratory organs, central nervous system, vaccines, vitamins, antibodies, and the like. In addition, the so-called event being inquired about refers to which event among various events (face-to-face advice, clinical trial notification, approval review, reliability survey, etc.) included in the manufacturing and sales approval application of drugs and the like is being inquired about. In addition, the so-called variety of the inquiry content refers to, for example, the type of drug. In addition, the so-called reply period refers to the year, month, and day when the input-completed document data 800 is sent to the authority terminal 110.

[0159] <Explanation of registration data>

[0160] Next, a specific example of the registration data stored in the database 151 of the server device 150 by the data management device 120 will be described. Figure 10 It is a diagram showing an example of the registration data.

[0161] As Figure 10 shown, the registration data 1000 includes "number", "authority inquiry ID", "authority inquiry document data", "attribute information", "input-completed document data", "input-completed divided document data", and "keyword" as information items.

[0162] In the "number", a number indicating the storage order when each registration data included in the registration data 1000 is stored in the database 151 is stored.

[0163] In the "authority inquiry ID", an identifier for identifying each authority inquiry document data sent from the authority terminal 110 is stored.

[0164] In the "authority inquiry document data", the authority inquiry document data sent from the authority terminal 110 is stored. In Figure 10 it, "Pharmacy 001" refers to the authority inquiry document data 400, for example.

[0165] In the "attribute information", the attribute information extracted from the corresponding authority inquiry document data or input-completed document data is stored. In Figure 10 it, "AAA" refers to the attribute information 900, for example.

[0166] In the "input completed file data", store the input completed file data generated based on the corresponding authority query file data. In Figure 10 "PDF001" refers to, for example, the input completed file data 800.

[0167] In the "input completed divided file data", store the generated input completed divided file data used to generate the corresponding input completed file data. In Figure 10 "Text 1001" refers to, for example, the input completed divided file data 710. In addition, "Text 1002" refers to, for example, the input completed divided file data 720. Also, "Text 1001" can be either the input completed divided file data 710 itself or the text data obtained by extracting the text data part from the input completed divided file data 710. Similarly, "Text 1002" can be either the input completed divided file data 720 itself or the text data obtained by extracting the text data part from the input completed divided file data 720.

[0168] In the "keywords", store the keywords given by applicant 122 based on the corresponding input completed divided file data. In Figure 10 "K1001" refers to, for example, the keyword given by applicant 122 based on the input completed divided file data 710. In addition, "K1002" refers to, for example, the keyword given by applicant 122 based on the input completed divided file data 720.

[0169] <Functional Structure of the Retrieval Unit of the Server Device>

[0170] Next, the functional structure of the retrieval unit of the server device 150 will be described. Figure 11 Figure 1 shows an example of the functional structure of the retrieval unit of the server device. As described above, in the retrieval stage, the server device 150 functions as the retrieval unit 201. In addition, as Figure 11 shown, the retrieval unit 201 further includes:

[0171] · A retrieval screen providing unit 1101,

[0172] · A retrieval condition acquisition unit 1102,

[0173] · A retrieval control unit 1103.

[0174] In the presence of access from staff terminals 130, 140, etc., the retrieval screen providing unit 1101 provides a retrieval screen. In addition, the retrieval screen providing unit 1101 receives retrieval requests from staff terminals 130, 140, etc. in response to the provision of the retrieval screen. In addition, when the retrieval result is notified from the retrieval control unit 1103, the retrieval screen providing unit 1101 provides the retrieval result included in the retrieval screen to staff terminals 130, 140, etc. In addition, the retrieval screen providing unit 1101 receives an instruction for detailed display from staff terminals 130, 140, etc. in response to the provision of the retrieval screen including the retrieval result. Furthermore, when the detailed information is notified from the retrieval control unit 1103, the retrieval screen providing unit 1101 provides the detailed information included in the retrieval screen to staff terminals 130, 140, etc.

[0175] The retrieval condition acquisition unit 1102 acquires the retrieval conditions included in the retrieval requests received by the retrieval screen providing unit 1101. In addition, the retrieval condition acquisition unit 1102 notifies the acquired retrieval conditions to the retrieval control unit 1103.

[0176] The retrieval control unit 1103 retrieves the database 151 and acquires the input-completed divided file data that matches the retrieval conditions notified from the retrieval condition acquisition unit 1102. In addition, the retrieval control unit 1103 notifies the acquired input-completed divided file data as the retrieval result to the retrieval screen providing unit 1101.

[0177] In addition, when the instruction for detailed display received in the retrieval screen providing unit 1101 is acquired, the retrieval control unit 1103 acquires registration data (attribute information, input-completed file data, keywords, etc.) other than the input-completed divided file data notified as the retrieval result from the database 151. In addition, the retrieval control unit 1103 notifies the registration data (attribute information, input-completed file data, keywords, etc.) other than the input-completed divided file data notified as the retrieval result as the detailed information to the retrieval screen providing unit 1101.

[0178] <Explanation of the retrieval screen>

[0179] Next, the retrieval screen provided by the server device 150 to staff terminals 130, 140, etc. will be explained. Figure 12 Figure 1 shows an example of the retrieval screen displayed on the staff terminal.

[0180] Among them, the retrieval screen 1210 is a screen for inputting retrieval conditions and sending retrieval requests. For example, as Figure 12As shown, in the search screen 1210, "character string", "period of inquiry by the authorities", "disease area", "event", "variety", and "reply period" are included as search items.

[0181] The "character string" is an item input when retrieving the input-completed divided file data based on the character string included in the inquiry content within the divided file data after input completion and the character string included in the keyword associated with the input-completed divided file data.

[0182] "Period of inquiry by the authorities" to "reply period" are items input when retrieving the input-completed divided file data based on the attribute information associated with the input-completed divided file data. Additionally, for the "period of inquiry by the authorities" and "reply period", the first day and the last day of the period are input. Thus, in the search control unit 1103, it is possible to retrieve the input-completed divided file data in which the period of inquiry by the authorities or the reply period is included within the input period.

[0183] Furthermore, for "disease area" to "variety", options selected by the staff members 131, 141, etc. from the pre-determined options are input. Thus, in the search control unit 1103, it is possible to retrieve the input-completed divided file data associated with the attribute information that matches the input options.

[0184] In addition, in the search screen 1210, a search button 1211 is included. If the input to the search items is completed and the search button 1211 is pressed, a search request setting the information input to the search items as search conditions is sent to the server device 150.

[0185] On the other hand, the search screen 1220 is a screen that is displayed on the staff terminals 130, 140, etc. when the server device 150 retrieves the input-completed divided file data that matches the search conditions according to the sent search request and provides a search screen including the search results.

[0186] As Figure 12 shown, the number of input-completed divided file data that match the search conditions is displayed in the search screen 1220. In addition, the input-completed divided file data that match the search conditions and the registration data other than the input-completed divided file data (however, for the registration data other than the input-completed divided file data, only the file name is displayed here) are displayed in the search screen 1220.

[0187] In addition, a detailed display button 1221 is included in the search screen 1220 and is pressed when, for any of the registration data other than the input-completed divided file data that matches the search conditions, it is desired to display the details. After the staff members 131, 141, etc. select the registration data (file name) other than the input-completed divided file data for which they want to display the details, they press the detailed display button 1221. As a result, an instruction for detailed display of the selected registration data (file name) other than the input-completed divided file data is sent to the server device 150.

[0188] <Process of Business Support System (Accumulation Phase)>

[0189] Next, the process flow in the accumulation phase of the business support system 100 will be described. Figure 13 This is the first timing chart showing the process flow in the accumulation phase of the business support system.

[0190] In step S1301, the authority terminal 110 sends an authority inquiry file data including a plurality of inquiry items, and the data management device 120 receives the authority inquiry file data sent from the authority terminal 110.

[0191] In step S1302, the data management device 120 divides the received authority inquiry file data for each inquiry item to generate a plurality of divided file data.

[0192] In step S1303, the data management device 120 sends the generated plurality of divided file data to the corresponding staff terminals 130, 140, etc., and the staff terminals 130, 140, etc. receive the divided file data sent from the data management device 120.

[0193] In step S1304, the staff terminals 130, 140, etc. respectively display the received divided file data and respectively receive the reply contents for the inquiry contents input by the staff members 131, 141, etc. As a result, the staff terminals 130, 140, etc. generate each input-completed divided file data.

[0194] In step S1305, the staff terminals 130, 140, etc. send the generated input-completed divided file data to the data management device 120.

[0195] In step S1306, the data management device 120 collects the input-completed divided file data sent from the staff terminals 130, 140, etc.

[0196] In step S1307, the data management device 120 generates input-completed file data by setting the collected input-completed divided file data in a single file format.

[0197] In step S1308, the data management device 120 extracts attribute information from the document data queried from the authority and the input-completed document data.

[0198] In step S1309, the data management device 120 sends the generated input-completed document data to the authority terminal 110.

[0199] In step S1310, the data management device 120 receives the keywords input by the applicant 122 based on the input-completed divided document data.

[0200] In step S1311, the data management device 120 sends the registration data to the server device 150.

[0201] In step S1312, the server device 150 stores the registration data sent from the server device 150 in the database 151.

[0202] <Processing of Business Support System (Retrieval Phase)>

[0203] Next, the process flow of the processing in the retrieval phase of the business support system 200 will be described. Figure 14 It is the first timing diagram showing the process flow of the processing in the retrieval phase of the business support system.

[0204] In step S1401, the staff terminals 130, 140, etc. access the retrieval unit 201 of the server device 150.

[0205] In step S1402, in response to the access from the staff terminals 130, 140, etc., the server device 150 starts to provide a retrieval screen.

[0206] In step S1403, the staff terminals 130, 140, etc. display the retrieval screen provided by the server device 150.

[0207] In step S1404, the staff terminals 130, 140, etc. receive the input of the retrieval conditions of the staff 131, 141, etc.

[0208] In step S1405, the staff terminals 130, 140, etc. receive the retrieval instructions of the staff 131, 141, etc. Thus, the staff terminals 130, 140, etc. send a retrieval request including the input retrieval conditions to the server device 150.

[0209] In step S1406, the server device 150 receives the retrieval request sent from the staff terminals 130, 140, etc., and performs retrieval processing based on the retrieval conditions included in the received retrieval request.

[0210] In step S1407, the server device 150 obtains a search result and provides a search screen including the obtained search result to the staff terminals 130, 140, etc.

[0211] In step S1408, the staff terminals 130, 140, etc. display a search screen including the search result.

[0212] In step S1409, the staff terminals 130, 140, etc. accept an instruction for detailed display from the staff members 131, 141, etc. Thus, the staff terminals 130, 140, etc. send the instruction for detailed display to the server device 150, and in the server device 150, the instruction for detailed display sent from the staff terminals 130, 140, etc. is received.

[0213] In step S1410, the server device 150 obtains detailed information based on the received instruction for detailed display, and provides a search screen including the obtained detailed information to the staff terminals 130, 140, etc.

[0214] In step S1411, the staff terminals 130, 140, etc. display a search screen including the detailed information.

[0215] <Summary>

[0216] As can be seen from the above description, in the accumulation stage of the business support system 100 according to the first embodiment,

[0217] · By dividing a plurality of inquiry items included in the official inquiry document data by each inquiry item, a plurality of divided document data are generated.

[0218] · A plurality of divided document data with reply contents input for each of the plurality of inquiry items are obtained.

[0219] · The input-completed document data, which is in the form of a single file and is generated based on the plurality of obtained input-completed divided document data, is obtained.

[0220] · The plurality of obtained input-completed divided document data are respectively associated with the input-completed document data and stored in the database.

[0221] In this way, in the business support system 100 according to the first embodiment, in the accumulation stage, a database is constructed for enabling simple retrieval of past similar cases of official inquiries. Thus, according to the business support system 100 according to the first embodiment, the business efficiency when the staff conducts a search is improved, and the reply business of the staff can be supported. That is, according to the first embodiment, a structure for supporting the reply business for official inquiries can be provided.

[0222] [Second Embodiment]

[0223] In the above-described first embodiment, the applicant 122 was shown the input-completed divided file data, and the applicant 122 assigned keywords based on the shown input-completed divided file data. However, the method of assigning keywords is not limited to this. For example, it may be configured to assign keywords using a learned keyword model generated by having a model learn past cases. Hereinafter, the second embodiment will be described centering on the differences from the above-described first embodiment.

[0224] <System Structure of Business Support System>

[0225] First, the system structure of the business support system according to the second embodiment will be described. In the second embodiment, since the business support system executes the processing in the learning phase and the processing in the accumulation phase, the system structure will be described separately for each phase here.

[0226] (1) Business Support System in Learning Phase

[0227] Figure 15 FIG. 3, which shows an example of the system structure of the business support system, shows the system structure in the learning phase. The learning phase refers to a phase that is executed separately from the response service to the authority inquiry from the authority, and is a phase for generating a learned keyword model for automatically assigning keywords. The difference from the business support system 100 in the accumulation phase shown in Figure 1 is that in the case of the business support system 1500 in the learning phase, it has a learning device 1510.

[0228] A learning program is installed in the learning device 1510, and by executing this program in the learning phase, the learning device 1510 functions as a learning unit 1511.

[0229] The learning unit 1511 reads out the registered data stored in the database 151 of the server device 150 and generates a learning data set. The learning data set generated by the learning unit 1511 is a learning data set that uses the input-completed divided file data as input data and the corresponding keyword as ground truth data.

[0230] In addition, the learning unit 1511 uses the generated learning data set to perform a learning process on a keyword model (a model for predicting keywords) and generates a learned keyword model.

[0231] (2) Business Support System in Accumulation Phase

[0232] Figure 16FIG. 4 shows an example of the system configuration of the business support system, showing the system configuration during the accumulation phase.

[0233] The difference from the system configuration during the learning phase is that, in the case of the business support system 1600 during the accumulation phase, the function of the management unit 1601 of the data management device 120 is different, and a prediction device 1610 is newly provided.

[0234] The management unit 1601 of the data management device 120 obtains keywords suitable for the input-completed divided file data from the prediction device 1610.

[0235] A prediction program is installed in the prediction device 1610, and by executing this program during the accumulation phase, the prediction device 1610 functions as a prediction unit 1611.

[0236] The prediction unit 1611 includes a learned keyword model generated by the learning unit 1511. Whenever the authority query file data is sent from the authority terminal 110 and the input-completed divided file data is collected by the data management device 120, the prediction unit 1611 predicts keywords and notifies the data management device 120.

[0237] Specifically, whenever the input-completed divided file data collected is notified from the management unit 1601 of the data management device 120, the prediction unit 1611 predicts keywords suitable for the input-completed divided file data and notifies the management unit 1601 of the data management device 120.

[0238] In addition, the keywords notified from the prediction unit 1611 are associated with the input-completed divided file data and sent to the server device 150. Thus, in the server device 150, the registered data with keywords assigned by the prediction unit 1611 can be stored in the database 151.

[0239] <Function Structure of the Learning Unit of the Learning Device>

[0240] Next, the function structure of the learning device 1510 will be described. Figure 17 It is a figure showing an example of the function structure of the learning unit of the learning device. As Figure 17 shown, the learning unit 1511 reads out the registered data 1000 from the database 151 and generates a learning dataset. Figure 17 The example of

[0241] · Based on the read registered data 1000, each query content and response content included in the input-completed divided file data are used as input data,

[0242] ·Set the keywords recalled from each query content and response content included in the file data divided according to the input completion as the correct answer data.

[0243] In addition, as Figure 17 shown, the learning department 1511 has a morpheme analysis department 1711, a vectorization model 1712, a keyword model 1713, and a comparison / change department 1714.

[0244] The morpheme analysis department 1711 reads the input completion divided file data included in the learning dataset, and performs morpheme analysis on the article included in the query content and response content of the input completion divided file data. In addition, the morpheme analysis department 1711 notifies the analysis result of the morpheme analysis to the vectorization model 1712.

[0245] The vectorization model 1712 uses TF-IDF (Term Frequency-Inverse Document Frequency), LSI (Latent Semantic Indexing), LDA (Latent Dirichlet Allocation), Doc2Vec, Sent2Vec, RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), GRU (Gated-Recurrent Unit), USE (Universal Sentence Encoder), BERT (Bidirectional Encoder Representations from Transformers), Sentence-Transformers, etc., and vectorization techniques obtained by combining them respectively, processes the analysis result notified from the morpheme analysis department 1711, and generates vector data. In addition, the vectorization model 1712 notifies the vector data to the keyword model 1713.

[0246] The keyword model 1713 processes the vector data notified from the vectorization model 1712 using classification techniques such as support vector machines, decision trees, k-means, nearest neighbors, logistic regression, random forests, gradient boosting decision trees, neural networks, deep learning (e.g., RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), GRU (Gated-Recurrent Unit), Seq2seq, Sentence-Transformers, etc.) and combinations thereof, and outputs output data.

[0247] The comparison / change unit 1714 compares the output data output from the keyword model 1713 with the keywords included in the learning dataset, and updates the model parameters of the keyword model 1713 to reduce the error between the two. Thus, the learning unit 1511 can generate a learned keyword model.

[0248] <Function Structure of the Management Unit of the Data Management Device and Function Structure of the Prediction Device>

[0249] Next, the function structure of the management unit 1601 of the data management device 120 and the function structure of the prediction unit 1611 of the prediction device 1610 will be described. Figure 18 It is a diagram showing an example of the function structure of the management unit of the data management device and the function structure of the prediction unit of the prediction device.

[0250] The management unit 1601 of the data management device 120 is the same as Figure 5 the management unit 121 shown, and includes:

[0251] · File data acquisition unit 501,

[0252] · Division unit 503,

[0253] · Allocation unit 504,

[0254] · Attribute information extraction unit 505,

[0255] · Collection unit 506,

[0256] · File data generation unit 507,

[0257] · File data transmission unit 508,

[0258] · Storage unit 1801.

[0259] Among them, the document data acquisition unit 501 to the document data transmission unit 508 have already been used Figure 5 The description is completed, so the description is omitted here.

[0260] If the collection unit 506 notifies the input-completed divided document data, the storage unit 1801 notifies the notified input-completed divided document data to the prediction device 1610. In addition, the storage unit 1801 acquires the keywords notified from the prediction device 1610 in response to the notification of the input-completed divided document data.

[0261] In addition, similar to the storage unit 509, the storage unit 1801 stores the registration data in the database 151 of the server device 150. In addition, the registration data includes:

[0262] · The input-completed divided document data notified from the collection unit 506,

[0263] · The input-completed document data notified from the document data generation unit 507,

[0264] · The attribute information notified from the attribute information extraction unit 505,

[0265] · The keywords notified from the prediction device 1610,...

[0266] etc.

[0267] The prediction unit 1611 of the prediction device 1610 includes:

[0268] · The morpheme analysis unit 1711,

[0269] · The vectorization model 1712,

[0270] · The learned keyword model 1802.

[0271] Among them, the morpheme analysis unit 1711 and the vectorization model 1712 have already been used Figure 17 The description is completed, so the description is omitted here.

[0272] The learned keyword model 1802 is a model generated by the learning unit 1511. If the vector data is notified from the vectorization model 1712, the keyword is predicted. The keyword predicted by the learned keyword model 1802 is notified to the storage unit 1801.

[0273] <Business support system processing (learning stage)>

[0274] Next, the processing flow of the business support system 1500 in the learning stage will be described. Figure 19 It is a timing chart showing the processing flow in the learning stage of the business support system.

[0275] In step S1901, the server device 150 reads the registration data stored in the database 151 and sends it to the learning device 1510.

[0276] In step S1902, the learning device 1510 acquires the sent registration data.

[0277] In step S1903, the learning device 1510 generates a learning data set based on the acquired registration data.

[0278] In step S1904, the learning device 1510 uses the generated learning data set to perform a learning process on the keyword model 1713 and generates a learned keyword model.

[0279] In step S1905, the learning device 1510 stores the generated learned keyword model together with the morpheme analysis unit 1711 and the vectorization model 1712 in the storage unit of the prediction device 1610.

[0280] <Process of Business Support System (Accumulation Phase)>

[0281] Next, the process flow of the business support system 1600 in the accumulation phase will be described. Figure 20 It is the second timing chart showing the process flow in the accumulation phase of the business support system.

[0282] In addition, Figure 20 each step of steps S1301 to S1309, S1311, and S1312 of Figure 13 is the same as each step of steps S1301 to S1309, S1311, and S1312 of

[0283] In step S2001, the prediction device 1610 predicts keywords based on the input-completed divided file data notified from the data management device 120 and notifies the predicted keywords to the data management device 120.

[0284] <Summary>

[0285] As can be seen from the above description, in the learning phase of the business support system 1500 according to the second embodiment,

[0286] · Generate a learning data set that includes each query content and response content read from the database and included in the input-completed divided file data, and keywords recalled based on each query content and response content included in the input-completed divided file data.

[0287] ·Generate a learned keyword model by performing a learning process using a learning dataset.

[0288] In addition, in the accumulation stage, the business support system 1600 according to the second embodiment

[0289] ·Predict keywords that are recalled based on each inquiry content and response content included in the input-completed file data by using the learned keyword model.

[0290] In this way, in the business support system 1600 according to the second embodiment, it is configured to automatically assign keywords in the accumulation stage when constructing a database for easily retrieving past similar cases of authority inquiries.

[0291] Accordingly, according to the second embodiment, the business load on the applicant when storing registration data in the database can be reduced.

[0292] [Third Embodiment]

[0293] In the above first and second embodiments, for example, the case where the retrieval items shown in the retrieval screen 1210 where the retrieval conditions are input by the staff members 131, 141, etc. are used to retrieve past cases similar to the inquiry content included in the authority inquiry file data has been described. Figure 12 However, the retrieval method when retrieving past cases similar to the inquiry content included in the authority inquiry file data is not limited to this. For example, it may be configured to retrieve past similar cases by inputting the inquiry content itself included in the authority inquiry file data. Hereinafter, the third embodiment will be described centering on the differences from the above first and second embodiments.

[0294] <System Structure of Business Support System>

[0295] First, as the business support system according to the third embodiment, the system structure of the business support system in the retrieval stage will be described.

[0296] FIG. 5 is an example showing the system structure of the business support system. Figure 21 It is different from the business support system 200 in the retrieval stage shown in

[0297] In the case of the business support system 2100 in the retrieval stage shown in Figure 2 that, when a retrieval request is made, the inquiry content itself included in the authority inquiry file data of the processing object is input. Figure 21

[0298] In addition, it is different from the business support system 200 in the retrieval stage shown in In the case of the business support system 2100 in the retrieval stage shown in Figure 2 that, when a retrieval request is made, the inquiry content itself included in the authority inquiry file data of the processing object is input.Figure 21 In the case of the business support system 2100 in the retrieval stage shown, the function of the retrieval unit 2101 is different from the function of the retrieval unit 201. Specifically, in the case of the retrieval unit 2101, the inquiry content included in the retrieval request (the inquiry content of the authority inquiry file data to be processed) is vectorized to generate vector data. In addition, in the case of the retrieval unit 2101, the inquiry content included in each input-completed divided file data stored in the database 151 is vectorized to generate vector data. Furthermore, in the case of the retrieval unit 2101, by calculating the similarity degree of this vector data, the input-completed divided file data including the inquiry content similar to the inquiry content included in the retrieval request is retrieved.

[0299] <Function Structure of Retrieval Unit of Server Device>

[0300] Next, the function structure of the retrieval unit 2101 of the server device 150 will be described. Figure 22 This is the second figure showing an example of the function structure of the retrieval unit of the server device. As described above, in the retrieval stage of the third embodiment, the server device 150 functions as the retrieval unit 2101. In addition, as Figure 22 shown, the retrieval unit 2101 further includes:

[0301] · A retrieval screen providing unit 2201,

[0302] · An extraction unit 2202,

[0303] · A first vectorization unit 2203,

[0304] · A separation unit 2204,

[0305] · A second vectorization unit 2205,

[0306] · A similarity calculation unit 2206,

[0307] · An output unit 2207.

[0308] In the case of access from staff terminals 130, 140, etc., the retrieval screen providing unit 2201 provides a retrieval screen. In addition, the retrieval screen providing unit 2201 receives retrieval requests from staff terminals 130, 140, etc. in response to the provision of the retrieval screen. In addition, when the retrieval result is notified from the output unit 2207, the retrieval screen providing unit 2201 includes the retrieval result in the retrieval screen and provides it to staff terminals 130, 140, etc. In addition, the retrieval screen providing unit 2201 receives an instruction for detailed display from staff terminals 130, 140, etc. in response to the provision of the retrieval screen including the retrieval result. Furthermore, when the detailed information is notified from the output unit 2207, the retrieval screen providing unit 2201 includes the detailed information in the retrieval screen and provides it to staff terminals 130, 140, etc.

[0309] The extraction unit 2202 extracts the inquiry content included in the retrieval request received by the retrieval screen providing unit 2201. In addition, the extraction unit 2202 notifies the extracted inquiry content to the first vectorization unit 2203.

[0310] The first vectorization unit 2203 vectorizes the inquiry content notified from the extraction unit 2202 and generates vector data. In addition, the first vectorization unit 2203 notifies the generated vector data to the similarity calculation unit 2206.

[0311] The separation unit 2204 reads the input-completed partition file data from the database 151 and separates the read input-completed partition file data into inquiry content and response content. In addition, the separation unit 2204 notifies the inquiry content in the separated inquiry content and response content to the second vectorization unit 2205.

[0312] The second vectorization unit 2205 vectorizes the inquiry content notified from the separation unit 2204 and generates vector data. In addition, the second vectorization unit 2205 notifies the generated vector data to the similarity calculation unit 2206.

[0313] The similarity calculation unit 2206 calculates the similarity between the vector data notified from the first vectorization unit 2203 and the vector data notified from the second vectorization unit 2205. In addition, the similarity calculation unit 2206 notifies the calculated similarity to the output unit 2207.

[0314] The output unit 2207 retrieves the similarity that satisfies the specified similarity condition (for example, the highest similarity or the top n (n is an arbitrary number) similarities) from the similarities notified by the similarity calculation unit 2206, and determines the corresponding input-completed partition file data. In addition, the output unit 2207 notifies the determined input-completed partition file data as the retrieval result to the retrieval screen providing unit 2201.

[0315] In addition, when an instruction for detailed display received in the retrieval screen providing unit 2201 is obtained, the output unit 2207 acquires registration data other than the input-completed divided file data (attribute information, input-completed file data, keywords, etc.) notified as the retrieval result from the database 151. In addition, the output unit 2207 notifies the registration data other than the input-completed divided file data (attribute information, input-completed file data, keywords, etc.) notified as the retrieval result to the retrieval screen providing unit 2201 as detailed information. That is, the output unit 2207 notifies the data other than the input-completed divided file data included in the registration data corresponding to the retrieved query content to the retrieval screen providing unit 2201 as detailed information.

[0316] <Explanation of Retrieval Screen>

[0317] Next, an explanation will be given of the retrieval screen provided by the server device 150 to the staff terminals 130, 140, etc. Figure 23 FIG. 2 shows an example of the retrieval screen displayed on the staff terminal.

[0318] Among them, the retrieval screen 2310 is a screen for inputting query content and sending a retrieval request. As Figure 23 shown, the retrieval screen 2310 includes a column for inputting query content.

[0319] Staff members 131, 141, etc. input the query content in the column for inputting query content, for example, by copying and pasting the query content included in the authority query file data of the processing object (actually the query content included in the divided file data sent from the data management device 120).

[0320] In addition, the retrieval screen 2310 includes a retrieval button 1211. If the input of the query content in the column for inputting query content is completed and the retrieval button 1211 is pressed, a retrieval request including the query content is sent to the server device 150.

[0321] On the other hand, the retrieval screen 1220 is a screen displayed on the staff terminals 130, 140, etc. when a retrieval request is sent, the server device 150 retrieves the input-completed divided file data including query content similar to the query content included in the retrieval request, and provides a retrieval screen including the retrieval result.

[0322] In addition, Figure 23 the retrieval screen 1220 shown is the same as Figure 12 the retrieval screen 1220 shown, so the explanation is omitted here.

[0323] <Specific Example of Processing of Each Unit Included in the Retrieval Unit of the Server Device>

[0324] Next, a specific example of the processing of each part included in the retrieval unit 2101 of the server device 150 (here, the first vectorization unit 2203, the separation unit 2204, and the second vectorization unit 2205) will be described. Figure 24 FIG. 1 is a diagram showing a specific example of the processing of the first vectorization unit, the separation unit, and the second vectorization unit included in the retrieval unit of the server device.

[0325] As Figure 24 shown in (a), the first vectorization unit 2203 includes a morpheme analysis unit 2401 and a vectorization model 2402.

[0326] When the interrogation content is notified from the extraction unit 2202, the morpheme analysis unit 2401 performs morpheme analysis on each article included in the notified interrogation content using the pharmaceutical development term dictionary stored in the pharmaceutical development term dictionary storage unit 2403. In addition, the pharmaceutical development term dictionary refers to a dictionary containing inherent terms in pharmaceutical development. Further, the morpheme analysis unit 2401 notifies the analysis result obtained by the morpheme analysis to the vectorization model 2402.

[0327] The vectorization model 2402 is an example of a vectorization unit, and generates vector data by processing the analysis result of the morpheme analysis notified from the morpheme analysis unit 2401. Specifically, in the vectorization model 2402, for example, by using TF-IDF or the like, vectorization is performed with values corresponding to the appearance frequencies of each word extracted from the interrogation content to generate vector data. Specifically, in the generation of vector data, TF-IDF, LSI (Latent Semantic Indexing), LDA (Latent Dirichlet Allocation), Doc2Vec, Sent2Vec, Word2Vec, RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), GRU (Gated-Recurrent Unit), USE (Universal Sentence Encoder), BERT (Bidirectional Encoder Representations from Transformers), Sentence-Transformers, etc., and vectorization techniques obtained by combining them can also be used.

[0328] As Figure 24As shown in (b), the separation unit 2204 sequentially reads out the input-completed partition file data from the database 151, separates the query content from the read input-completed partition file data, and notifies the separated query content to the second vectorization unit 2205.

[0329] The second vectorization unit 2205 includes a morpheme analysis unit 2411 and a vectorization model 2412. When the query content is notified from the separation unit 2204, the morpheme analysis unit 2411 uses the drug development term dictionary stored in the drug development term dictionary storage unit 2413 to perform morpheme analysis on each article included in the notified query content. In addition, the morpheme analysis unit 2411 notifies the analysis result obtained by the morpheme analysis to the vectorization model 2412.

[0330] The vectorization model 2412 is an example of a vectorization unit, and generates vector data by processing the analysis result of the morpheme analysis notified from the morpheme analysis unit 2411. Specifically, in the vectorization model 2412, for example, by using TF-IDF or the like, vectorization is performed with values corresponding to the occurrence frequencies of the respective words extracted from the query content to generate vector data. Specifically, in the generation of vector data, TF-IDF, LSI (Latent Semantic Indexing), LDA (Latent Dirichlet Allocation), Doc2Vec, Sent2Vec, Word2Vec, RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), GRU (Gated-Recurrent Unit), USE (Universal Sentence Encoder), BERT (Bidirectional Encoder Representations from Transformers), Sentence-Transformers, etc. and vectorization techniques obtained by combining them can also be used.

[0331] In Figure 24 For ease of explanation, the vectorization model and the morpheme analysis unit are shown as two functional blocks. However, the vectorization model and the morpheme analysis unit for processing the query content included in the retrieval request and the vectorization model and the morpheme analysis unit for processing the query content separated from the input-completed partition file data may use different vectorization models and morpheme analysis units, but preferably the same vectorization model and morpheme analysis unit.

[0332] Similarly, in Figure 24 , for ease of explanation, two drug development term dictionary storage units are shown. However, the drug development term dictionary used when processing the query content included in the retrieval request and the drug development term dictionary used when processing the query content separated from the input-completed divided file data may be different, but preferably the same dictionary.

[0333] <Business Support System Processing (Retrieval Phase)>

[0334] Next, the process of the processing in the retrieval phase of the business support system 2100 will be described. Figure 25 is the second timing chart showing the process of the processing in the retrieval phase of the business support system. The difference from the processing in the retrieval phase of the business support system 200 described using Figure 14 lies in steps S2501 to S2503.

[0335] In step S2501, the staff terminals 130, 140, etc. receive the input of the query content from the staff members 131, 141, etc.

[0336] In step S2502, the staff terminals 130, 140, etc. receive the retrieval instructions from the staff members 131, 141, etc. Thereby, the staff terminals 130, 140, etc. send a retrieval request including the input query content to the server device 150.

[0337] In step S2503, the server device 150 receives the retrieval request sent from the staff terminals 130, 140, etc., and retrieves the input-completed divided file data including query content similar to the query content included in the received retrieval request.

[0338] <Summary>

[0339] From the above description, it can be seen that in the retrieval phase, the business support system 2100 according to the third embodiment

[0340] · uses a dictionary containing the inherent terms in drug development to perform morpheme analysis on the query content included in the authority query file data to be processed.

[0341] · Based on each word extracted from the query content included in the authority query file data to be processed through morpheme analysis, vectorize the query content included in the authority query file data to be processed.

[0342] · Calculate the similarity with the vector data generated according to the query content included in the authority query file data to be processed through vectorization, and retrieve the input-completed divided file data including similar query content.

[0343] In this way, in the business support system 2100 according to the third embodiment, it is configured to input the inquiry content itself instead of the search conditions during the search phase to perform a search, so that past similar cases of the authority inquiry can be retrieved simply.

[0344] Accordingly, in the business support system 2100 according to the third embodiment, the work efficiency of the staff during the search is improved, and the reply work of the staff can be supported. That is, according to the third embodiment, a structure that supports the reply work for the authority inquiry can be provided.

[0345] [Fourth Embodiment]

[0346] In the above third embodiment, it is configured to retrieve the input-completed divided file data including similar inquiry content by inputting the inquiry content included in the authority inquiry file data. However, the retrieval method of the input-completed divided file data including similar inquiry content is not limited to this. For example, it may be configured to input predetermined search conditions, and after narrowing down the input-completed divided file data to be retrieved, retrieve the input-completed divided file data including inquiry content similar to the input inquiry content. Thereby, the retrieval accuracy can be improved. Hereinafter, the fourth embodiment will be described centering on the differences from the above third embodiment.

[0347] <Functional Structure of the Retrieval Unit of the Server Device>

[0348] First, the functional structure of the retrieval unit 2101 of the server device 150 according to the fourth embodiment will be described. Figure 26 FIG. 3 is an example showing the functional structure of the retrieval unit of the server device. The difference from the Figure 22 shown functional structure is that in the case of the retrieval unit 2101 shown in Figure 26 , the function of the extraction unit 2601 is different from that of the extraction unit 2202 shown in Figure 22 . In addition, the difference from the Figure 22 shown functional structure is that in the case of the retrieval unit 2101 shown in Figure 26 , it has a filtering unit 2602.

[0349] The extraction unit 2601 extracts the search conditions (in this embodiment, used for narrowing down the retrieval object, so it is called a filtering condition) and the inquiry content included in the retrieval request received by the retrieval screen providing unit 2201. In addition, the extraction unit 2601 notifies the extracted filtering condition to the filtering unit 2602, and notifies the extracted inquiry content to the first vectorization unit 2203.

[0350] If the extraction unit 2601 notifies the filtering conditions, the filtering unit 2602 retrieves the input-completed divided file data in the database 151 that matches the filtering conditions. In addition, the filtering unit 2602 notifies the obtained input-completed divided file data to the separation unit 2204.

[0351] As a result, in the separation unit 2204, the inquiry content included in the input-completed divided file data obtained by the filtering unit 2602 is notified to the second vectorization unit 2205, and the object for calculating the similarity in the similarity calculation unit 2206 can be reduced.

[0352] <Explanation of the search screen>

[0353] Next, the search screen provided by the server device 150 according to the fourth embodiment to the staff terminals 130, 140, etc. will be described. Figure 27 Fig. 3 is an example of the search screen displayed on the staff terminal.

[0354] Among them, the search screen 2710 is a screen for inputting inquiry content and filtering conditions and sending a search request.

[0355] As Figure 27 shown, the search screen 2710 includes a column for inputting inquiry content and a column for inputting filtering conditions. The staff members 131, 141, etc. input the inquiry content in the column for inputting inquiry content by, for example, copying and pasting the inquiry content included in the authority inquiry file data.

[0356] In addition, as Figure 27 shown, in the search screen 2710, as filtering conditions, it includes "string", "period of authority inquiry", "disease field", "event", "variety", "reply period". In addition, the "string", "period of authority inquiry", "disease field", "event", "variety", "reply period" input as filtering conditions have been described, so the description is omitted here. Figure 12 The description is completed, so the description is omitted here.

[0357] On the other hand, the search screen 1220 is a screen displayed on the staff terminals 130, 140, etc. when a search request is sent, and the server device 150 retrieves the input-completed divided file data including inquiry content similar to the inquiry content included in the search request and provides a search screen including the search results.

[0358] In addition, Figure 27 the search screen 1220 shown is the same as the Figure 12 search screen 1220 shown, so the description is omitted here.

[0359] <Summary>

[0360] As can be seen from the above description, in the retrieval stage of the business support system 2100 according to the fourth embodiment:

[0361] · Retrieve the inquiry content included in the input-completed divided file data that meets the specified filtering conditions among the multiple input-completed divided file data stored in the database 151.

[0362] · The filtering conditions include the inquiry period of the authority inquiry, the reply period for the authority inquiry, the disease field, the events being inquired about, and the variety.

[0363] In this way, in the business support system 2100 according to the fourth embodiment, it is configured to perform a retrieval in the retrieval stage by inputting the filtering conditions together with the input inquiry content itself, so as to be able to simply retrieve past similar cases of authority inquiries. Thus, according to the fourth embodiment, the retrieval accuracy when the staff performs a retrieval can be improved. In particular, in the past, it was necessary to select appropriate retrieval terms and use Boolean operators, etc., which are familiar with retrieval techniques during the retrieval, but now it is possible to simply retrieve past similar cases only by inputting the authority inquiry.

[0364] [Fifth Embodiment]

[0365] In the business support systems according to the above-described third and fourth embodiments, the following structure is adopted: The input-completed divided file data containing inquiry content similar to the inquiry content included in the authority inquiry file data is displayed as a retrieval result on the staff terminals 130, 140, etc. However, the structure of the business support system is not limited to this. For example, it may also be configured as follows: When the retrieval result is displayed on the staff terminals 130, 140, etc., an input of an evaluation for the retrieval result is accepted, and the accepted evaluation is fed back. Hereinafter, the fifth embodiment will be described centering on the differences from the above-described third and fourth embodiments.

[0366] <System Structure of Business Support System>

[0367] First, the system structure of the business support system according to the fifth embodiment will be described. In the fifth embodiment, the business support system performs the processing in the feedback stage. Figure 28 FIG. 6 is an example showing the system structure of the business support system. The feedback stage refers to a stage that is executed separately from the reply service for the authority inquiry from the authority, and is a stage for updating the retrieval unit 2101 in order to improve the retrieval accuracy when a certain amount of feedback information has been accumulated.

[0368] And Figure 1The difference in the business support system 100 during the accumulation phase shown is that in the case of the business support system 2800 during the feedback phase, it has an analysis device 2810. In addition, it is configured such that in the case of the business support system 2800 during the feedback phase, evaluations of the retrieval results are input by the staff members 131 and 141.

[0369] An analysis program is installed in the analysis device 2810, and by executing this program during the feedback phase, the analysis device 2810 functions as an analysis unit 2811. The analysis unit 2811 can also execute various processes under the operation instructions of the analyst 2822.

[0370] Specifically, the analysis unit 2811 obtains feedback information from the staff terminals 130, 140, etc. In addition, based on the obtained feedback information, the analysis unit 2811 updates the drug development term dictionaries stored in the drug development term dictionary storage units 2403 and 2413 of the retrieval unit 2101.

[0371] In addition, based on the obtained feedback information, the analysis unit 2811 updates the hyperparameters of the vectorization models 2402 and 2412 of the retrieval unit 2101.

[0372] In addition, based on the obtained feedback information, the analysis unit 2811 updates the vectorization models 2402 and 2412 of the retrieval unit 2101 themselves to new learned vectorization models.

[0373] In addition, the analysis unit 2811 realizes the above updates by sending update information to the server device 150.

[0374] If the input of an evaluation of the retrieval result is accepted, the staff terminals 130, 140, etc. obtain the input evaluation result and the corresponding information. In addition, the information corresponding to the evaluation result includes the query content included in the retrieval request, the query content of the retrieval result, keywords, categories, time periods, etc. corresponding to the query content and the reply content of the retrieval result. In addition, the staff terminals 130, 140, etc. send the input evaluation result and the corresponding information (keywords, categories, time periods, etc.) as feedback information to the analysis device 2810.

[0375] If update information is sent from the analysis device 2810, the server device 150 updates the retrieval unit 2101 according to the sent update information.

[0376] <Summary of the processing of the business support system (feedback phase)>

[0377] Next, a summary of the processing of the business support system 2800 during the feedback phase will be described. Figure 29This is a diagram for explaining an overview of processing in the feedback phase of a business support system.

[0378] In Figure 29 , the time axis 2901 shows the content and timing of operations performed by staff members 131, 141, etc. on staff terminals 130, 140, etc. during the accumulation phase. As Figure 29 shown, during the accumulation phase, staff members 131, 141, etc. repeatedly perform retrieval operations and input operations for evaluation results of retrieval results. Additionally, in staff terminals 130, 140, etc., each time an evaluation result input operation is performed, feedback information is sent to the analysis device 2810.

[0379] In Figure 29 , the time axis 2902 shows the content and timing of processing performed by the analysis device 2810 during the feedback phase.

[0380] As Figure 29 shown, during the feedback phase, each time feedback information is sent from staff terminals 130, 140, etc., the analysis device 2810 analyzes the drug development term dictionary, adds new terms, and deletes unnecessary terms.

[0381] Furthermore, during the feedback phase, if the analysis of the drug development term dictionary is performed for a certain period, the analysis device 2810 uses a certain amount of accumulated feedback information to analyze the vectorization models 2402, 2412. Specifically, the analysis device 2810 adjusts the hyperparameters of the vectorization models 2402, 2412.

[0382] Moreover, during the feedback phase, if the analysis of the vectorization models 2402, 2412 is performed for a certain period and the hyperparameter adjustment is carried out, the analysis device 2810 uses a certain amount of accumulated feedback information to generate a learning dataset. Furthermore, the analysis device 2810 performs a learning process by using the generated learning dataset to generate a new learned vectorization model, which is used to generate vector data based on query content.

[0383] In Figure 29 , the time axis 2903 shows the update timing of the drug development term dictionary storage units 2403, 2413 for the server device 150.

[0384] As Figure 29 shown, during the feedback phase, the analysis device 2810 sends an instruction to add new terms and delete unnecessary terms as update information. Thereby, in the retrieval unit 2101 of the server device 150, the drug development term dictionary is updated.

[0385] In Figure 29In this, the timeline 2904 shows the update content and update timing of the vectorization models 2402 and 2412 of the server device 150.

[0386] As Figure 29 shown, in the feedback phase, the adjusted hyperparameters are sent as update information from the analysis device 2810. Accordingly, in the retrieval unit 2101 of the server device 150, the hyperparameters of the vectorization models 2402 and 2412 are updated to the adjusted hyperparameters.

[0387] Furthermore, in the feedback phase, the learned vectorization model is sent as update information from the analysis device 2810. Accordingly, in the retrieval unit 2101 of the server device 150, the vectorization models 2402 and 2412 are updated to the learned vectorization model.

[0388] <Functional Structure of the Analysis Unit of the Analysis Device>

[0389] Next, the functional structure of the analysis unit 2811 of the analysis device 2810 will be described. Figure 30 is a diagram showing an example of the functional structure of the analysis unit of the analysis device. As described above, in the feedback phase, the analysis device 2810 functions as the analysis unit 2811. In addition, as Figure 30 shown, the analysis unit 2811 further includes:

[0390] · Feedback information acquisition unit 3010,

[0391] · Additional term extraction unit 3021,

[0392] · Deleted term extraction unit 3022,

[0393] · Dictionary update unit 3023 (an example of the first update unit),

[0394] · Hyperparameter update unit 3031 (an example of the second update unit),

[0395] · Learning data generation unit 3041,

[0396] · Learning unit 3042,

[0397] · Model update unit 3043 (an example of the third update unit).

[0398] The feedback information acquisition unit 3010 acquires the feedback information notified from the staff terminals 130, 140, etc. The feedback information notified from the staff terminals 130, 140, etc. includes:

[0399] · The query content included in the retrieval request,

[0400] · Input completion of search results, dividing the query content included in the file data,

[0401] · Evaluation results,

[0402] · Keywords associated with the input completion division file data of the search results,

[0403] · Categories associated with the input completion division file data of the search results,

[0404] · Query period and response period associated with the input completion division file data of the search results.

[0405] The additional term extraction unit 3021 extracts terms to be added to the drug development term dictionary based on the feedback information obtained by the feedback information acquisition unit 3010. In addition, the additional term extraction unit 3021 notifies the extracted terms to the dictionary update unit 3023.

[0406] The deletion term extraction unit 3022 extracts terms to be deleted from the drug development term dictionary based on the feedback information obtained by the feedback information acquisition unit 3010. In addition, the deletion term extraction unit 3022 notifies the extracted terms to the dictionary update unit 3023.

[0407] The dictionary update unit 3023 instructs the server device 150 to add the terms notified from the additional term extraction unit 3021 to the drug development term dictionary. In addition, the dictionary update unit 3023 adds the terms notified from the additional term extraction unit 3021 to the drug development term dictionary storage unit 3024. Additionally, in the analysis unit 2811, it is set to store the latest drug development term dictionary stored in the drug development term dictionary storage units 2403 and 2413 of the retrieval unit 2101 of the server device 150 in the drug development term dictionary storage unit 3024 as well.

[0408] The hyperparameter update unit 3031 adjusts the hyperparameters of the vectorization model 2402 and the vectorization model 2412 based on the feedback information accumulated in the feedback information storage unit 3030.

[0409] In addition, the hyperparameter update unit 3031 instructs the server device 150 to update the vectorization model 2402 of the first vectorization unit 2203 and the vectorization model 2412 of the second vectorization unit 2205 with the adjusted hyperparameters.

[0410] The learning data generation unit 3041 generates a learning data set based on the feedback information acquired by the feedback information acquisition unit 3010 and accumulated in the feedback information storage unit 3030. Specifically, the learning data generation unit 3041 generates a learning data set that uses the query content included in the retrieval request and the query content included in the input-completed partitioned file data of the retrieval result as input data, and the evaluation result as correct answer data.

[0411] The learning unit 3042 uses the learning data set generated by the learning data generation unit 3041 to perform a learning process on the vectorization model. In addition, in the learning unit 3042, when the input data of the learning data set is input to the vectorization model, morphological analysis is performed. At this time, in the learning unit 3042, the latest pharmaceutical development term dictionary stored in the pharmaceutical development term dictionary storage unit 3024 is used for morphological analysis.

[0412] The model update unit 3043 instructs the server device 150 to update the vectorization model 2402 and the vectorization model 2412 with the learned vectorization model, which is generated by the learning process performed by the learning unit 3042.

[0413] In this way, the update information sent from the analysis unit 2811 to the server device 150 includes:

[0414] · Instructions to append terms to the pharmaceutical development term dictionary, instructions to delete terms from the pharmaceutical development term dictionary,

[0415] · Instructions to update the hyperparameters of the vectorization model 2402 of the first vectorization unit 2203 and the vectorization model 2412 of the second vectorization unit 2205,

[0416] · Instructions to update the vectorization model 2402 of the first vectorization unit 2203 and the vectorization model 2412 of the second vectorization unit 2205 to a new learned vectorization model.

[0417] <Specific example of the processing of the learning unit included in the analysis unit of the analysis device>

[0418] Next, a specific example of the processing of the learning unit 3042 included in the analysis unit 2811 of the analysis device 2810 will be described. Figure 31 It is a diagram showing a specific example of the processing of the learning unit included in the analysis unit of the analysis device.

[0419] As Figure 31 shown, in the learning data set 3100, items including "input data" and "correct answer data" are included as information.

[0420] In the "input data", there are also included a "searching side" and "search results". In the "searching side", the query content included in the retrieval request in the feedback information is stored. In the "search results", the query content included in the input-completed divided file data of the retrieval results in the feedback information is stored.

[0421] In the "correct answer data", the evaluation result in the feedback information is stored. In Figure 31 the example of , the case where the evaluation result is stored as "good" or "no" is shown, but the evaluation result can also be stored as a numerical value from 0 to 100, for example.

[0422] In addition, as Figure 31 shown, the learning unit 3042 also includes a morpheme analysis unit 3101, a vectorization model 3102, a similarity calculation unit 3103, and a comparison / modification unit 3104.

[0423] The input data of the learning dataset 3100 is input to the morpheme analysis unit 3101. In the morpheme analysis unit 3101, morpheme analysis is performed on the input input data (the query content included in the retrieval request, the query content included in the input-completed divided file data of the retrieval results). In addition, in the morpheme analysis unit 3101, when performing morpheme analysis, the latest pharmaceutical development term dictionary stored in the pharmaceutical development term dictionary storage unit 3024 is used.

[0424] The analysis result of the morpheme analysis of the input data is input to the vectorization model 3102, and vector data is output. Specifically, if the analysis result of the morpheme analysis of the query content included in the retrieval request is input, the vectorization model 3102 outputs the vector data of the query content included in the retrieval request. In addition, if the analysis result of the morpheme analysis of the query content included in the input-completed divided file data of the retrieval results is input, the vectorization model 3102 outputs the vector data of the query content included in the input-completed divided file data of the retrieval results.

[0425] In addition, the type of the vectorization model 3102 is arbitrary. In the vectorization model 3102, for example, Transformers such as Sentence-Transformers can also be used. Alternatively, in the vectorization model 3102, TF-IDF, LSI (Latent Semantic Indexing), LDA (Latent Dirichlet Allocation), Doc2Vec, Sent2Vec, Word2Vec, RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), GRU (Gated-Recurrent Unit), USE (Universal Sentence Encoder), BERT (Bidirectional Encoder Representations from Transformers), Sentence-Transformers, etc. can also be used, as well as technologies obtained by combining them.

[0426] The similarity calculation unit 3103 calculates the similarity between the vector data of the query content included in the retrieval request output from the vectorization model 3102 and the vector data of the query content included in the input-completed partition file data of the retrieval result. In addition, the similarity calculation unit 3103 notifies the calculated similarity to the comparison / change unit 3104.

[0427] The comparison / change unit 3104 reads out the evaluation result corresponding to the input data from the similarity notified from the similarity calculation unit 3103 and the evaluation result stored in the "correct answer data" of the learning dataset 3100, and compares it with the notified similarity.

[0428] In addition, when the correct answer data is "good", the comparison / change unit 3104 updates the model parameters of the vectorization model 3102 so that the similarity notified from the similarity calculation unit 3103 becomes higher.

[0429] In addition, when the correct answer data is "no", the comparison / change unit 3104 updates the model parameters of the vectorization model 3102 so that the similarity notified from the similarity calculation unit 3103 becomes lower.

[0430] <Specific examples of the processing of each unit included in the retrieval unit of the server device>

[0431] Next, a specific example of the processing of each part (here, the first vectorization unit 2203, the separation unit 2204, and the second vectorization unit 2205) included in the updated retrieval unit 2101 of the server device 150 in the feedback phase will be described. Figure 32 FIG. 2 is a diagram showing a specific example of the processing of the first vectorization unit, the separation unit, and the second vectorization unit included in the retrieval unit of the server device.

[0432] As Figure 32 shown, by sending update information from the analysis device 2810 in the feedback phase, the drug development term dictionary storage unit 2403 of the first vectorization unit 2203 is updated to the drug development term dictionary storage unit 3203.

[0433] In addition, as Figure 32 shown, by sending update information from the analysis device 2810 in the feedback phase, the adjusted hyperparameters are applied to the hyperparameters of the vectorization model 2402. As a result, the vectorization model 2402 is updated to the vectorization model 3201. Similarly, by sending update information from the analysis device 2810 in the feedback phase, the adjusted hyperparameters are applied to the hyperparameters of the vectorization model 2412. As a result, the vectorization model 2412 is updated to the vectorization model 3211.

[0434] In addition, as Figure 32 shown, by sending update information from the analysis device 2810 in the feedback phase, a new learned vectorization model is applied to the vectorization model 3201. As a result, the vectorization model 3201 is updated to the vectorization model 3202 (learned vectorization model). Similarly, by sending update information from the analysis device 2810 in the feedback phase, a new learned vectorization model is applied to the vectorization model 3211. As a result, the vectorization model 3211 is updated to the vectorization model 3212 (learned vectorization model).

[0435] <Processing of Business Support System (Feedback Phase)>

[0436] Next, the process of the processing in the feedback phase of the business support system 2800 will be described. Figure 33 FIG. is a timing chart showing the process of the processing in the feedback phase of the business support system.

[0437] In step S3301, the staff terminals 130, 140, etc. receive the evaluation results of the retrieval results input by the staff 131, 141, etc., and send the feedback information to the analysis device 2810.

[0438] In step S3302, whenever feedback information is received, the analysis device 2810 analyzes the drug development term dictionary and sends update information including addition of new terms and deletion of unnecessary terms to the server device 150.

[0439] In step S3303, the server device 150 updates the drug development term dictionary based on the update information from the analysis device 2810.

[0440] In step S3304, the analysis device 2810 adjusts the hyperparameters of the vectorization models 2402 and 2412 based on the accumulated feedback information, and sends update information including the adjusted hyperparameters to the server device 150.

[0441] In step S3305, the server device 150 updates the hyperparameters of the vectorization model 2402 of the first vectorization unit 2203 and the vectorization model 2412 of the second vectorization unit 2205 based on the update information from the analysis device 2810.

[0442] In step S3306, the analysis device 2810 generates a learning dataset based on the accumulated feedback information.

[0443] In step S3307, the analysis device 2810 performs a learning process on the vectorization model by using the generated learning dataset, and generates a new learned vectorization model for generating vector data according to the query content. In addition, the analysis device 2810 sends update information including the generated learned vectorization model to the server device 150.

[0444] In step S3308, the server device 150 updates the vectorization model 2402 of the first vectorization unit 2203 and the vectorization model 2412 of the second vectorization unit 2205 to the new learned vectorization model based on the update information from the analysis device 2810.

[0445] <Summary>

[0446] As can be seen from the above description, the business support system 2800 according to the fifth embodiment:

[0447] · Updates the drug development term dictionary based on the evaluation result of the query content included in the retrieved input-completed partitioned file data.

[0448] · Updates the hyperparameters used by the vectorization model of the first vectorization unit and the vectorization model of the second vectorization unit when generating vector data based on the evaluation result of the query content included in the retrieved input-completed partitioned file data.

[0449] ·Generate a learning dataset that uses the query content included in the retrieval request and the query content included in the input-completed division file data of the retrieval result as input data, and the evaluation result as correct answer data.

[0450] ·Generate a new learned vectorization model by performing learning processing using the learning dataset, and use the generated learned vectorization model to update the vectorization models of the first vectorization unit and the second vectorization unit.

[0451] In this way, in the business support system 2800 according to the fifth embodiment, it is configured to update the retrieval unit 2101 of the server device 150 by feeding back the evaluation result for the retrieval result in the feedback phase. Thus, according to the fifth embodiment, the retrieval accuracy when the staff performs retrieval can be improved.

[0452] [Sixth Embodiment]

[0453] In each of the above embodiments, Figure 1 , 2 , 15, 16, 21, 28, etc. are used to illustrate the system structures of each stage. However, the system structures of each stage are not limited to the structures shown in the above embodiments. For example, in the accumulation stage of the second embodiment above, in Figure 16 , the data management device 120 and the prediction device 1610 are configured as separate entities, but the data management device 120 and the prediction device 1610 can also be configured as one body.

[0454] In addition, in the second and fifth embodiments above, the hardware structures of the learning device 1510, the prediction device 1610, and the analysis device 2810 are not mentioned. However, these hardware structures can also be the same as, for example, Figure 3 the data management device 120 shown. Therefore, for example, in the prediction device 1610, the storage units for storing the morpheme analysis unit 1711, the vectorization model 1712, and the learned keyword model 1802 can also be implemented in Figure 3 the auxiliary storage device 303.

[0455] In addition, in the fifth embodiment above, the use of feedback information for updating the retrieval unit 2101 is described. However, the feedback information can also be used for updating the prediction unit 1611 described in the second embodiment, for example.

[0456] [Seventh Embodiment]

[0457] In the above-described embodiments, the response content for generating the input-completed partitioned file data by the staff members 131 and 141 has been described. However, the response content for the input-completed partitioned file data can also be generated by a language model such as a large language model (LLM: Large Language Model), and the staff members 131 and 141 only need to input the inquiry content, confirm the response content generated by the language model, and make corrections as needed, which can significantly reduce the workload for generating the input-completed partitioned file data. Hereinafter, the seventh embodiment will be described centering on the differences from the above-described embodiments.

[0458] <System Structure of Business Support System (Part 1 of LM Learning Phase)>

[0459] First, the system structure of the business support system according to the seventh embodiment will be described. In the seventh embodiment, the business support system executes the processing in the LM learning phase and the processing in the generation phase. Therefore, the system structure of the business support system in the LM learning phase will be described here.

[0460] Figure 34 Fig. 7 shows an example of the system structure of the business support system and shows the system structure in the LM learning phase. The LM learning phase refers to the phase in which learning processing for a language model is executed. The difference from the business support system 2100 described in the above-described third embodiment is that, in the case of the business support system 3400, Figure 21 as described, Figure 34 in the case of the business support system 3400,

[0461] · The function of the data management device 3410 is different from the function of the data management device 120,

[0462] · The server device 150 does not have the retrieval unit 2101,

[0463] · The external server device 3420 is connected to the external network 160.

[0464] Specifically, the data management device 3410 has an LM learning data generation unit 3411 (an example of a first generation unit). The LM learning data generation unit 3411 reads out the input-completed partitioned file data stored in the database 151 of the server device 150, generates an LM learning data set with the inquiry content as the input data and the response content as the correct answer data, and stores it in the LM learning data storage unit 3412.

[0465] In addition, the LM learning data generation unit 3411 sends the stored LM learning data set to the external server device 3420, thereby performing learning processing on the language model possessed by the external server device 3420.

[0466] The external server device 3420 functions as an LM service providing unit 3421. The LM service providing unit 3421 provides services corresponding to various language processing tasks by executing the language model. In the seventh embodiment, for the language model possessed by the external server device 3420, learning processing is performed using the LM learning data set sent from the LM learning data generation unit 3411. As a result, the language model is fine-tuned to generate response content.

[0467] In addition, in Figure 34 the business support system 3400 shown, the external server device 3420 is configured on the external network 160, but the external server device 3420 may also be configured on the internal network 170, for example.

[0468] <Explanation of the LM learning data set>

[0469] Next, an explanation will be given of the LM learning data set used when fine-tuning the language model possessed by the external server device 3420 to generate response content. Figure 35 FIG. 1 shows an example of the LM learning data set.

[0470] As Figure 35 shown, in the LM learning data set 3500, items of information such as "number" and "input-completed divided file data" are included. In "number", the numbers of the registration data 1000 stored in the database 151 of the storage server device 150 are stored. In "input-completed divided file data", items of information such as "query content" and "response content" are also included.

[0471] In "query content", the query content included in the input-completed divided file data corresponding to the "number" in the input-completed divided file data included in the registration data 1000 is stored. In "response content", the response content corresponding to the query content included in the input-completed divided file data corresponding to the "number" in the input-completed divided file data included in the registration data 1000 is stored.

[0472] <Processing of the business support system (LM learning stage part 1)>

[0473] Next, an explanation will be given of the processing flow in the LM learning stage of the business support system 3400. Figure 36It is the first timing chart showing the process of the LM learning stage of the business support system.

[0474] In step S3601, the server device 150 sends the input-completed partitioned file data including the query content and the response content from the registration data 1000 stored in the database 151 to the data management device 3410. In addition, the sending of the input-completed partitioned file data from the server device 150 to the data management device 3410 can be executed at each specific cycle or whenever the input-completed partitioned file data is newly stored in the registration data 1000.

[0475] In step S3602, the data management device 3410 generates the LM learning dataset 3500 based on the input-completed partitioned file data sent from the server device 150 and stores it in the LM learning data storage unit 3412.

[0476] In step S3603, the data management device 3410 sends the LM learning dataset 3500 stored in the LM learning data storage unit 3412 to the external server device 3420. Thereby, the data management device 3410 instructs to perform the learning process on the language model possessed by the external server device 3420. In addition, the sending of the LM learning dataset 3500 from the data management device 3410 to the external server device 3420 can also be executed at each specific cycle. Or, the sending of the LM learning dataset 3500 from the data management device 3410 to the external server device 3420 can also be executed whenever the input-completed partitioned file data is sent from the server device 150 and the LM learning dataset 3500 is updated.

[0477] In step S3604, the external server device 3420 uses the LM learning dataset 3500 sent from the data management device 3410 to perform the learning process on the language model, thereby fine-tuning the language model to generate the response content.

[0478] <System structure of the business support system (LM learning stage part 2)>

[0479] Next, as the business support system according to the seventh embodiment, the system structure of a business support system different from Figure 34 the business support system shown will be described. Figure 37A It is the eighth diagram showing an example of the system structure of the business support system, showing another system structure in the LM learning stage. The difference from the business support system 3400 described using Figure 34 is that in the case of the business support system 3700 of Figure 37A

[0480] · The data management device 3410 has a first LM learning data generation unit 3411 (another example of the first generation unit) instead of the LM learning data generation unit 3411.

[0481] · The external server device 3420 has a second LM learning data generation unit 3710 (an example of the second generation unit).

[0482] The first LM learning data generation unit 3411 reads the input-completed partitioned file data stored in the database 151 of the server device 150. In addition, the first LM learning data generation unit 3411 generates a first LM learning data set in which the query content of the read input-completed partitioned file data is used as input data and the response content is used as correct answer data, and stores it in the LM learning data storage unit 3412. In addition, the first LM learning data set is an example of the first learning data set.

[0483] In addition, the first LM learning data generation unit 3411 sends the stored first LM learning data set to the external server device 3420, so that the external server device 3420 performs learning processing on the language model included in the LM service providing unit 3421.

[0484] The second LM learning data generation unit 3710 included in the external server device 3420 further includes a learning object information collection unit 3711 and a self-supervised training data generation unit 3712.

[0485] The learning object information collection unit 3711 accesses the Web server device 3720 via the external network 160 to collect learning object information. The learning object information refers to any information that helps the LM service providing unit 3421 generate a response corresponding to the latest information. The learning object information collection unit 3711 is configured to automatically collect learning object information from a preset access destination at a preset timing. Alternatively, the learning object information collection unit 3711 may be configured to collect learning object information from the indicated access destination at the timing indicated by the applicant 122. In addition, the learning object information collection unit 3711 may be configured to accept the input of the applicant 122 when logging in is required when accessing the indicated access destination.

[0486] The self-supervised learning data generation unit 3712 generates a second LM learning dataset based on the learning object information collected by the learning object information collection unit 3711. For example, the self-supervised learning data generation unit 3712 generates a second LM learning dataset that includes a masked file obtained by masking a part of the collected learning object information and a file (correct answer file) composed of the collected learning object information (unmasked learning object information). In addition, the second LM learning dataset is an example of the second learning dataset.

[0487] The self-supervised learning data generation unit 3712 notifies the generated second LM learning dataset to the LM service providing unit 3421 to cause the learning process of the language model possessed by the LM service providing unit 3421 to be executed.

[0488] In addition, in Figure 37A the business support system 3700 shown, it is shown that the external server device 3420 is configured on the external network 160, but the external server device 3420 can also be configured on the internal network 170, for example.

[0489] Furthermore, in Figure 37A the business support system 3700 shown, it is shown that the second LM learning data generation unit 3710 is implemented in the external server device 3420. However, the second LM learning data generation unit 3710 can also be implemented in any device configured on the internal network 170 (for example, in the data management device 3410), for example. In addition, the second LM learning data generation unit 3710 can also be implemented in both the external server device 3420 and any device configured on the internal network 170 (for example, in the data management device 3410). That is, the second LM learning data generation unit 3710 can also be implemented in multiple devices.

[0490] Furthermore, in Figure 37A the business support system 3700 shown, the description is made on the premise that the LM service providing unit 3421 has one language model, but the LM service providing unit 3421 can also have three language models.

[0491] Specifically, the LM service providing unit 3421 can also have:

[0492] · A language model fine-tuned using the first LM learning dataset,

[0493] · A language model fine-tuned using the second LM learning dataset,

[0494] · A language model fine-tuned using both the first LM learning dataset and the second LM learning dataset.

[0495] <System Structure of Business Support System (LM Learning Phase 3)>

[0496] Next, as the business support system according to the 7th embodiment, a modified example of the Figure 37A business support system shown will be described. Figure 37B Fig. 9 shows an example of the system structure of the business support system and shows another system structure in the LM learning phase. The difference from the business support system 3700 described using Figure 37A is that in the case of the business support system 3700 of Figure 37B

[0497] · The data management device 3410 has a first generation control unit 3721,

[0498] · The external server device 3420 has a second generation control unit 3722.

[0499] The first generation control unit 3721 appropriately selects and discards the first LM learning dataset stored in the LM learning data storage unit 3412 by referring to the learning object information collected by the learning object information collection unit 3711 of the external server device 3420. This is because the first LM learning dataset also contains data that is not suitable for fine-tuning the language model according to the latest information for generating response content.

[0500] The second generation control unit 3722 appropriately changes the collection source when the learning object information collection unit 3711 collects learning object information by referring to the first LM learning dataset generated by the first LM learning data generation unit 3411 of the data management device 3410. This is because the learning object information to be collected by the learning object information collection unit 3711 is important information for replacing the first LM learning dataset with the latest information.

[0501] <Explanation of LM Learning Datasets>

[0502] Next, the first LM learning dataset and the second LM learning dataset used when fine-tuning the language model of the external server device 3420 for generating response content will be described. Figure 38 Fig. 2 shows an example of the LM learning dataset.

[0503] In Figure 38 , the first LM learning dataset 3500 is the same as the LM learning dataset 3500 shown in Figure 35 , so the description is omitted here.

[0504] In Figure 38Among them, the second LM learning dataset 3820 is a learning dataset generated by the self-supervised learning data generation unit 3712 based on the learning object information 3810 collected by the learning object information collection unit 3711.

[0505] In addition, in the learning object information 3810, items of information such as "number", "information name", "file name", "collection source", and "collection date and time" are included.

[0506] In "number", the number assigned to the file collected as learning object information is stored. In "information name", the name of the file collected as learning object information is stored. In Figure 38 In the example of, the files collected as learning object information include, for example:

[0507] · Files showing guidelines for drugs,

[0508] · Files related to drug development,

[0509] · Files related to the achievements of industry activities,

[0510] · Drug-related magazines,

[0511] · Medical / chemical department magazines,

[0512] · Journals in which medical / chemical papers are published.

[0513] In "file name", the file name of the file stored in "information name" is stored. In "collection source", the access destination when collecting the file stored in "information name" is stored. In "collection date and time", the collection date and time when collecting the file stored in "information name" is stored.

[0514] In the second LM learning dataset 3820, items of information such as "number" and "self-supervised learning data" are included. In "number", the number of the learning object information used when the second LM learning dataset is generated by the self-supervised learning data generation unit 3712 is stored.

[0515] In "self-supervised learning data", items of information such as "mask file" and "correct answer file" are also included. In "mask file", a masked file with a part masked, which is included in the learning object information determined by the corresponding "number" (for example, "N1"), is stored. In "correct answer file", the file before the corresponding mask file is masked is stored. As Figure 38 shown, it is also possible to generate combinations of multiple mask files and multiple correct answer files based on one file of learning object information.

[0516] If the learning object information 3810 is collected, the self-supervised learning data generation unit 3712 automatically (i.e., without the participation of applicant 122) generates the second LM learning dataset 3820.

[0517] <Process of Business Support System (LM Learning Phase Part 2)>

[0518] Next, the process flow of the LM learning phase in the business support system 3700 will be described. Figure 39A This is the second timing diagram showing the process flow in the LM learning phase of the business support system, showing Figure 37A the process flow in the LM learning phase of the business support system 3700 shown.

[0519] In Figure 39A the processing of steps S3601 to S3604 is substantially the same as the processing of steps S3601 to S3604 in Figure 36 so the description is omitted here.

[0520] In step S3901, the external server device 3420 accesses the Web server device 3720 via the external network 160 to collect learning object information.

[0521] In step S3902, the external server device 3420 generates the second LM learning dataset based on the collected learning object information.

[0522] In step S3903, the external server device 3420 uses the generated second LM learning dataset to perform a learning process on the language model and fine-tune the language model to generate response content.

[0523] In addition, Figure 39A in the sending of the first LM learning dataset from the data management device 3410 to the external server device 3420 can also be executed at each specific period. Or, the sending of the first LM learning dataset from the data management device 3410 to the external server device 3420 can also be executed whenever the input completion partitioned file data is sent from the server device 150 and the first LM learning dataset is updated.

[0524] Furthermore, in Figure 39A the external server device 3420 can also execute the learning process (step S3903) using the second LM learning dataset at each specific period. In addition, the external server device 3420 can execute the learning process (step S3903) using the second LM learning dataset either at the same period as the learning process (step S3604) using the first LM learning dataset or at a different period.

[0525] In addition, the external server device 3420 repeatedly executes the learning process using the first LM learning dataset (step S3604) and the learning process using the second LM learning dataset (step S3903). As a result, the LM service providing unit 3421 can output a response content that incorporates the latest information into the past input-completed partitioned file data. Alternatively, when the LM service providing unit 3421 outputs a response content incorporating the latest information, it can output a response content incorporating the expression method of the recent input-completed partitioned file data.

[0526] <Process of Business Support System (LM Learning Phase 3)>

[0527] Next, the process flow in the LM learning phase of the business support system 3700 will be described. Figure 39B This is the third timing diagram showing the process flow in the LM learning phase of the business support system, showing Figure 37B the process flow in the LM learning phase of the business support system 3700 shown.

[0528] In Figure 39B the processing of steps S3601 to S3604 is substantially the same as the processing of steps S3601 to S3604 in Figure 36 so the description thereof is omitted here.

[0529] In step S3911, the external server device 3420 changes the collection source when the learning object information collection unit 3711 collects learning object information by referring to the first LM learning dataset generated by the data management device 3410.

[0530] The processing of steps S3901 to S3902 and S3903 is the same as the processing of steps S3901 to S3902 and S3903 in Figure 39A so the description thereof is omitted here.

[0531] In step S3912, the external server device 3420 sends the second LM learning dataset generated in step S3902 to the data management device 3410.

[0532] In step S3913, the data management device 3410 selects and rejects the first LM learning dataset by referring to the learning object information included in the second LM learning dataset.

[0533] <System Structure of Business Support System (Generation Phase)>

[0534] Next, the system structure in the generation phase of the business support system according to the seventh embodiment will be described. Figure 40FIG. 10 shows an example of the system structure of a business support system, showing the system structure in the generation stage. The generation stage refers to the stage of generating input-completed file data using a language model that has been fine-tuned for generating response content. Different from the business support system 3400 in the LM learning stage described using Figure 34 in the case of the business support system 4000 of Figure 40

[0535] · The function of the data management device 4010 is different from that of the data management device 3410,

[0536] · The data management device 4010 has an LM utilization unit 4011,

[0537] · The data management device 4010 does not have the LM learning data storage unit 3412.

[0538] As Figure 40 shown, if the authority inquiry file data ((1)) is sent from the authority terminal 110, the management unit 121 of the data management device 4010 generates a plurality of divided file data by dividing the received authority inquiry file data according to each inquiry item. In addition, the management unit 121 of the data management device 4010 sends the generated plurality of divided file data to the corresponding staff terminals 130, 140, etc. ((2)).

[0539] If the divided file data is sent from the data management device 4010, the staff member 131 extracts the inquiry content from the divided file data and generates a prompt including the extracted inquiry content. In addition, the staff member 131 sends the generated prompt to the LM utilization unit 4011 of the data management device 4010 via the staff terminal 130, thereby instructing the generation of a response content for the extracted inquiry content ((3)).

[0540] Based on the prompt received from the staff terminal 130, the LM utilization unit 4011 sends the inquiry content to the external server device 3420 and instructs the generation of a response content for the sent inquiry content.

[0541] The LM service providing unit 3421 of the external server device 3420 has a language model that has been fine-tuned for generating response content. If the inquiry content and the generation instruction for the response content for the inquiry content are received from the data management device 4010, the LM service providing unit 3421 of the external server device 3420 generates a response content according to the generation instruction.

[0542] ​The response content generated by the external server device 3420 is obtained by the LM utilization unit 4011 of the data management device 4010 ((4)), and is sent to the staff terminal 130 of the instruction source. Thus, while referring to the response content generated by the language model that has been fine-tuned for generating the response content, the staff member 131 inputs the response content to the inquiry content, and generates the input-completed divided file data.

[0543] The input-completed divided file data generated by the staff member 131 is sent to the management unit 121 of the data management device 4010 via the staff terminal 130 ((5)).

[0544] Based on the input-completed divided file data collected from the staff terminal 130, the staff terminal 140, etc., the management unit 121 of the data management device 4010 generates the input-completed file data, and sends it to the authority terminal 110 via the external network 160 ((6)).

[0545] In this way, according to the business support system 4000, when generating the input-completed divided file data, the staff members 131, 141, etc. can generate while referring to the response content obtained by using the language model that has been fine-tuned for generating the response content. That is, according to the business support system 4000, it is possible to support the operations of the staff members 131, 141, etc. to generate the input-completed divided file data.

[0546] <Business Support System Processing (Generation Phase)>

[0547] Next, the processing flow in the generation phase of the business support system 4000 will be described. Figure 41 It is the first timing diagram showing the processing flow in the generation phase of the business support system.

[0548] Among them, the processing of steps S1301~S1303, S1305~S1309 is the same as that of Figure 13 steps S1301~S1303, S1305~S1309, so the description is omitted here.

[0549] In step S4101, the staff terminals 130, 140, etc. respectively display the received divided file data.

[0550] In step S4102, when the staff terminals 130, 140, etc. receive the prompt that contains the generation instruction of the inquiry content and the response content to the inquiry content input by the staff members 131, 141, etc. in response to the display of the divided file data, they accept the prompt. In addition, the staff terminals 130, 140, etc. send the accepted input prompt to the data management device 4010.

[0551] In step S4103, the data management device 4010 sends an inquiry content and an instruction to generate a response content for the inquiry content to the external server device 3420.

[0552] In step S4104, if the external server device 3420 receives an inquiry content and an instruction to generate a response content for the inquiry content from the data management device 4010, the external server device 3420 causes a language model that has been fine-tuned for generating the response content to execute, and generates the response content. In addition, the external server device 3420 sends the response content generated by the data management device 4010.

[0553] In step S4105, the data management device 4010 receives the response content sent from the external server device 3420 and sends it to the staff terminals 130, 140, etc.

[0554] In step S4106, the staff terminals 130, 140, etc. display the response content sent from the data management device 4010. Thus, the staff members 131, 141, etc. can generate the input-completed divided file data while referring to the response content displayed on the staff terminals 130, 140, etc.

[0555] <Summary>

[0556] From the above description, it can be seen that the business support system 3400 according to the seventh embodiment:

[0557] · Generates a learning dataset obtained by associating the inquiry content and the response content included in multiple input-completed divided file data, and sends the generated learning dataset, thereby causing the learning process of the language model to be executed.

[0558] In addition, the business support system 4000 according to the seventh embodiment:

[0559] · For the authority inquiry file data to be processed, in the case where multiple divided file data are generated, the inquiry content of each of the multiple inquiry items is sent to the language model that has undergone the learning process, and the response content for the inquiry content of each of the multiple inquiry items is obtained.

[0560] · Obtains multiple input-completed divided file data generated based on the obtained response content.

[0561] Thus, according to the business support system according to the seventh embodiment, it is possible to support the work of the staff who generate the input-completed divided file data.

[0562] [Eighth Embodiment]

[0563] In the above-described seventh embodiment, the following structure is adopted: when fine-tuning the language model, the LM learning dataset 3500 including the inquiry content and the response content is used. However, the LM learning dataset used when fine-tuning the language model is not limited to the LM learning dataset 3500. For example, an LM learning dataset including the authority submission materials (an example of submission file data) submitted to the authority before sending the authority inquiry file data from the authority terminal 110 and the inquiry content sent from the authority terminal 110 can also be used. Hereinafter, the eighth embodiment will be described centering on the differences from the above-described seventh embodiment.

[0564] <System Structure of Business Support System (LM Learning Phase)>

[0565] First, the system structure of the business support system according to the eighth embodiment will be described. In the eighth embodiment, the business support system executes the processing in the LM learning phase, the processing in the pharmaceutical affairs application phase, and the processing in the pharmaceutical affairs inquiry phase. Here, the system structure of the business support system in the LM learning phase will be described.

[0566] Figure 42 FIG. 11 shows an example of the system structure of the business support system and shows the system structure in the LM learning phase. Different from the business support system 3400 described above using Figure 34 In the case of the business support system 4200 Figure 42 described,

[0567] · The function of the data management device 4210 is different from the function of the data management device 3410,

[0568] · The server device 4220 is connected to the internal network 170.

[0569] Specifically, the data management device 4210 has an LM learning data generation unit 4211. The LM learning data generation unit 4211 reads out the authority submission materials stored in the authority submission material storage unit 4221 of the server device 4220.

[0570] In addition, in response to the situation where the authority submission materials have been submitted to the authority, the LM learning data generation unit 4211 reads out the inquiry content included in the authority inquiry file data sent from the authority from the database 151 of the server device 150. Furthermore, the LM learning data generation unit 4211 generates an LM learning dataset with the read authority submission materials as input data and the inquiry content as correct answer data, and stores it in the LM learning data storage unit 4212.

[0571] In addition, the LM learning data generation unit 4211 sends the stored LM learning data set to the external server device 3420, thereby performing the learning process on the language model possessed by the external server device 3420.

[0572] In the eighth embodiment, for the language model possessed by the external server device 3420, a learning process is performed using the LM learning data set sent from the LM learning data generation unit 4211. As a result, the language model is fine-tuned for generating query content.

[0573] In addition, in the eighth embodiment, in the "materials submitted to the authority", for example, in addition to "main materials" such as drug manufacturing and sales approval application documents (application documents), it also includes "annexed materials" submitted to the authority in connection with the submission of drug manufacturing and sales approval application documents. The so-called "annexed materials" refer to, for example, consultation records, test result data, etc. However, even materials equivalent to consultation records, test result data, etc., internal materials not submitted to the authority as "annexed materials" are not included in the LM learning data set.

[0574] In addition, when the materials submitted to the authority include data other than character information such as images, graphs, and tables, the LM learning data generation unit 4211 can also generate the LM learning data set by attaching annotations to the data other than the character information. Or, it is assumed that the materials submitted to the authority include data other than character information such as images, graphs, and tables, and the LM service providing unit 3421 can also have a multimodal model instead of the language model.

[0575] <Explanation of the LM learning data set>

[0576] Next, an explanation will be given of the LM learning data set used when fine-tuning the language model possessed by the external server device 3420 for generating query content. Figure 43 It is FIG. 3 showing an example of the LM learning data set.

[0577] As Figure 43As shown, in the LM learning dataset 4300, items including "number", "materials submitted to the authority", and "inquiry content" are included as information. In "number", the numbers assigned to each material submitted to the authority are stored. In "materials submitted to the authority", the main materials submitted to the authority and the supplementary materials submitted to the authority in association with the submission of the main materials are stored. In "inquiry content", in response to the situation where the main materials and the supplementary materials are submitted to the authority, each inquiry content included in the authority inquiry document data sent from the authority is stored.

[0578] <Process of the business support system (LM learning stage)>

[0579] Next, the process of the LM learning stage in the business support system 4000 will be described. Figure 44 It is the 4th timing diagram showing the process of the LM learning stage in the business support system.

[0580] In step S4401, the server device 4220 sends the materials submitted to the authority stored in the authority submission material storage unit 4221 to the data management device 4210. In addition, the sending of the materials submitted to the authority from the server device 4220 to the data management device 4210 can be executed at a specific cycle or whenever new materials submitted to the authority are stored in the authority submission material storage unit 4221.

[0581] In step S4402, the server device 150 sends the authority inquiry document data from the registered data 1000 stored in the database 151 to the data management device 4210. In addition, the sending of the authority inquiry document data from the server device 150 to the data management device 4210 can be executed at a specific cycle or whenever the input-completed divided file data is newly stored in the registered data 1000.

[0582] In step S4403, the data management device 4210 generates the LM learning dataset 4300 based on the materials submitted to the authority sent from the server device 4220 and the authority inquiry document data sent from the server device 150. In addition, the data management device 3410 stores the generated LM learning dataset 4300 in the LM learning data storage unit 4212.

[0583] In step S4404, the data management device 4210 sends the LM learning dataset 4300 stored in the LM learning data storage unit 4212 to the external server device 3420. Thereby, the data management device 4210 causes the learning process of the language model possessed by the external server device 3420 to be executed. In addition, the sending of the LM learning dataset 4300 from the data management device 4210 to the external server device 3420 can be executed at a specific cycle or whenever the LM learning dataset 4300 is updated.

[0584] In step S4405, the external server device 3420 uses the LM learning dataset 4300 sent from the data management device 4210 to perform a learning process on the language model, thereby fine-tuning the language model for use in generating inquiry content.

[0585] <System Structure of Business Support System (Pharmaceutical Affairs Application Stage)>

[0586] Next, the system structure in the pharmaceutical affairs application stage of the business support system according to the eighth embodiment will be described. Figure 45 FIG. 12 is an example showing the system structure of the business support system, showing the system structure in the pharmaceutical affairs application stage. The pharmaceutical affairs application stage refers to the stage of generating materials to be submitted to the authorities when applying for approval of the manufacture and sale of drugs, etc., using a language model fine-tuned for generating inquiry content. Different from the business support system 4200 in the LM learning stage described above, in the case of the business support system 4500, Figure 42 as described above, Figure 45 in the case of the business support system 4500,

[0587] · The function of the data management device 4510 is different from that of the data management device 4210,

[0588] · The data management device 4510 has a material generation unit 4511 and an LM utilization unit 4512,

[0589] · The data management device 4510 does not have the LM learning data storage unit 4212.

[0590] As Figure 45 shown, first, the applicant 122 drafts a material draft ((1)) to be submitted to the authorities when applying for approval of the manufacture and sale of drugs, etc., via the material generation unit 4511. Next, the applicant 122 generates a prompt including the generated material draft to be submitted to the authorities, and inputs the generated prompt into the LM utilization unit 4512 to instruct the generation of inquiry content for the material draft to be submitted to the authorities.

[0591] The LM utilization unit 4512 sends the draft official submission materials to the external server device 3420 based on the prompt input by the applicant 122, and instructs to generate the inquiry content for the sent draft official submission materials.

[0592] The LM service providing unit 3421 of the external server device 3420 has a language model that has been fine-tuned for generating inquiry content. Then, if it receives the draft official submission materials and the generation instruction for the inquiry content for the draft official submission materials from the data management device 4510, the LM service providing unit 3421 of the external server device 3420 generates the inquiry content according to the generation instruction.

[0593] The inquiry content generated by the external server device 3420 is obtained by the LM utilization unit 4512 of the data management device 4510 ((2)), and is displayed to the applicant 122. Thus, the applicant 122 corrects the draft official submission materials while referring to the inquiry content for the draft official submission materials. Specifically, the applicant 122 makes corrections to the main materials, deletes or adds supplementary materials ((3)) so that the official inquiry document data of the displayed inquiry content is not sent from the authority.

[0594] Next, the applicant 122 generates a prompt including the corrected draft official submission materials, and inputs the generated prompt into the LM utilization unit 4512, thereby instructing to generate the inquiry content for the corrected draft official submission materials.

[0595] The LM utilization unit 4512 sends the corrected draft official submission materials to the external server device 3420 based on the prompt input by the applicant 122, and instructs to generate the inquiry content for the sent corrected draft official submission materials.

[0596] If it receives the corrected draft official submission materials and the generation instruction for the inquiry content for the corrected draft official submission materials from the data management device 4510, the LM service providing unit 3421 of the external server device 3420 generates the inquiry content according to the generation instruction.

[0597] The inquiry content generated by the external server device 3420 is obtained by the LM utilization unit 4512 of the data management device 4510 ((4)), and is displayed to the applicant 122. Thus, the applicant 122 makes a second correction to the corrected draft official submission materials while referring to the inquiry content for the corrected draft official submission materials. Specifically, the applicant 122 makes corrections to the main materials, deletes or adds supplementary materials ((5)) so that the official inquiry document data of the displayed inquiry content is not sent from the authority.

[0598] Next, applicant 122 generates a prompt including the draft of the authorities submission materials after the second amendment, and inputs the generated prompt into the LM utilization unit 4512, thereby instructing the generation of inquiry content for the draft of the authorities submission materials after the second amendment.

[0599] Based on the prompt input by applicant 122, the LM utilization unit 4512 sends the draft of the authorities submission materials after the second amendment to the external server device 3420, instructing the generation of inquiry content for the sent draft of the authorities submission materials after the second amendment.

[0600] If receiving the draft of the authorities submission materials after the second amendment and the generation instruction for the inquiry content for the draft of the authorities submission materials after the second amendment from the data management device 4510, the LM service providing unit 3421 of the external server device 3420 generates inquiry content according to the generation instruction.

[0601] The inquiry content generated by the external server device 3420 is obtained by the LM utilization unit 4512 of the data management device 4510 ((6)) and displayed to applicant 122. Thus, applicant 122 re-amends the draft of the authorities submission materials after the second amendment while referring to the inquiry content for the draft of the authorities submission materials after the second amendment. Specifically, applicant 122 amends the main materials, deletes or adds supplementary materials ((7)) so that the authorities inquiry document data of the displayed inquiry content is not sent from the authorities, and completes the authorities submission materials.

[0602] The completed authorities submission materials are sent to the authorities terminal 110 via the material generation unit 4511, and the manufacturing and sales approval application for drugs, etc. is completed ((8)).

[0603] In this way, according to the business support system 4500, when generating the authorities submission document, applicant 122 can generate the authorities submission materials while referring to the inquiry content obtained by using the language model that has been fine-tuned for generating the inquiry content. That is, according to the business support system 4500, the operation of applicant 122 for generating the authorities submission document can be supported.

[0604] <Process of the Business Support System (Pharmaceutical Affairs Application Stage)>

[0605] Next, the process of the processing in the pharmaceutical affairs application stage of the business support system 4500 will be described. Figure 46 It is a timing chart showing the process of the processing in the pharmaceutical affairs application stage of the business support system.

[0606] In step S4601, when a prompt including a draft of the authority submission material generated by applicant 122 is input, the data management device 4510 accepts the prompt. In addition, the data management device 4510 sends an instruction to generate the draft of the authority submission material and the query content for the draft of the authority submission material to the external server device 3420. Additionally, the draft of the authority submission material generated by applicant 122 may also be generated via the staff terminal 130 and the staff terminal 140 based on the draft of the partial authority submission material generated by the staff member 131 or the staff member 141.

[0607] In step S4602, if the external server device 3420 receives an instruction to generate the draft of the authority submission material and the query content for the draft of the authority submission material from the data management device 4510, the external server device 3420 causes the language model that has been fine-tuned for generating the query content to execute. Thereby, the external server device 3420 generates the query content. In addition, the external server device 3420 sends the generated query content to the data management device 4510.

[0608] In step S4603, when a prompt including a draft of the authority submission material revised by applicant 122 is input, the data management device 4510 accepts the prompt. In addition, the data management device 4510 sends the revised draft of the authority submission material and an instruction to generate the query content for the revised draft of the authority submission material to the external server device 3420. Additionally, the draft of the authority submission material revised by applicant 122 may also be revised via the staff terminal 130 and the staff terminal 140 based on the revised draft of the partial authority submission material generated by the staff member 131 or the staff member 141.

[0609] In step S4604, if the external server device 3420 receives the revised draft of the authority submission material and an instruction to generate the query content for the revised draft of the authority submission material from the data management device 4510, the external server device 3420 generates the query content. In addition, the external server device 3420 sends the generated query content to the data management device 4510.

[0610] In step S4605, when a prompt including a draft of the authority submission material revised again by applicant 122 is input, the data management device 4510 accepts the prompt. In addition, the data management device 4510 sends the draft of the authority submission material revised again and an instruction to generate the query content for the draft of the authority submission material revised again to the external server device 3420. Additionally, the draft of the authority submission material revised again by applicant 122 may also be revised again via the staff terminal 130 and the staff terminal 140 based on the draft of the partial authority submission material revised again by the staff member 131 or the staff member 141.

[0611] In step S4606, if the external server device 3420 receives the redrafted authority submission materials after re - amendment and an instruction to generate inquiry content for the redrafted authority submission materials from the data management device 4510, the external server device 3420 generates the inquiry content. In addition, the external server device 3420 sends the generated inquiry content to the data management device 4510.

[0612] In step S4607, the data management device 4510 accepts the authority submission materials that have been completed through re - amendment by applicant 122. In addition, the data management device 4510 sends the completed authority submission materials to the authority terminal 110, thereby submitting the authority submission materials to the authority. Additionally, the draft authority submission materials redrafted by applicant 122 can also be redrafted based on the draft partial authority submission materials redrafted by staff member 131 or staff member 141 via the staff terminal 130 and staff terminal 140.

[0613] <System Structure of Business Support System (Pharmaceutical Affairs Inquiry Phase)>

[0614] Next, the system structure in the pharmaceutical affairs inquiry phase of the business support system according to the eighth embodiment will be described. Figure 47 FIG. 13 is an example showing the system structure of the business system, which shows the system structure in the pharmaceutical affairs inquiry phase. The pharmaceutical affairs inquiry phase refers to the phase of generating the input - completed file data using a language model that has been fine - tuned for generating inquiry content. It is basically the same as the generation phase described in the seventh embodiment above, but is different in that the language model used for generating the input - completed file data has been fine - tuned for generating inquiry content. Therefore, compared with the business support system 4000 in the generation phase described using Figure 40 :

[0615] · The functions of the data management device 4710 are different from those of the data management device 4010.

[0616] · It is different in that the server device 4220 is connected.

[0617] As Figure 47 shown, if the authority inquiry file data ((1)) is sent from the authority terminal 110, the management unit 121 of the data management device 4710 generates a plurality of divided file data by dividing the received authority inquiry file data for each inquiry item. In addition, the management unit 121 of the data management device 4710 sends the generated plurality of divided file data to the corresponding staff terminals 130, 140, etc. ((2)).

[0618] When dividing file data is sent from the data management device 4710, the staff member 131 extracts the inquiry content from the dividing file data and generates a prompt including the extracted inquiry content. In addition, the staff member 131 sends the generated prompt to the LM utilization unit 4711 of the data management device 4710 via the staff terminal 130, thereby instructing the generation of a response content for the extracted inquiry content ((3)).

[0619] Based on the prompt received from the staff terminal 130, the LM utilization unit 4711 sends the inquiry content to the external server device 3420 and instructs the generation of a response content for the sent inquiry content.

[0620] When receiving the inquiry content and an instruction to generate a response content for the inquiry content from the data management device 4710, the LM service providing unit 3421 of the external server device 3420 generates a response content according to the generation instruction.

[0621] The response content generated by the external server device 3420 is obtained by the LM utilization unit 4711 of the data management device 4710 ((4)) and sent to the staff terminal 130 of the instruction source. Thus, the staff member 131 inputs a response content for the inquiry content while referring to the response content generated by the language model, and generates a divided file data with input completed.

[0622] Next, the staff member 131 generates a prompt including the generated divided file data with input completed. In addition, the staff member 131 sends the generated prompt to the LM utilization unit 4711 of the data management device 4710 via the staff terminal 130, thereby instructing the generation of an inquiry content for the generated divided file data with input completed ((5)).

[0623] Based on the prompt received from the staff terminal 130, the LM utilization unit 4711 sends the divided file data with input completed to the external server device 3420 and instructs the generation of an inquiry content for the sent divided file data with input completed.

[0624] When receiving the divided file data with input completed and an instruction to generate an inquiry content for the divided file data with input completed from the data management device 4710, the LM service providing unit 3421 of the external server device 3420 generates an inquiry content according to the generation instruction.

[0625] The inquiry content generated by the external server device 3420 is obtained by the LM utilization unit 4711 of the data management device 4710 ((6)) and sent to the staff terminal 130 of the instruction source.

[0626] Next, the staff member 131 generates a prompt including the inquiry content, and sends the generated prompt to the LM utilization unit 4711 of the data management device 4710 via the staff terminal 130, thereby instructing the generation of a response content for the inquiry content ((7)).

[0627] Based on the prompt received from the staff terminal 130, the LM utilization unit 4711 of the data management device 4710 sends the inquiry content to the external server device 3420, instructing the generation of a response content for the sent inquiry content.

[0628] If the data management device 4710 receives the inquiry content and the generation instruction for the response content for the inquiry content, the LM service providing unit 3421 of the external server device 3420 generates a response content according to the generation instruction.

[0629] The response content generated by the external server device 3420 is obtained by the LM utilization unit 4711 of the data management device 4710 ((8)), and sent to the staff terminal 130 of the instruction source. Thus, the staff member 131 inputs the response content for the inquiry content while referring to the response content generated by the language model, and generates the input-completed partition file data.

[0630] Next, the staff member 131 sends the generated input-completed partition file data to the management unit 121 of the data management device 4710 via the staff terminal 130 ((9)).

[0631] Based on the input-completed partition file data collected from the staff terminal 130, the staff terminal 140, etc., the management unit 121 of the data management device 4710 generates the input-completed file data, and sends it to the authority terminal 110 via the external network 160 ((10)).

[0632] In this way, according to the business support system 4700, when generating the input-completed partition file data, the staff members 131, 141, etc. use the language model that has been fine-tuned for generating the inquiry content, and repeatedly obtain the inquiry content and the response content. Thus, the staff members 131, 141, etc. can generate the input-completed partition file data for which it is difficult to receive further inquiries from the authority. That is, according to the business support system 4700, it is possible to support the operation of the staff members 131, 141, etc. in generating the input-completed partition file data.

[0633] <Business Support System Processing (Pharmaceutical Affairs Inquiry Phase)>

[0634] Next, the process of the processing in the pharmaceutical affairs inquiry phase of the business support system 4700 will be described. Figure 48A And Figure 48BThey are the first and second timing diagrams showing the processes in the pharmaceutical inquiry phase of the business support system.

[0635] Among them, the processes of steps S1301 to S1303, S1305 to S1309 are the same as Figure 13 the processes of steps S1301 to S1303, S1305 to S1309 in , so the description is omitted here.

[0636] In step S4801, the staff terminals 130, 140, etc. respectively display the received divided file data.

[0637] In step S4802, in response to the display of the divided file data, when a prompt including an instruction to generate a reply content for the inquiry content is input by the staff 131, 141, etc., the staff terminals 130, 140, etc. accept the prompt. In addition, the staff terminals 130, 140, etc. send the accepted input prompt to the data management device 4710.

[0638] In step S4803, the data management device 4710 sends the inquiry content and an instruction to generate a reply content for the inquiry content to the external server device 3420.

[0639] In step S4804, if the inquiry content and an instruction to generate a reply content for the inquiry content are received from the data management device 4710, the external server device 3420 generates a reply content. In addition, the external server device 3420 sends the generated reply content to the data management device 4710.

[0640] In step S4805, the data management device 4710 receives the reply content sent from the external server device 3420 and sends it to the staff terminals 130, 140, etc.

[0641] In step S4806, the staff terminals 130, 140, etc. display the reply content sent from the data management device 4710. Thus, the staff 131, 141, etc. can generate the input-completed divided file data while referring to the reply content displayed on the staff terminals 130, 140, etc.

[0642] In step S4807, when a prompt including an instruction to generate an inquiry content for the input-completed divided file data generated by the staff 131, 141, etc. is input, the staff terminals 130, 140, etc. accept the prompt. Furthermore, the staff terminals 130, 140, etc. send the accepted input prompt to the data management device 4710.

[0643] In step S4808, the data management device 4710 sends an instruction to generate the input-completed divided file data and the inquiry content for the input-completed divided file data to the external server device 3420.

[0644] In Figure 48B step S4809, if the external server device 3420 receives the input-completed divided file data and the instruction to generate the inquiry content for the input-completed divided file data from the data management device 4710, the external server device 3420 generates the inquiry content. In addition, the external server device 3420 sends the generated inquiry content to the data management device 4710.

[0645] In step S4810, the data management device 4710 receives the inquiry content sent from the external server device 3420 and sends it to the staff terminals 130, 140, etc.

[0646] In step S4811, the staff terminals 130, 140, etc. display the inquiry content sent from the data management device 4710.

[0647] In step S4812, when the staff members 131, 141, etc. input a prompt including the inquiry content and the instruction to generate the reply content for the inquiry content, the staff terminals 130, 140, etc. accept the prompt. In addition, the staff terminals 130, 140, etc. send the accepted prompt with the input to the data management device 4710.

[0648] In step S4813, the data management device 4710 sends the inquiry content and the instruction to generate the reply content for the inquiry content to the external server device 3420.

[0649] In step S4814, if the external server device 3420 receives the inquiry content and the instruction to generate the reply content for the inquiry content from the data management device 4710, the external server device 3420 generates the reply content. In addition, the external server device 3420 sends the generated reply content to the data management device 4710.

[0650] In step S4815, the data management device 4710 receives the reply content sent from the external server device 3420 and sends it to the staff terminals 130, 140, etc.

[0651] In step S4816, the staff terminals 130, 140, etc. display the reply content sent from the data management device 4710. Thus, the staff members 131, 141, etc. can generate the input-completed divided file data while referring to the reply content displayed on the staff terminals 130, 140, etc.

[0652] <Summary>

[0653] As can be seen from the above description, the business support system 4200 according to the eighth embodiment:

[0654] · Generate a learning dataset obtained by associating the query contents of each of the multiple query items included in the authority submission materials and the authority inquiry document data, and send the generated learning dataset, thereby enabling the execution of the learning process for the language model.

[0655] In addition, the business support system 4500 according to the eighth embodiment:

[0656] · When generating a draft of the authority submission materials and sending the generated draft of the authority submission materials to the language model that has undergone the learning process, repeatedly perform the following processing:

[0657] The process of obtaining the query content,

[0658] The process of sending the revised draft of the authority submission materials, which has been revised based on the obtained query content, to the language model that has undergone the learning process.

[0659] · Repeatedly obtain the query content, and send the authority submission materials completed through repeated revisions to the authority.

[0660] In addition, the business support system 4700 according to the eighth embodiment:

[0661] · Send the query contents included in the multiple partition file data generated by partitioning the authority inquiry document data of the processing target to the language model that has undergone the learning process, and obtain the response contents for the query contents.

[0662] · When generating multiple completed input partition file data based on the obtained response contents, repeatedly perform the following processing:

[0663] The process of sending the multiple completed input partition file data to the language model that has undergone the learning process and obtaining the query contents for the multiple completed input partition file data,

[0664] The process of sending the obtained query contents to the language model that has undergone the learning process and obtaining the response contents for the obtained query contents.

[0665] · Based on the multiple completed input partition file data completed by repeatedly obtaining the response contents, generate the completed input file data and send it to the authority.

[0666] Thus, according to the business support system according to the eighth embodiment, it is possible to support the work of the staff who generate the completed input partition file data.

[0667] [Embodiment 9]

[0668] In the above-described seventh and eighth embodiments, an appropriate response is obtained from the language model by performing pre-learning processing on the language model and fine-tuning the language model. In contrast, in the ninth embodiment, it is configured to obtain an appropriate response from the language model regardless of whether the language model is fine-tuned. Specifically, the business support system according to the ninth embodiment is configured to transmit auxiliary information beneficial for generating an appropriate response content together when sending an instruction for generating response content, and the language model can generate response content while referring to this auxiliary information. Hereinafter, the ninth embodiment will be described centering on the differences from the above-described seventh or eighth embodiment.

[0669] <System Structure of Business Support System (Retrieval Phase)>

[0670] First, the system structure of the business support system according to the ninth embodiment will be described. In the ninth embodiment, the business support system executes processing in the retrieval phase of retrieving auxiliary information beneficial for generating response content and processing in the generation phase of generating input-completed file data based on the response content generated by referring to the auxiliary information. Here, the system structure of the business support system in the retrieval phase will be described.

[0671] Figure 49A FIG. 14, which shows an example of the system structure of the business support system, shows the system structure in the retrieval phase. The difference from the business support system 4000 described in the above-described seventh embodiment is that in the case of the business support system 4900: Figure 40 the difference from the business support system 4000 described in the above-described seventh embodiment is that in the case of the business support system 4900: Figure 49A the business support system 4900:

[0672] · The function of the data management device 4910 is different from the function of the data management device 4010,

[0673] · The data management device 4910 has an agent unit 4911,

[0674] · The server device 150 and the server device 4930 are connected to the internal network 170.

[0675] As shown in Figure 49A if the authority inquiry file data ((1)) is sent from the authority terminal 110, the management unit 121 of the data management device 4910 generates a plurality of divided file data by dividing the received authority inquiry file data for each inquiry item. In addition, the management unit 121 of the data management device 4910 sends the generated plurality of divided file data to the corresponding staff terminals 130, 140, etc. ((2)).

[0676] If divided file data is sent from the data management device 4910, the staff member 131 extracts the inquiry content from the divided file data and generates a prompt including the extracted inquiry content. In addition, the staff member 131 sends the generated prompt to the proxy section 4911 of the data management device 4910 via the staff terminal 130, thereby instructing the retrieval of auxiliary information ((3)) beneficial for generating a response content to the extracted inquiry content.

[0677] Based on the prompt received from the staff terminal 130, the proxy section 4911 determines the retrieval destination and sends a retrieval request for the auxiliary information to the determined retrieval destination. Figure 49A Examples of the retrieval destination when including the following devices, etc. as the retrieval auxiliary information are shown:

[0678] · The server device 150 storing the registered data including the divided file data, the input-completed divided file data, the input-completed file data, etc.

[0679] · Any server device 4930 other than the server device 150 connected to the internal network 170 (the server device 4930 having a database 4931 other than the specific database (other than the database 151)).

[0680] · Any Web server device 3720 accessible via the external network 160.

[0681] The proxy section 4911 sends a retrieval request for the auxiliary information to the server device 150, the server device 4930, and the Web server device 3720 ((4a), (4b), (4c)).

[0682] In addition, the proxy section 4911 can be configured to send a retrieval request to a preset retrieval destination, or can be configured to send a retrieval request after narrowing down the preset retrieval destination according to the inquiry content included in the prompt.

[0683] Alternatively, the staff member 131 can also generate a prompt guiding the retrieval destination and send it to the proxy section 4911. The prompt guiding the retrieval destination refers to a prompt specifying the object to be included in the retrieval destination, a prompt specifying the object to be excluded from the retrieval destination, etc.

[0684] In addition, the staff member 131 can also generate a prompt guiding the retrieval content and send it to the proxy section 4911. The prompt guiding the retrieval content refers to a prompt specifying the material to be retrieved (for example, the latest guideline), a prompt specifying the material to be excluded, etc.

[0685] <System Structure of the Business Support System (Generation Phase)>

[0686] Next, the system structure of the business support system in the generation stage will be described. Figure 49B FIG. 15 shows an example of the system structure of the business support system, showing the system structure in the generation stage.

[0687] As Figure 49B shown, in response to a retrieval request from the proxy unit 4911, the server device 150 sends the retrieval result to the proxy unit 4911 ((5a)). The retrieval result sent by the server device 150 includes, for example, a response content associated with past query contents similar to the query content included in the retrieval request. The response content sent by the server device 150 as the retrieval result is a response content (referred to as a reference response content) that serves as a reference when the language model generates a response content.

[0688] In addition, in response to a retrieval request from the proxy unit 4911, the server device 4930 sends the retrieval result to the proxy unit 4911 ((5b)). The retrieval result sent by the server device 4930 includes, for example, an enterprise retrieval result. The enterprise retrieval result refers to any retrieval result retrieved from a vast amount of digital data scattered within an enterprise.

[0689] In addition, in response to a retrieval request from the proxy unit 4911, the Web server device 3720 sends the Web retrieval result to the proxy unit 4911 ((5c)).

[0690] The proxy unit 4911 sends the retrieval result sent by the server device 150, the enterprise retrieval result sent by the server device 4930, and the Web retrieval result sent by the Web server device 3720 to the staff terminal 130 as auxiliary information ((6)).

[0691] If the staff terminal 130 receives the auxiliary information from the proxy unit 4911, the staff member 131 appropriately selects and rejects the auxiliary information. In addition, the staff member 131 generates a prompt including the query content and the selected and rejected auxiliary information, and sends it to the proxy unit 4911 of the data management device 4910 via the staff terminal 130 ((7)). Alternatively, the proxy unit 4911 may select and reject the auxiliary information and generate a prompt including the query content and the selected and rejected auxiliary information.

[0692] If a prompt including an inquiry content and auxiliary information selected by the agent unit is received from the staff terminal 130, the agent unit 4911 instructs the external server device 3420 to generate a response content for the inquiry content while referring to the auxiliary information. Additionally, the agent unit 4911 may also send a prompt including the inquiry content and the auxiliary information selected by the agent unit 4911 to the external server device 3420 to instruct it to generate a response content for the inquiry content while referring to the auxiliary information.

[0693] If an inquiry content, auxiliary information, and a generation instruction for generating a response content for the inquiry content while referring to the auxiliary information are received from the data management device 4910, the LM service providing unit 3421 of the external server device 3420 generates a response content according to the generation instruction.

[0694] The response content generated by the external server device 3420 is obtained by the agent unit 4911 of the data management device 4910 ((8)) and sent to the staff terminal 130 of the instruction source. Thus, the staff member 131 generates an input completion partition file data while referring to the response content generated by the language model and inputting the response content for the inquiry content.

[0695] The input completion partition file data generated by the staff member 131 is sent to the management unit 121 of the data management device 4910 via the staff terminal 130 ((9)).

[0696] The management unit 121 of the data management device 4910 generates an input completion file data based on the input completion partition file data collected from the staff terminal 130, the staff terminal 140, etc., and sends it to the authority terminal 110 via the external network 160 ((10)).

[0697] In this way, according to the business support system 4900, when the staff members 131, 141, etc. send a generation instruction for the response content, they can send the auxiliary information beneficial to generating an appropriate response content together. Or, according to the business support system 4900, the agent unit 4911 can also select the auxiliary information, generate a prompt including the inquiry content and the selected auxiliary information, and instruct the external server device 3420 to generate a response content for the inquiry content while referring to the auxiliary information.

[0698] Thereby, according to the business support system 4900, the language model can generate a response content while referring to the auxiliary information. As a result, regardless of whether the language model is fine-tuned, the staff members 131, 141, etc. can obtain an appropriate response content.

[0699] In addition, inFigure 49A and Figure 49B In the description of Figure 49B , it is assumed that the staff member 131 separately generates a prompt for retrieving auxiliary information and a prompt for generating a response content. However, the staff member 131 may also generate the prompt for retrieving auxiliary information and the prompt for generating a response content as an integrated prompt. In addition, the agency department 4911 may also generate the prompt for retrieving auxiliary information and the prompt for generating a response content as an integrated prompt without going through the staff member 131. In this case, if the collection of the auxiliary information is completed, the agency department 4911 automatically sends the collected auxiliary information together with the inquiry content to the external server device 3420. In addition, if the agency department 4911 receives the response content from the external server device 3420, it is sent to the staff terminal 130 of the instruction source.

[0700] <Business Support System Processing (Retrieval Phase and Generation Phase)>

[0701] Next, the processing flow in the retrieval phase and the generation phase of the business support system 4900 will be described. Figure 50A and Figure 50B are the first and second timing diagrams showing the processing flow in the retrieval phase and the generation phase of the business support system.

[0702] Among them, the processing of steps S1301~S1303, S1305~S1309 is the same as that of Figure 13 steps S1301~S1303, S1305~S1309, so the description is omitted here.

[0703] In step S5001, the staff terminals 130, 140, etc. respectively display the received divided file data.

[0704] In step S5002, in response to the display of the divided file data, when the staff terminals 130, 140, etc. receive a prompt input by the staff members 131, 141, etc. indicating the retrieval of auxiliary information beneficial to generating a response content to the inquiry content, the prompt is accepted. In addition, the staff terminals 130, 140, etc. send the accepted input prompt to the data management device 4910.

[0705] In step S5003, the data management device 4910 sends a retrieval request to the server device 150, the server device 4930, the Web server device 3720, etc.

[0706] In step S5004, the server device 150 sends the retrieval result including the reference response content to the data management device 4910.

[0707] In step S5005, the server device 4930 sends the enterprise search result to the data management device 4910.

[0708] In step S5006, the Web server device 3720 sends the Web search result to the data management device 4910.

[0709] In step S5007, the data management device 4910 sends the search result including the reference reply content, the enterprise search result, and the Web search result to the staff terminals 130, 140, etc. as auxiliary information.

[0710] In Figure 50B In step S5010, if the staff terminals 130, 140, etc. receive the input of the prompt including the inquiry content and the auxiliary information, it is sent to the data management device 4910.

[0711] In step S5011, the data management device 4910 instructs the external server device 3420 based on the prompt received from the staff terminals 130, 140, etc., so as to generate the reply content for the inquiry content while referring to the auxiliary information. In addition, the data management device 4910 may also instruct the external server device 3420 based on the prompt including the inquiry content and the auxiliary information selected by the proxy unit 4911 without going through step S5007 and step S5010, so as to generate the reply content for the inquiry content while referring to the auxiliary information. Further, the prompt may also include the inquiry content, the auxiliary information selected by the staff terminal 130, etc., and the auxiliary information selected by the proxy unit 4911.

[0712] In step S5012, the external server device 3420 generates the reply content according to the instruction from the data management device 4910 and sends it to the data management device 4910.

[0713] In step S5013, the data management device 4910 receives the reply content sent from the external server device 3420 and sends it to the staff terminals 130, 140, etc.

[0714] In step S5014, the staff terminals 130, 140, etc. display the reply content sent from the data management device 4910. The staff 131, 141, etc. generate the input-completed partition file data while referring to the reply content displayed on the staff terminals 130, 140, etc.

[0715] <Summary>

[0716] As can be seen from the above description, the business support system 4900 according to the 9th embodiment:

[0717] · For the document data subject to the authority inquiry of the processing object, in the case where a plurality of divided document data are generated and a retrieval instruction based on the inquiry content included in the plurality of divided document data is received, auxiliary information referred to by the language model when generating a response content for the inquiry content is collected.

[0718] · The response content included in the input-completed divided document data corresponding to the inquiry content, the enterprise retrieval results retrieved based on the inquiry content, and the Web retrieval results retrieved based on the inquiry content are collected as auxiliary information.

[0719] · The collected auxiliary information and the inquiry content are sent to the language model to obtain a response content for the inquiry content.

[0720] Thus, according to the business support system according to the 9th embodiment, regardless of whether the language model is fine-tuned, an appropriate response content can be obtained from the language model. In this way, according to the business support system according to the 9th embodiment, the work of the staff who generate the input-completed divided document data can be supported.

[0721] [Other Embodiments]

[0722] In addition, the present invention is not limited to the structures and combinations with other elements listed in the above embodiments, such as the structures shown here. Regarding these points, changes can be made without departing from the gist of the present invention, and can be appropriately determined according to its application form.

[0723] This application is based on Japanese Patent Application No. 2022-209989 filed on December 27, 2022 and claims the priority thereof. By referring to the entire content of the Japanese patent application, it is incorporated herein for this application.

[0724] Explanation of Reference Numerals

[0725] 100: Business Support System; 110: Authority Terminal; 120: Data Management Device; 121: Management Department; 130, 140: Staff Terminal; 150: Server Device; 151: Database; 200: Business Support System; 201: Retrieval Department; 400: Authority Inquiry File Data; 501: File Data Acquisition Department; 502: User Interface Department; 503: Division Department; 504: Allocation Department; 505: Attribute Information Extraction Department; 506: Collection Department; 507: File Data Generation Department; 508: File Data Sending Department; 610, 620: Divided File Data; 710, 720: Divided File Data after Input Completion; 800: File Data after Input Completion; 900: Attribute Information; 1000: Registration Data; 1101: Retrieval Screen Provision Department; 1102: Retrieval Condition Acquisition Department; 1103: Retrieval Control Department; 1210, 1220: Retrieval Screen; 1500: Business Support System; 1510: Learning Device; 1511: Learning Department; 1600: Business Support System; 1610: Prediction Device; 1611: Prediction Department; 1801: Storage Department; 2100: Business Support System; 2101: Retrieval Department; 2201: Retrieval Screen Provision Department; 2202: Extraction Department; 2203: First Vectorization Department; 2204: Separation Department; 2205: Second Vectorization Department; 2206: Similarity Calculation Department; 2207: Output Department; 2310: Retrieval Screen; 2401: Morpheme Analysis Department; 2402: Vectorization Model; 2411: Morpheme Analysis Department; 2412: Vectorization Model; 2601: Extraction Department; 2602: Filtering Department; 2710: Retrieval Screen; 2800: Business Support System; 2810: Analysis Device; 2811: Analysis Department; 3010: Feedback Information Acquisition Department; 3021: Additional Term Extraction Department; 3022: Deleted Term Extraction Department; 3023: Dictionary Update Department; 3031: Hyperparameter Update Department; 3041: Learning Data Generation Department; 3042: Learning Department; 3043: Model Update Department; 3201, 3202: Vectorization Models; 3211, 3212: Vectorization Models; 3400: Business Support System; 3410: Data Management Device; 3411: LM Learning Data Generation Department; 3420: External Server Device; 3421: LM Service Provision Department; 3500: LM Learning Dataset; 3700: Business Support System; 3710: Second LM Learning Data Generation Department; 3711: Learning Object Information Collection Department; 3712: Self-Supervised Learning Data Generation Department; 3721: First Generation Control Department; 3722: Second Generation Control Department; 3820: Second LM Learning Dataset; 4000: Business Support System; 4010: Data Management Device; 4011: LM Utilization Department; 4200: Business Support System; 4211: LM Learning Data Generation Department; 4220: Server Device; 4300: LM Learning Dataset; 4500: Business Support System4510: Data management device; 4511: Material generation unit; 4512: LM utilization unit; 4520: Server device; 4700: Business support system; 4710: Data management device; 4711: LM utilization unit; 4900: Business support system; 4910: Data management device; 4911: Agent unit; 4930: Server device.;

Claims

1. A business support system having: A partitioning unit that generates a plurality of partitioned file data by partitioning a plurality of inquiry items included in the file data of an authority inquiry for each inquiry item; A first acquisition unit that acquires a plurality of partitioned file data with input completed, in which response contents for the inquiry contents of each of the plurality of inquiry items are input; A second acquisition unit that acquires input-completed file data, which is file data in a single file format generated based on the plurality of input-completed partitioned file data acquired; And A storage unit that associates the acquired plurality of input-completed partitioned file data with the input-completed file data respectively and stores them in a database.

2. The business support system according to claim 1, further having: An attribute information extraction unit that extracts attribute information from the file data of the authority inquiry or the input-completed file data, The storage unit associates the acquired plurality of input-completed partitioned file data with the attribute information and the input-completed file data respectively and stores them in the database.

3. The business support system according to claim 2, wherein The attribute information includes at least one piece of information selected from the inquiry period of the authority inquiry, the disease field of the inquiry content, the event being inquired about, the variety of the inquiry content, and the response period for the authority inquiry.

4. The business support system according to claim 2, further having: A third acquisition unit that acquires keywords recalled based on each inquiry content and response content included in the acquired input-completed partitioned file data, The storage unit associates the acquired plurality of input-completed partitioned file data with the keywords, the attribute information, and the input-completed file data respectively and stores them in the database.

5. The business support system according to claim 4, further having: A storage unit that stores a learned model learned using a learning dataset. Among them, This learning dataset includes: Each inquiry content and response content included in the input-completed partitioned file data; and Keywords recalled based on each inquiry content and response content included in the input-completed partitioned file data, The third acquisition unit acquires keywords recalled based on each inquiry content and response content included in the acquired input-completed file data by inputting each inquiry content and response content included in the acquired input-completed file data into the learned model.

6. The business support system according to claim 4 or 5, further having: A retrieval unit that retrieves input-completed partitioned file data that matches a specified condition from the acquired plurality of input-completed partitioned file data stored in the database.

7. The business support system according to claim 6, wherein Among the input-completed partitioned file data that matches the specified condition, any one of the following input-completed partitioned file data is included: Input-completed partitioned file data including a specified string; or Input-completed partitioned file data in which the associated keyword matches the specified string; or The input-completed divided document data in which the inquiry period of the authority inquiry included in the associated attribute information or the response period to the authority inquiry is included in a specified period; or The input-completed divided document data in which the disease field included in the associated attribute information is a specified disease field; or The input-completed divided document data in which the event being inquired about included in the associated attribute information is a specified event; Or The input-completed divided document data in which the variety included in the associated attribute information is a specified variety.

8. The service support system according to claim 6, wherein when receiving an instruction for detailed display, the retrieval unit outputs the attribute information, the input-completed document data, or the keyword that is associated with and stored in the retrieved input-completed divided document data.

9. The service support system according to claim 1, further comprising: a first generation unit that generates a first learning dataset obtained by associating the inquiry content and the response content included in the multiple input-completed divided document data obtained by the first acquisition unit, and transmits the generated first learning dataset to instruct the learning process of the language model.

10. The service support system according to claim 9, further comprising: a second generation unit that generates a second learning dataset for self-supervised learning by collecting information helpful for generating the response content at a specific period via a network, and instructs the learning process of the language model based on the generated second learning dataset.

11. The service support system according to claim 10, further comprising: a first generation control unit that selects, based on the second learning dataset, the combination of the inquiry content and the response content included in the first learning dataset; and a second generation control unit that changes the collection source of the information helpful for generating the response content based on the first learning dataset.

12. The service support system according to claim 9, further comprising: a utilization unit that, when the divided unit generates multiple divided document data for the document data of the authority inquiry for the processing target, transmits the inquiry content of each of the multiple inquiry items to the language model that has undergone the learning process, obtains the response content for the inquiry content of each of the multiple inquiry items, and the first acquisition unit acquires the multiple input-completed divided document data generated based on the response content obtained by the utilization unit.

13. The service support system according to claim 1, further comprising: a generation unit that generates a learning dataset obtained by associating the submission document data submitted to the authority and the inquiry content of each of the multiple inquiry items included in the document data of the authority inquiry, and transmits the generated learning dataset to instruct the learning process of the language model.

14. The service support system according to claim 13, further comprising: a material generation unit that generates the submission document data to be submitted to the authority; and a utilization unit that, when transmitting the submission document data to be submitted to the authority to the language model that has undergone the learning process, repeatedly performs the following process: Processing of obtaining inquiry content for submission file data submitted to the authority; and Processing of sending the submission file data submitted to the authority, which has been corrected based on the inquiry content for the submission file data submitted to the authority, to the language model that has undergone the learning process, The material generation unit sends the submission file data submitted to the authority, which has been corrected by repeatedly obtaining the inquiry content by the utilization unit, to the authority.

15. The business support system according to claim 13, further comprising: A utilization unit, in the case of file data of an authority inquiry for a processing target, when sending the inquiry content included in the multiple divided file data generated by the dividing unit to the language model that has undergone the learning process and obtaining a reply content for the generated inquiry content, thereby generating multiple input-completed divided file data, repeatedly performs the following processing: Processing of sending the multiple input-completed divided file data to the language model that has undergone the learning process and obtaining the inquiry content for the multiple input-completed divided file data; and Processing of sending the obtained inquiry content to the language model that has undergone the learning process and obtaining a reply content for the obtained inquiry content, The first acquisition unit acquires the multiple input-completed divided file data generated by repeatedly obtaining the reply content by the utilization unit.

16. The business support system according to claim 9 or 13, further comprising: An agency unit, in the case of file data of an authority inquiry for a processing target, when receiving a retrieval instruction based on the inquiry content included in the multiple divided file data generated by the dividing unit, collects auxiliary information referred to by the language model that has undergone the learning process when generating a reply content for the inquiry content.

17. The business support system according to claim 16, wherein The auxiliary information includes any one of the reply content included in the input-completed divided file data corresponding to the inquiry content of the retrieval instruction, which is a retrieval result collected by retrieving the database, or an enterprise retrieval result collected by retrieving a database other than the database, or a Web retrieval result collected by retrieving a Web server device.

18. The business support system according to claim 17, wherein The agency unit sends the inquiry content included in the file data of the authority inquiry for the processing target and the collected auxiliary information to the language model, and obtains a reply content for the inquiry content included in the file data of the authority inquiry for the processing target.

19. The business support system according to claim 9 or 13, wherein The language model is a large language model.

20. A service support method, wherein, The following steps are performed by a computer: A generation step of generating multiple divided file data by dividing multiple inquiry items included in the file data of the authority inquiry by each inquiry item; A first acquisition step of acquiring multiple input-completed divided file data input with a reply content for each of the multiple inquiry items; The second acquisition step of acquiring the input-completed file data which is file data in the form of one file generated based on the acquired multiple input-completed divided file data; and a storage step of associating the acquired multiple input-completed divided file data with the input-completed file data respectively and storing them in a database.

21. A data management program for causing a computer to execute the following steps: a generation step of generating multiple divided file data by dividing multiple inquiry items included in the file data of an authority inquiry by each inquiry item; the first acquisition step of acquiring multiple input-completed divided file data in which response contents for the inquiry contents of each of the multiple inquiry items are input; the second acquisition step of acquiring the input-completed file data which is file data in the form of one file generated based on the acquired multiple input-completed divided file data; and a storage step of associating the acquired multiple input-completed divided file data with the input-completed file data respectively and storing them in a database.

22. A business support system having: A database stores registration data of a plurality of input-completed divided file data in which response content has been input for each of the divided file data for a plurality of divisions, where The multiple divided file data are generated by dividing the inquiry contents of each of the multiple inquiry items included in the file data of an authority inquiry by each inquiry item; a retrieval unit that, in the database, retrieves an inquiry content having vector data satisfying a specified similarity condition by calculating a similarity with vector data obtained by vectorizing the inquiry content included in the file data of an authority inquiry to be processed; and an output unit that outputs the input-completed divided file data corresponding to the retrieved inquiry content or data other than the input-completed divided file data included in the registered data corresponding to the retrieved inquiry content.

23. The business support system according to claim 22, wherein the retrieval unit has: a morpheme analysis unit that performs morpheme analysis on the inquiry content included in the file data of an authority inquiry to be processed using a dictionary including terms inherent in drug development; a vectorization unit that vectorizes the inquiry content included in the file data of an authority inquiry to be processed based on each word extracted from the inquiry content included in the file data of an authority inquiry to be processed by morpheme analysis; and a similarity calculation unit that calculates a similarity with the vector data generated based on the inquiry content included in the file data of an authority inquiry to be processed by vectorization.

24. The business support system according to claim 23, wherein the vectorization unit vectorizes with values corresponding to the occurrence frequencies of each word extracted from the inquiry content included in the file data of an authority inquiry to be processed.

25. The business support system according to claim 24, wherein the vectorization unit vectorizes based on TF-IDF.

26. The business support system according to claim 22, wherein the database also associates the following with the multiple input-completed divided file data and stores them as the registered data: the input-completed file data which is file data in the form of one file generated based on the multiple input-completed divided file data; Attribute information extracted from the document data queried from the authority or the input-completed document data; and Keywords recalled based on each query content and response content included in the multiple input-completed divided document data.

27. The service support system according to claim 26, wherein at least one of the following information is included in the attribute information: the query period of the authority query, the disease field of the query content, the event being queried, the variety of the query content, and the response period for the authority query.

28. The service support system according to claim 27, wherein the retrieval unit retrieves the query content included in the input-completed divided document data that satisfies a specified filtering condition among the multiple input-completed divided document data stored in the database.

29. The service support system according to claim 28, wherein among the input-completed divided document data that satisfies the specified filtering condition, any one of the following input-completed divided document data is included: input-completed divided document data containing a specified string; or input-completed divided document data in which the associated keyword is consistent with a specified string; or input-completed divided document data in which the query period of the authority query or the response period for the authority query included in the associated attribute information is included in a specified period; or input-completed divided document data in which the disease field included in the associated attribute information is a specified disease field; or input-completed divided document data in which the event being queried included in the associated attribute information is a specified event; or input-completed divided document data in which the variety included in the associated attribute information is a specified variety.

30. The service support system according to claim 26, wherein when an instruction for detailed display is received, the output unit outputs the attribute information, the input-completed document data, or the keyword that is associated with and stored in the input-completed divided document data corresponding to the retrieved query content.

31. The service support system according to claim 23, further comprising: a first update unit that updates a dictionary containing the inherent terms in the drug development based on the evaluation result for the retrieved query content.

32. The service support system according to claim 23, further comprising: a second update unit that updates the parameters used by the vectorization unit when performing vectorization based on the evaluation result for the retrieved query content.

33. The service support system according to claim 32, further comprising: The learning department generates a learned model by performing learning processing using a learning dataset. Among them, the learning dataset includes: the query content included in the document data queried by the authority, the retrieved query content, and the evaluation result; and a third update unit that updates the vectorization unit using the learned model.

34. The service support system according to claim 33, wherein The learning unit learns the model in such a manner that the similarity between the vector data output by inputting the query content included in the document data queried by the authority read from the learning dataset into the model and the vector data output by inputting the retrieved query content read from the learning dataset into the model approaches the similarity corresponding to the evaluation result read from the learning dataset, thereby generating the learned model.

35. The service support system according to claim 22, further comprising: An agency unit that, when receiving a retrieval instruction based on the query content included in the document data of the authority query for the processing target, collects auxiliary information referred to by the language model when generating a response content for the query content.

36. The service support system according to claim 35, wherein The auxiliary information includes, as a retrieval result collected by retrieving the database, a response content included in the input-completed divided document data corresponding to the query content of the retrieval instruction, or an enterprise retrieval result collected by retrieving a database other than the database, or a Web retrieval result collected by retrieving a Web server device.

37. The service support system according to claim 36, wherein The agency unit sends the query content included in the document data of the authority query for the processing target and the collected auxiliary information to the language model, and obtains a response content for the query content included in the document data of the authority query for the processing target.

38. The service support system according to claim 35, wherein The language model is a language model that has been learned using a learning dataset obtained by associating the query content and the response content included in each of the plurality of input-completed divided document data, or a learning dataset obtained by associating the submission document data submitted to the authority and the query content of each of the plurality of query items included in the document data of the authority query.

39. A service support method, wherein A computer of a server device having a database executes the following steps. The database stores registration data of a plurality of input-completed divided document data in which response contents are input for each of a plurality of divided document data, and the plurality of divided document data are generated by dividing the query content of each of the plurality of query items included in the document data of the authority query for each query item: A retrieval step of retrieving, in the database, a query content having vector data that satisfies a specified similarity condition by calculating the similarity with the vector data obtained by vectorizing the query content included in the document data of the authority query for the processing target; and An output step of outputting the input-completed divided document data corresponding to the retrieved query content, or data other than the input-completed divided document data included in the registration data corresponding to the retrieved query content.

40. A retrieval program for causing a computer of a server device having a database to execute the following steps, the database storing registration data of a plurality of input-completed divided file data in which response contents are input for each of the plurality of divided file data, wherein, The plurality of divided document data are generated by dividing the query content of each of the plurality of query items included in the document data of the authority query for each query item: A retrieval step of calculating a similarity of vector data obtained by vectorizing query content included in document data queried from an authority of a processing object, and retrieving, in the database, query content having vector data that satisfies a specified similarity condition; and An output step of outputting input-completed division document data corresponding to the retrieved query content, or data other than the input-completed division document data included in the registration data corresponding to the retrieved query content.

Images