Prompt generation system and prompt management method

Abstract

This system provides a prompt generation system capable of accurately extracting desired information from documents. [Solution] The data processing unit 20 has a prompt statement that specifies the information to be extracted from the document 31. The data processing unit 20 generates an input prompt including the prompt statement and inputs it to the generating AI model 3. The data processing unit 20 also determines any additional information included in the extracted information 33 extracted by the generating AI model 3 in response to the input prompt that does not correspond to the information to be extracted, and adds a statement to the prompt statement that specifies new information to be extracted according to that additional information.

Description

Technical Field

[0001] This disclosure relates to a prompt generation system and a prompt management method.

Background Art

[0002] In recent years, for the purpose of business efficiency improvement and the like, it has been desired to automatically extract desired information from documents such as product catalogs, system specifications, parts lists, and manuals. However, since there are various formats in documents, it is not easy to accurately extract desired information.

[0003] In contrast, Patent Document 1 discloses a form processing apparatus that reads description information described in a form image indicating a form. In the form processing apparatus, based on the form image, the format of the form is specified, and among a plurality of pre-registered form patterns, the description information is read using a form pattern having the same format as the specified format.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in the technique described in Patent Document 1, since it is necessary to pre-register form patterns in which the format is defined, there is a problem that it is difficult to apply when the number of formats is very large or when a document having an unknown format is included.

[0006] Another approach is to use generative AI (Artificial Intelligence) models, which have become increasingly popular in recent years, to extract desired information from documents. However, accurately extracting desired information using generative AI models requires appropriately adjusting the prompts input to the model, which is not easy.

[0007] The purpose of this disclosure is to provide a prompt generation system and a prompt management method that can accurately extract desired information from a document. [Means for solving the problem]

[0008] A prompt generation system according to one aspect of the present disclosure is a prompt generation system that generates input prompts for a generating AI model to extract information from a document, comprising a processor and memory, wherein the memory stores prompt statements that specify information to be extracted from the document, the processor generates the input prompt including the prompt statements and inputs it to the generating AI model, the generating AI model identifies additional information included in the extracted information extracted in response to the input prompt that does not correspond to the information to be extracted, and appends a statement to the prompt statement that specifies new information to be extracted in response to the additional information. [Effects of the Invention]

[0009] According to the present invention, it becomes possible to extract desired information from documents with high accuracy. [Brief explanation of the drawing]

[0010] [Figure 1] This figure shows the functional configuration of the prompt generation system of the first embodiment of the present disclosure. [Figure 2] This figure shows an example of a document. [Figure 3] This figure shows an example of an item type table. [Figure 4] This figure shows an example of extracted information. [Figure 5] It is a diagram showing an example of discrimination result information. [Figure 6] It is a diagram showing an example of a document input UI. [Figure 7] It is a diagram showing an example of a table editing UI. [Figure 8] It is a diagram showing an example of an extraction information browsing UI. [Figure 9] It is a diagram showing an example of the hardware configuration of a prompt generation system. [Figure 10] It is a flowchart for explaining an example of extraction processing. [Figure 11] It is a diagram for explaining a specific example of determination processing. [Figure 12] It is a diagram for explaining an example of update processing. [Figure 13] It is a flowchart for explaining an example of update processing. [Figure 14] It is a diagram showing the functional configuration of the prompt generation system according to the second embodiment of the present disclosure. [Figure 15] It is a diagram showing an example of a document type table. [Figure 16] It is a diagram showing another example of update processing.

Mode for Carrying Out the Invention

[0011] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0012] (First Embodiment) FIG. 1 is a diagram showing the functional configuration of a prompt generation system 1 according to the first embodiment of the present disclosure. The prompt generation system 1 shown in FIG. 1 includes an input / output unit 10 and a data processing unit 20.

[0013] The input / output unit 10 performs input / output of information with the user 2 who uses the prompt generation system 1. Specifically, the input / output unit 10 presents the document input UI (User Interface) 11, the table editing UI 12, and the extracted information browsing UI 13 to the user 2, and receives various information through these UIs.

[0014] The document input UI 11 is an interface for inputting the document 31 which is the extraction source for extracting the desired information. The document 31 is a product catalog or a system specification, etc. In the present embodiment, in the document 31, for each item related to a predetermined object, item information corresponding to the item is described. There may be a plurality of predetermined objects. For example, when the document 31 is a product catalog, each product is the object. Also, when the document 31 is a system specification, each function of the system, each component (such as hardware) or both of them, etc. are the objects. The item information is, for example, a description text or a value indicating the content of the item.

[0015] The table editing UI 12 is an interface for editing the item type table 32 which is extraction management information regarding the extraction target information to be extracted from the document 31. In the present embodiment, the extraction target information is item information corresponding to a predetermined target item. There may be a plurality of target items. The item type table 32 shows, for each target item, an item type which is a type (attribute) indicating the characteristics of the target item, a discrimination criterion for discriminating the target item, and a prompt text for specifying the extraction target information to the generation AI model 3. The item type indicates, for example, "format" or "language", etc. The discrimination criterion specifically indicates the characteristics of the target item.

[0016] The extracted information browsing UI 13 is an interface for outputting the extracted information 33 which is the information extracted from the document 31 so that the user 2 can view it.

[0017] The data processing unit 20 is a processing unit that inputs input prompts to the generating AI model 3 to extract information to be extracted from the document 31, and retrieves the information extracted by the generating AI model 3 from the document 31. Specifically, the data processing unit 20 includes a document reading unit 21, a prompt input unit 22, and a condition determination unit 23.

[0018] The document reading unit 21 loads the document 31 into the generation AI model 3 and instructs the generation AI model 3 to identify the item type of the target item contained in the document 31 using the discrimination criteria in the item type table 32.

[0019] The prompt input unit 22 inputs an input prompt containing a prompt statement corresponding to the target item to the generating AI model 3 based on the item type table 32, causing the generating AI model 3 to extract the item information of the target item from the document 31 as extracted information 33.

[0020] The condition determination unit 23 updates the prompt statements in the item type table 32 based on the item type table 32 and the extraction information 33. Specifically, for each target item, the condition determination unit 23 determines any additional information that does not correspond to the extraction target information specified by the prompt statements in the extraction information 33, and generates judgment result information 34 indicating that additional information. Then, the condition determination unit 23 updates the prompt statements in the item type table 32 based on the judgment result information 34. Specifically, the condition determination unit 23 adds a statement to the prompt statement that specifies new extraction target information corresponding to the additional information.

[0021] Figure 2 shows an example of document 31. In the document 31 shown in Figure 2, item information 312 is described for each item 311. In this embodiment, document 31 is divided into pages for each object, and Figure 2 shows the first page of document 31.

[0022] Figure 3 shows an example of an item type table 32. The item type table 32 shown in Figure 3 contains fields 321 to 326 for each record.

[0023] Field 321 stores the item type of the target item. Field 322 stores the name of the target item. Field 323 stores the characteristics of the target item as a discrimination criterion for identifying the target item. Field 324 stores a prompt statement for specifying the item information of the target item as information to be extracted to the generating AI model 3. Field 325 stores an additional template, which is template information for additional statements to be added to the prompt statement. Field 326 stores a programmatic judgment condition, which is a judgment condition for determining whether or not the prompt statement needs to be updated.

[0024] Figure 4 shows an example of extracted information 33. The extracted information 33 shown in Figure 4 includes 331 to 336 for each record.

[0025] Field 331 stores a code, which is identification information for identifying the object. Fields 332 to 336 store item information for the object. In the example in Figure 3, field 332 stores the name of the object, field 333 stores the model number of the object, field 334 stores the color of the object, field 335 stores the price of the object, and field 336 stores the characteristics of the object.

[0026] Figure 5 shows an example of the discrimination result information 34. The discrimination result information 34 shown in Figure 5 includes fields 341 to 346 for each record.

[0027] Field 341 stores a code, which is identification information for identifying the object. Fields 342 to 346 store the results of the determination of additional information for each of the object's items. In the example in Figure 4, field 332 stores the determination results for the product name, field 333 for the model number, field 334 for the color, field 335 for the price, and field 336 for the features. The determination result shows "○" if there is no additional information, and if there is additional information, it shows that additional information.

[0028] Figure 6 shows an example of a document input UI 11. The document input UI 11 shown in Figure 6 includes a document selection button 111, a document preview display button 112, a document deselection button 113, a document selection confirmation button 114, and a document preview display screen 115.

[0029] The document selection button 111 is used to select document 31 to be input into the prompt generation system 1. The document preview display button 112 is used to preview document 31 selected by document selection button 111. The document deselection button 113 is used to deselect document 31 by document selection button 111. The document selection confirmation button 114 is used to confirm the selection of document 31 by document selection button 111 and input that document 31 into the prompt generation system 1. The document preview display screen 115 is an area that displays a preview of document 31 selected by document selection button 111, for example, the document 31 shown in Figure 2.

[0030] Figure 7 shows an example of the table editing UI 12. The document input UI 11 shown in Figure 7 includes a table information display button 121, a table information registration button 122, a table information deletion button 123, and a table information display screen 124.

[0031] The Table Information Display button 121 is a button for displaying the Item Type Table 32. The Table Information Registration button 122 is a button for registering the Item Type Table 32. The Table Information Delete button 123 is a button for deleting the Item Type Table 32. The Table Information Display screen 124 is an area for displaying the Item Type Table 32, and the Item Type Table 32 displayed on the Table Information Display screen 124 may be edited by user 2.

[0032] Figure 8 shows an example of the extracted information viewing UI 13. The extracted information viewing UI 13 shown in Figure 8 includes an extracted information selection button 131, an extracted information display button 132, an extracted information deselection button 133, and an extracted information display screen 134.

[0033] The Extraction Information Selection Button 131 is a button for selecting Extraction Information 33. The Extraction Information Display Button 132 is a button for displaying the Extraction Information 33 selected by the Extraction Information Selection Button 131. The Extraction Information Deselection Button 133 is a button for deselecting the Extraction Information 33 selected by the Extraction Information Selection Button 131. The Extraction Information Display Screen 135 is an area for displaying the Extraction Information 33 selected by the Extraction Information Selection Button 131.

[0034] Figure 9 shows an example of the hardware configuration of the prompt generation system 1. As shown in Figure 9, the prompt generation system 1 has a storage device 51, main memory 52, processor 53, input device 54, display device 55, and communication device 56, which are connected via a bus 57.

[0035] The storage device 51 is a device that records data in a writable and readable manner, and stores a program that defines the operation of the processor 53, and various information used and generated by that program. The main memory 52 is used as the work area of ​​the processor 53 and stores at least temporarily various information used and generated by that program. Various information such as the document 31, item type table 32, extraction information 33, and discrimination result information 34 shown in Figure 1 are stored in at least the storage device 51 or the main memory 52 that constitute the memory of the prompt generation system 1.

[0036] The processor 53 reads the program stored in the storage device 51 into the main memory 52 and uses the main memory 52 as a work area to implement the functional units corresponding to the program. Specifically, the processor 53 implements each functional unit of the prompt generation system 1 shown in Figure 1 (such as the document reading unit 21, the prompt input unit 22, and the condition determination unit 23). Therefore, in this specification, the subject of the processing performed by each functional unit as the operating entity may be the processor 53.

[0037] The input device 54 is a device that receives various information from the user 2 of the prompt generation system 1, and this information is used by the processor 53. The display device 55 is a device that displays various information such as the UI 11 to 13 shown in Figure 1. The communication device 56 is connected to an external device via a network 58 or the like and transmits and receives information with that external device. In this embodiment, an example of an external device is an AI server 59 having a generation AI model 3. The generation AI model 3 may be provided inside the prompt generation system 1.

[0038] Figure 10 is a flowchart illustrating an example of an extraction process that extracts target information using a prompt generation system.

[0039] In the extraction process, the document reading unit 21 of the data processing unit 20 reads the document 31, which was input using the document input UI 11, into the generation AI model 3 (step S101). Then, the document reading unit 21 instructs the generation AI model 3 to determine, based on the item type table 32, which items in the document 31 match the discrimination criteria and are designated as target items (step S102).

[0040] The prompt input unit 22 selects a prompt statement corresponding to the target item from the item type table 32 (step S103), generates an input prompt containing the selected prompt statement, and inputs it to the generated AI model 3 (step S104). In addition to the selected prompt statement, the input prompt may also include other prompt statements, such as common statements common to all target items. A common statement might be, for example, "There may be no information in the relevant item. In that case, output a blank field." The common statement may also be defined within the item type table 32.

[0041] The generating AI model 3 extracts and outputs item information for the target item from document 31 in response to the input prompt. The condition determination unit 23 acquires the information output from the generating AI model 3 as extracted information 33 and displays it using the extracted information viewing UI 13 of the input / output unit 10 (step S105). Note that the generating AI model 3 may extract additional information in response to the input prompt, in addition to the information directly specified by the input prompt.

[0042] The condition determination unit 23, for each target item, determines additional information that does not correspond to the extraction target information included in the item information in the extraction information 33 based on the programmatic determination conditions corresponding to the target item in the item type table 32, and executes a determination process to generate determination result information 34 that shows the determination result (step S106).

[0043] The condition determination unit 23 performs an update process to update the item type table 32 based on the determination result information 34 (step S107), and then terminates the process.

[0044] Figure 11 is a diagram illustrating a specific example of the determination process in step S106 of Figure 10.

[0045] In the determination process, the condition determination unit 23 determines whether the item information of each target item included in the extracted information 33 (i.e., each field of the extracted information 33) matches the programmatic determination conditions corresponding to that target item. The condition determination unit 23 then identifies item information that does not match the programmatic determination conditions as additional information that does not correspond to the extracted target items.

[0046] For example, the item information for "Features" of code "1" is "●Waterproof, ●Washable," which matches the program's judgment condition ("text[0]=●or■", meaning the beginning of the item information (list marker) starts with either "●" or "■"). In this case, the judgment result corresponding to the "Features" of code "1" in the judgment result information 34 is "〇".

[0047] On the other hand, the item information for the feature of code "3" is "▲For rainy days," which does not match the program's judgment condition. In other words, the generating AI model 3 also supplemented the item information of the "feature" with the bulleted sentence starting with "▲" which was not directly specified in the input prompt. In this case, the condition judgment unit 23 determines that the item information does not match the program's judgment condition and identifies "▲For rainy days" as additional information that does not match the information to be extracted. The condition judgment unit 23 stores the program judgment condition "text[0]=▲" indicating the additional information in the judgment result information 34 as the judgment result corresponding to the "feature" of code "3." Similarly, the judgment result corresponding to the "type" of code "3" in the judgment result information 34 is "character code=hiragana."

[0048] Figure 12 is a diagram illustrating an example of the update process in step S107 of Figure 10, and Figure 13 is a flowchart illustrating an example of the update process.

[0049] As shown in Figure 12, the update process includes a condition update process S2 that updates the programmatic determination conditions in the item type table 32 based on the determination result information 34, and a prompt update process S3 that updates the prompt statement based on the updated programmatic determination conditions and additional templates.

[0050] The condition determination unit 23 first performs a condition update process S2, which determines whether or not there is a determination result other than "○" in the determination result information 34, that is, whether or not additional information exists (step S201).

[0051] If no additional information exists (step S201: No), the condition determination unit 23 terminates the update process.

[0052] On the other hand, if additional information exists (step S201: Yes), a new condition corresponding to the determination result is added to the programmatic determination condition in the item type table 32 that corresponds to the determination result that additional information exists (the programmatic determination condition in the item type table 32 for rows whose row name matches the column name of the field in the determination result information 34 in which the additional information exists) (step S202). Specifically, the new condition is a condition for determining that the additional information is information to be extracted.

[0053] For example, in the example in Figure 12, the discrimination result corresponding to the item "Features" for code "3" in the discrimination result information 34 is "text[0]=▲". Therefore, "text[0]=▲" is added to the programmatic judgment condition "text[0]=●or■" corresponding to the target item "Features" in the item type table 32. As a result, the programmatic judgment condition becomes "text[0]=●or■or▲". Similarly, the discrimination result corresponding to the item "Type" for code "3" in the discrimination result information 34 is "Character code=Hiragana". Therefore, "Character code=Hiragana" is added to the programmatic judgment condition "Character code=Number" corresponding to the target item "Format" in the item type table 32. As a result, the programmatic judgment condition becomes "Character code=Number or Hiragana".

[0054] When the processing in step S202 is completed, the condition determination unit 23 moves to the prompt update process S3 and generates a temporary prompt statement (step S301) in which the value of the condition added to the programmatic determination condition in step S202 is inserted into the insertion location (placeholder) in the additional template corresponding to the programmatic determination condition to which the condition was added.

[0055] For example, in the example in Figure 12, the additional template for the target item "Features" is "May also start with XX," where "XX" indicates the insertion point. In this case, the new condition added to the programmatic judgment condition for the target item "Features" is "text[0]=▲," so the temporary prompt statement 41 corresponding to the target item "Features" becomes "May also start with ▲." Similarly, the temporary prompt statement for the target item "Format" becomes "May also be written in hiragana."

[0056] Then, the condition determination unit 23 appends the temporary prompt statement to the prompt statement (step S302) and terminates the process. For example, in the example in Figure 12, the prompt statement for the target item "Features" has the temporary prompt statement "May also start with ▲." appended to it, so it becomes "The relevant item may be in the form of a black circle bulleted list starting with ●, a black square bulleted list starting with ■, etc. May also start with ▲."

[0057] As described above, according to this embodiment, the main memory 52 stores a prompt statement that describes the information to be extracted from the document 31. The processor 53 generates an input prompt including the prompt statement and inputs it to the generating AI model 3. The processor 53 also identifies additional information included in the extracted information 33 extracted by the generating AI model 3 in response to the input prompt that does not correspond to the information to be extracted, and adds an additional statement to the prompt statement that specifies that additional information as the information to be extracted. Therefore, it becomes possible to optimize the input prompt to allow the generating AI model 3 to extract the desired information to be extracted, and thus it becomes possible to extract the desired information from the document 31 with high accuracy.

[0058] Furthermore, in this embodiment, the main memory 52 stores a prompt statement that describes the item information corresponding to each predetermined target item included in the document 31 as the information to be extracted. Therefore, it is possible to extract the desired item information with high accuracy.

[0059] In this embodiment, the main memory 52 stores discrimination criteria, which are characteristics of each target item. The processor 53 instructs the generating AI model 3 to identify items in the document 31 that match the discrimination criteria as target items, and generates an input prompt that includes a prompt statement corresponding to the identified target item. In this case, since an input prompt containing a prompt statement corresponding to the target item in the document can be input to the generating AI model 3, it becomes possible to extract the desired item information with greater accuracy.

[0060] Furthermore, in this embodiment, the main memory 52 stores programmatic determination conditions for determining the information to be extracted included in the extracted information 33, and the processor 53 determines, based on the programmatic determination conditions, that information included in the extracted information 33 that does not match the programmatic determination conditions is additional information, and adds new conditions to the determination conditions for determining that the additional information is information to be extracted. In this case, it becomes possible to improve the accuracy of the determination of the additional information.

[0061] Furthermore, in this embodiment, the processor 53 uses an additional template to add a statement to the prompt that explains the additional information as information to be extracted. This makes it possible to update the prompt statement appropriately.

[0062] (Second embodiment) Figure 14 shows the functional configuration of the prompt generation system 1 of the second embodiment of this disclosure. The prompt generation system 1 shown in Figure 14 differs from the prompt generation system 1 of the first embodiment shown in Figure 1 in that it further has a document type table 35 that shows the characteristics of the document type, which is the type of document 31, for each document type. In addition, an item type table 32 is provided for each document type. Each item type table 32 has discrimination criteria and prompt statements corresponding to the document type.

[0063] The document type table 35 is stored, for example, in the storage device 51 shown in Figure 9. Alternatively, the table editing UI 12 may be provided with a function for editing the document type table 35, or a separate UI for editing the document type table 35 may be provided in addition to the table editing UI 12.

[0064] Figure 15 shows an example of a document type table 35. The document type table 35 shown in Figure 15 includes fields 351 and 352 for each record.

[0065] Field 351 stores the document type. Field 352 stores feature information that describes the characteristics of the document type.

[0066] In this embodiment, when the document reading unit 21 reads the document 31 into the generation AI model 3, it instructs the generation AI model 3 to identify the document type of the document 31 based on the document type table 35. Based on the item type table 32 corresponding to the identified document type, the document reading unit 21 determines that the items in the document 31 that match the discrimination criteria are the target items.

[0067] According to this embodiment, it becomes possible to use an appropriate input prompt depending on the type of document 31, thereby enabling more accurate extraction of desired information from the document 31.

[0068] (Third embodiment) In the first embodiment, if additional information existed in the extracted information 33, a new condition was added to the prompt statement in the item type table 32. In contrast, this embodiment differs from the first embodiment in that the prompt statement may be modified (for example, deleted).

[0069] The prompt generation system 1, for example, performs an extraction process as described in Figure 10, etc., each time a document 31 is input, and obtains multiple pieces of extracted information 33. The condition determination unit 23 calculates an extraction rate, which is the percentage of information that matches the programmatic determination condition, based on the multiple pieces of extracted information 33 and the item type table 32. If the extraction rate is less than a threshold, it deletes the statement that specifies the information that matches the programmatic determination condition from the prompt statement in the item type table 32.

[0070] Figure 16 shows an example of the update process in this embodiment.

[0071] In the example in Figure 16, the proportion of item information starting with "●" or "■" in the "Features" item of extracted information 33 is below the threshold, and there is additional information starting with "▲". In this case, the condition determination unit 23 updates the input prompt corresponding to "Features" in the item type table 32 from "The relevant item has formats such as black circle bullet points starting with ●, black square bullet points starting with ■, etc." to "The relevant item has formats such as black triangle bullet points starting with ▲, etc.".

[0072] In this embodiment, it becomes possible to optimize the prompt message.

[0073] The embodiments of the Disclosure described above are illustrative for illustrative purposes and are not intended to limit the scope of the Disclosure to those embodiments only. Those skilled in the art can implement the Disclosure in various other forms without departing from the scope of the Disclosure. [Explanation of Symbols]

[0074] 1: Prompt generation system 2: User 3: Generated AI model 10: Input / output unit 20: Data processing unit 21: Document reading unit 22: Prompt input unit 23: Condition judgment unit 31: Document 32: Item type table 33: Extracted information 34: Discrimination result information 35: Document type table 51: Storage device 52: Processor 53: Memory 54: Input device 55: Display device 56: Communication device 57: Bus 58: Network 59: AI server

Claims

1. A prompt generation system that generates input prompts for a generative AI model to extract information from a document, It has a processor and memory, The memory stores a prompt statement that specifies the information to be extracted from the document. The aforementioned processor, The input prompt including the aforementioned prompt statement is generated and input to the generated AI model. A prompt generation system that identifies additional information included in the extracted information extracted by the generating AI model in response to the input prompt that does not correspond to the information to be extracted, and adds a sentence to the prompt sentence that specifies new information to be extracted according to the additional information.

2. The aforementioned document contains item information for each item, corresponding to that item. The prompt generation system according to claim 1, wherein the memory stores a prompt statement that specifies item information corresponding to the target item as the extraction target information for each predetermined target item included in the item.

3. The memory stores the characteristics of each target item, The prompt generation system according to claim 2, wherein the processor causes the generating AI model to identify items that match the characteristics from the items in the document as target items, and generates the input prompt including a prompt sentence corresponding to the identified target items.

4. The memory stores determination conditions for determining the information to be extracted included in the extracted information, The prompt generation system according to claim 1, wherein the processor determines, based on the determination conditions, that information included in the extracted information that does not match the determination conditions is the additional information, and adds a new condition to the determination conditions for determining that the additional information is the information to be extracted.

5. The memory stores template information for a statement that specifies the new information to be extracted. The prompt generation system according to claim 1, wherein the processor uses the template information to add a statement to the prompt statement that specifies the new information to be extracted.

6. The memory stores, for each type of document, the characteristics of that type of document and the prompt statement. The prompt generation system according to claim 1, wherein the processor determines the type of document based on the characteristics of the document for the generated AI model, and generates the input prompt including the prompt statement corresponding to that type.

7. The prompt generation system according to claim 1, wherein the processor calculates the percentage of the target information to be extracted based on the plurality of extracted pieces of information extracted by the generating AI model from the plurality of documents, and if the percentage is less than a threshold, deletes the prompt statement specifying the target information to be extracted from the memory.

8. The prompt generation system according to claim 1, wherein the processor displays the extracted information.

9. A prompt management method using a prompt generation system that generates input prompts for a generative AI model to extract information from a document, The prompt generation system comprises a processor and memory, The memory stores a prompt statement that specifies the information to be extracted from the document. The aforementioned processor, The input prompt including the aforementioned prompt statement is generated and input to the generated AI model. A prompt management method comprising: determining additional information included in the extracted information extracted by the generating AI model in response to the input prompt that does not correspond to the information to be extracted; and adding a statement to the prompt statement that specifies new information to be extracted according to the additional information.

Images