Method and program for outputting causal relationships

Abstract

In addition to the validity of the causal relationship, the level of user interest in the causal relationship is taken into consideration when outputting the causal relationship. [Solution] The causal relationship output method is a computer-operated causal relationship output method that obtains information indicating the user's profile, extracts information regarding the causal relationships of the data from a storage unit that stores analysis data, inputs the information regarding each of the causal relationships into an analysis model to calculate a first index value indicating the validity of each of the causal relationships, inputs the information regarding each of the causal relationships and the information indicating the user's profile into the analysis model to calculate a second index value indicating the degree of interest the user has in each of the causal relationships, and outputs causal relationships in which the first index value is equal to or greater than a first threshold, while also outputting causal relationships in which the second index value is equal to or greater than a second threshold among causal relationships in which the first index value is less than the first threshold.

Description

Technical Field

[0001] The present invention relates to a causal relationship output method and a causal relationship output program.

Background Art

[0002] The use of strategy maps is increasing in order to obtain measures that help achieve the objectives of individual companies and organizations. A strategy map is a graph (causal graph) that includes strategic objectives (goals) classified into several perspectives, with causal relationships established between the goals. For example, when considering measures based on the analysis results of business data, it is necessary to understand and take into account the causal relationships between the data.

[0003] Since the number of causal relationships between data increases in proportion to the number of data items, if all causal relationships are displayed on a strategy map (causal graph), it becomes difficult for a person to check them, and there is a possibility of overlooking important parts. In response to this, conventionally, a technique is known that uses LLM (Large Language Models) to calculate the validity of causal relationships and display a strategy map that takes the validity into account (see Non-Patent Document 1, etc.).

Prior Art Documents

Non-Patent Documents

[0004]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, as described in Non-Patent Document 1 above, when a strategy map (causal graph) is displayed considering the validity of causal relationships, there is a possibility that the causal relationships that the user actually wants to emphasize may not be displayed as important.

[0006] In one aspect, the present invention aims to provide a causal relationship output method and a causal relationship output program that can output causal relationships while taking into account not only the validity of the causal relationship but also the user's level of interest in the causal relationship. [Means for solving the problem]

[0007] In one embodiment, the causal relationship output method is a computer-operated causal relationship output method in which the process is performed by a computer, which acquires information indicating the user's profile, extracts information regarding the causal relationships of the data from a storage unit that stores analysis data, inputs the information regarding each of the causal relationships into an analysis model to calculate a first index value indicating the validity of each of the causal relationships, inputs the information regarding each of the causal relationships and the information indicating the user's profile into the analysis model to calculate a second index value indicating the degree of interest the user has in each of the causal relationships, and outputs causal relationships in which the first index value is equal to or greater than a first threshold, while also outputting causal relationships in which the first index value is less than the first threshold but the second index value is equal to or greater than a second threshold. [Effects of the Invention]

[0008] In addition to the validity of the causal relationship, the level of user interest in the causal relationship can also be taken into consideration when outputting the causal relationship. [Brief explanation of the drawing]

[0009] [Figure 1] Figure 1 is a schematic diagram showing the configuration of the causal relationship output system according to the first embodiment. [Figure 2] Figure 2(a) shows the hardware configuration of the user terminal, and Figure 2(b) shows the hardware configuration of the server. [Figure 3]This is a functional block diagram of the user terminal and server. [Figure 4] Figure 4(a) shows an example of a causal relationship database, and Figure 4(b) shows the state in which the validity value has been stored in the calculation result database. [Figure 5] Figure 5(a) shows the state in which the interest level values ​​are stored in the calculation results DB, and Figure 5(b) shows an example of the extraction results DB. [Figure 6] Figure 6 shows an example of a causal model output in the first embodiment. [Figure 7] Figure 7 is a flowchart showing the server's processing. [Figure 8] Figures 8(a) and 8(b) are diagrams illustrating comparative examples. [Figure 9] Figure 9 is a diagram illustrating the effects and advantages of the first embodiment. [Figure 10] Figure 10 is a diagram illustrating the method for outputting a causal model according to a modified example of the first embodiment. [Figure 11] Figure 11 is a flowchart showing the server processing according to the second embodiment. [Figure 12] Figures 12(a) and 12(b) are diagrams illustrating the processing of the server (extraction unit) according to the second embodiment. [Modes for carrying out the invention]

[0010] 《First Embodiment》 The first embodiment of the causal relationship output system will be described in detail below with reference to Figures 1 to 10.

[0011] The causal relationship output system 100 of this first embodiment is a system that performs causal analysis between data when companies and organizations perform data analysis and provide consulting, and outputs (presents to the user) the causal relationships obtained from the causal analysis in a causal graph.

[0012] For example, in the case of financial and non-financial analysis, financial information and non-financial information are used to generate and output a causal graph that indicates how the non-financial information affects the financial information. Users utilize the output causal graph for management strategies and plans. The causal relationships included in the causal graph in this case include, for example, "Increase in the proportion of female executive employees" → "Improvement in engagement" → "Leads to an increase in operating profit".

[0013] Further, in the case of supply chain risk analysis, when a risk event (earthquake, regulation, supplier bankruptcy, etc.) occurs in the supply chain, a causal graph showing the causal relationships between the impacts on the supply chain (transport delay, production stoppage, etc.) and the risk event is generated and output. From such a causal graph, the scope of impact of the risk event can be grasped, so new insights can be provided to the user, and the user's consideration can be supported. Examples of causal relationships in this case include "Earthquake occurs" → "Road is damaged" → "Transportation route is blocked" → "Component shortage" → "Production stoppage".

[0014] Figure 1 schematically illustrates the configuration of a causal relationship output system 100 according to the first embodiment. As shown in Figure 1, the causal relationship output system 100 includes a user terminal 70 and a server 10. The user terminal 70 and the server 10 are connected to a network 80 such as the Internet.

[0015] The user terminal 70 is a PC (Personal Computer) or other terminal used by a user (for example, an employee belonging to the company's human resources department, or an employee in charge of measures to improve engagement or recruitment). Figure 2(a) shows the hardware configuration of the user terminal 70. As shown in Figure 2(a), the user terminal 70 includes a CPU (Central Processing Unit) 90, ROM (Read Only Memory) 92, RAM (Random Access Memory) 94, storage (SSD (Solid State Drive) or HDD (Hard Disk Drive)) 96, network interface 97, display unit 93, input unit 95, and a portable storage medium drive 99, etc. The display unit 93 includes an LCD display, etc., and the input unit 95 includes a keyboard, mouse, touch panel, etc. Each of these components of the user terminal 70 is connected to the bus 98. In server 10, the CPU 90 executes programs stored in ROM 92 or storage 96, or programs read from the portable storage medium 91 by the portable storage medium drive 99, thereby realizing the functions of each component as shown in Figure 3.

[0016] The server 10 is an information processing apparatus that analyzes causal relationships between data, and outputs the analyzed results to the user terminal 70 as a causal graph. Fig. 2(b) shows the hardware configuration of the server 10. As shown in Fig. 2(b), the server 10 is provided with a CPU (Central Processing Unit) 190, a ROM (Read Only Memory) 192, a RAM (Random Access Memory) 194, a storage (SSD (Solid State Drive) or HDD (Hard Disk Drive)) 196, a network interface 197, a portable storage medium drive 199, and the like. These respective components of the server 10 are connected to a bus 198. In the server 10, the CPU 190 executes a program (including a causal relationship output program) stored in the ROM 192 or the storage 196, or a program read by the portable storage medium drive 199 from the portable storage medium 191, thereby realizing the functions of the respective units shown in Fig. 3. Note that Fig. 3 also illustrates various tables stored in the storage 196 or the like of the server 10. The functions of the respective units in Fig. 3 may be realized by, for example, an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

[0017] Fig. 3 shows a functional block diagram of the user terminal 70 and the server 10. As shown in Fig. 3, in the user terminal 70, the CPU 90 executes a program to function as an information input unit 72, a causal analysis request unit 74, and a causal graph display unit 76. On the other hand, in the server 10, the CPU 190 executes a program to function as a causal analysis processing unit 12, a persona creation unit 14, a validity calculation unit 16, an interest level calculation unit 18, an extraction unit 20, and an output unit 22.

[0018] (Regarding each function of the user terminal 70) The information input unit 72 receives input from the user via the input unit 95, including analysis data that is the target of causal analysis and user attribute information data, and transmits it to the server 10. The analysis data is stored in the analysis data DB 30, which is a storage unit of the server 10. For example, if there is analysis data for the item "employee productivity," this analysis data includes data showing the time-series changes in employee productivity. Similarly, if there is analysis data for the item "operating profit," this analysis data includes data showing the time-series changes in operating profit. User attribute information data is stored in the attribute information DB 32.

[0019] The causal analysis request unit 74 requests the server 10 (causal analysis processing unit 12) to perform a causal analysis when the user requests the start of a causal analysis via the input unit 95.

[0020] The causal graph display unit 76 receives the causal graph output from the server 10 (see, for example, Figure 6) and displays it on the display unit 93.

[0021] (Regarding each function of Server 10) The causal analysis processing unit 12 searches for causal relationships in each item of data included in the analysis data DB 30, for example, using known causal search methods. Causal relationships include information indicating the causal relationship (cause and effect) and causal coefficients indicating the influence of the causal relationship. Examples of causal search methods include, but are not limited to, LiNGAM (Linear Non-Gaussian Acyclic Model), SAM (Structural Agnostic Modelling), Bayesian networks, and LiNG (Linear Non-Gaussian SEM). The information on causal relationships searched by the causal analysis processing unit 12 is stored in the causal relationship DB 36.

[0022] Figure 4(a) shows an example of the causal relationship DB36. As shown in Figure 4(a), the causal relationship DB36 stores the "cause" and "effect" of causal relationships in association with each other, as well as the "causal coefficient," which represents the influence of causality. The causal coefficient takes values ​​between -1.0 and 1.0. Positive values ​​indicate a positive relationship, negative values ​​indicate a negative relationship, and when the causal coefficient is close to 0, it means that there is little to no causal relationship between the items. The causal analysis processing unit 12 may, for example, acquire information on causal relationships analyzed by an external device (a device other than the server 10) capable of causal exploration.

[0023] The persona creation unit 14 generates information (personas) that represent the user's personality based on the user's attribute information stored in the attribute information DB 32. The persona creation unit 14 creates multiple commonly assumed personas by inputting the attribute information into an LLM (Large-Scale Language Model) as an analysis model. Personas include information (elements) such as age (age group), gender, and job title. Each persona created by the persona creation unit 14 for a given user is stored in the persona DB 34, associated with each user's identification information.

[0024] The validity calculation unit 16 inputs a causal relationship description and a question prompting the LLM to evaluate validity (first index value) as queries, and calculates the validity of the causal relationship by prompting the LLM to respond with a numerical value for validity. The validity is assumed to be a value in the range of 0 to 1.0, for example, with a higher value indicating higher validity. The validity calculation unit 16 stores the calculated validity of the causal relationship in the calculation result DB 38. Figure 4(b) shows an example of the calculation result DB 38. As shown in Figure 4(b), the calculation result DB 38 stores "cause," "effect," and "causal coefficient," similar to the causal relationship DB 36 (Figure 4(a)). In addition, the calculation result DB 38 stores values ​​for "validity" and "level of interest." The validity value calculated by the validity calculation unit 16 is stored in the "validity" column in Figure 4(b).

[0025] The interest level calculation unit 18 uses LLM to calculate the user's level of interest (second indicator value) for each causal relationship. Here, the interest level calculation unit 18 calculates the user's level of interest for a certain causal relationship (referred to as the "first causal relationship"). In this case, the interest level calculation unit 18 inputs the following as queries into the LLM: a description of the first causal relationship, the user's persona, and a question that asks the user to evaluate values ​​indicating the relationship between the first causal relationship and each piece of information (element) included in the persona. The interest level calculation unit 18 then obtains a response from the LLM indicating the relationship between each piece of information (element) included in the persona and the first causal relationship. In this case, the interest level calculation unit 18 calculates the average value of the obtained relationship values ​​and takes the calculation result as the user's level of interest for the first causal relationship. The level of interest is, for example, a value in the range of 0 to 1.0, and a larger value indicates a greater level of user interest in that causal relationship. The interest level calculation unit 18 is stored in the "Interest Level" column of the calculation result DB 38, as shown in Figure 5(a).

[0026] The extraction unit 20 refers to the calculation result DB 38 in Figure 5(a) and extracts causal relationships where the validity value is equal to or greater than the first threshold, as well as causal relationships where the level of interest value is equal to or greater than the second threshold. Here, as an example, the first threshold = 0.5 and the second threshold = 0.5. The first and second thresholds can be set as appropriate by the user or administrator. The information on causal relationships extracted by the extraction unit 20 is stored in the extraction result DB 40 (Figure 5(b)). As shown in Figure 5(b), the extraction result DB 40 stores only the data from the calculation result data included in the calculation result DB 38 in Figure 5(a) where the validity value is equal to or greater than the first threshold, or the level of interest value is equal to or greater than the second threshold.

[0027] The output unit 22 generates a causal graph based on the data stored in the extraction results DB 40 and outputs it to the user terminal 70 (causal graph display unit 76). For example, if the extraction results DB 40 contains data like that shown in Figure 5(b), the output unit 22 generates a causal graph like the one shown in Figure 6 based on the cause and effect information stored in Figure 5(b) and outputs it to the user terminal 70.

[0028] (Regarding the processing on Server 10) Next, we will explain in detail the processing of server 10, following the flowchart in Figure 7.

[0029] As a prerequisite for the process shown in Figure 7 to begin, the analysis data DB 30 is assumed to contain analysis data entered from the information input unit 72 of the user terminal 70. The attribute information DB 32 is assumed to contain user attribute information data entered from the information input unit 72 of the user terminal 70. Furthermore, the process shown in Figure 7 begins when a request for causal analysis is received from the causal analysis request unit 74 of the user terminal 70. When a request for causal analysis is received from the causal analysis request unit 74, the user identification information of the user using the user terminal 70 is also sent to the server 10.

[0030] When the process shown in Figure 7 begins, in step S12, the causal analysis processing unit 12 first searches for causal relationships in each item of the analysis data contained in the analysis data DB 30. As a result, data like that shown in Figure 4(a) is stored in the causal relationship DB 36.

[0031] Next, in step S14, the persona creation unit 14 generates a persona for the user using the user terminal 70 based on the user's attribute information stored in the attribute information DB 32. For example, if the attribute information of the user using the user terminal 70 is name "〇no ×ko", gender "female", age "32", and department "Human Resources Department", then a persona such as "female, 30-35 years old, Human Resources Department" will be generated. The personas for each user created by the persona creation unit 14 are stored in the persona DB 34, associated with the user's identification information.

[0032] Next, in step S16, the validity calculation unit 16 calculates the validity of the causal relationship using LLM and stores the calculated validity value in the calculation result DB 38.

[0033] Next, in step S18, the interest level calculation unit 18 uses LLM to calculate the user's level of interest in the causal relationship and stores the calculated interest level value in the calculation result DB 38.

[0034] Next, in step S20, the extraction unit 20 selects unselected causal relationship data from the calculation result DB38. Here, as an example, it is assumed that the data in the first row of Figure 5(a) (cause "employee productivity", result "operating profit") is selected.

[0035] Next, in step S22, the extraction unit 20 determines whether the validity of the selected data is equal to or greater than the first threshold. If the first threshold is 0.5, the determination here is affirmed, and the extraction unit 20 proceeds to step S26. In step S26, the extraction unit 20 stores the selected causal relationship data in the extraction result DB 40 (see the first row of Figure 5(b)).

[0036] After step S26, in step S28, the extraction unit 20 determines whether all the data in the calculation result DB 40 has been selected. If the determination in step S28 is negative, the process returns to step S20, where the extraction unit 20 selects the unselected causal relationship data from the calculation result DB 38. Here, as an example, it is assumed that the data in the second row of Figure 5(a) (cause "GHG emissions", result "operating profit") has been selected.

[0037] Next, in step S22, the extraction unit 20 determines whether the validity of the selected data is equal to or greater than the first threshold. Here, the validity is 0.2, which is less than the first threshold (=0.5), so the determination in step S22 is rejected, and the extraction unit 20 proceeds to step S24.

[0038] When the system moves to step S24, the extraction unit 20 determines whether the level of interest of the selected data is equal to or greater than the second threshold. For example, if the second threshold is 0.5, the level of interest is 0.4, so the determination in step S24 is rejected, and the extraction unit 20 moves to step S28. When the system moves to step S28, the extraction unit 20 determines whether all the data in the calculation result DB 40 has been selected. If this determination in step S26 is rejected, the system returns to step S20, and the extraction unit 20 selects the unselected causal relationship data (in this case, the data in the third row of Figure 5(a)) from the calculation result DB 38.

[0039] Next, in step S22, the extraction unit 20 determines whether the validity of the selected data is equal to or greater than the first threshold. Here, the validity is 0.3, which is less than the first threshold (=0.5), so the determination in step S22 is rejected, and the extraction unit 20 proceeds to step S24. In step S24, the extraction unit 20 determines whether the degree of interest of the selected data is equal to or greater than the second threshold. Here, the degree of interest is 0.9, which is greater than the second threshold (=0.5), so the determination in step S24 is affirmed, and the extraction unit 20 proceeds to step S26. In step S26, the extraction unit 20 stores the selected causal relationship data in the extraction result DB 40 (see the second row of Figure 5(b)). Subsequently, in step S28, the extraction unit 20 determines whether all the data in the calculation result DB 40 has been selected.

[0040] Subsequently, the extraction unit 20 repeats the processing and decision-making in steps S20 to S28, and proceeds to step S30 when the decision in step S28 is confirmed, that is, when processing of all data in the calculation result DB38 is completed.

[0041] When the process moves to step S30, the output unit 22 generates a causal graph based on the data stored in the extraction results DB 40 and outputs it to the user terminal 70 (causal graph display unit 76). If the extraction results DB 40 contains data as shown in Figure 5(b) when the process moves to step S30, the output unit 22 generates a causal graph as shown in Figure 6 and outputs it to the user terminal 70. As a result, the causal graph shown in Figure 6 is displayed on the display unit 93 of the user terminal 70.

[0042] With this, all the processes shown in Figure 7 are completed.

[0043] Figure 8(a) shows the extracted results DB40 when the extraction unit 20 extracts causal data to be displayed in the causal graph based solely on validity. In other words, the extracted results DB40 in Figure 8(a) contains causal data from the data stored in the calculation results DB38 whose validity is equal to or greater than the first threshold (0.5).

[0044] Figure 8(b) shows a causal graph generated based on the extracted results DB40 in Figure 8(a). As can be seen by comparing Figure 8(b) with Figure 6, Figure 6 displays information not shown in Figure 8(b) (causal data that is of low validity but of high user interest). Specifically, Figure 6 also displays the causal relationship between the cause "percentage of female executives" and the result "engagement," as well as the causal relationship between the cause "engagement" and the result "operating profit." As a result, in this first embodiment, it is possible to extract all causal relationships that are of low validity but of importance to users from among the large number of existing causal relationships and display them in a causal graph.

[0045] As described in detail above, according to this first embodiment, the persona creation unit 14 creates a persona of a user using the user terminal 70 (S14), and the causal analysis processing unit 12 extracts causal relationships between items from the storage unit that stores the analysis data DB 30 (S12). The validity calculation unit 16 calculates the validity of the causal relationships by inputting the causal relationships into the LLM (S16), and the interest level calculation unit 18 calculates the user's interest level in the causal relationships by inputting the causal relationships and the user's persona into the LLM (S18). The extraction unit 20 extracts causal relationships whose validity is above the first threshold and causal relationships whose validity is below the first threshold but whose interest level is above the second threshold (S20-S28). The output unit 22 outputs causal relationships whose validity is above the first threshold, as well as causal relationships whose validity is below the first threshold but whose interest level is above the second threshold. Here, Figure 9 shows a graph with validity on the vertical axis and interest level on the horizontal axis, with the range divided using a first threshold and a second threshold. Conventionally, since only validity was considered, only causal relationships with "high validity" (1) and (2) were output to the causal graph. In contrast, in this first embodiment, in addition to (1) and (2), (4) "low validity but high interest" is also output to the causal graph. As a result, in this embodiment, it is possible to extract and output all causal relationships that have low validity but are important to the user from the large amount of causal relationships extracted from the analysis data DB30. In this way, users can analyze measures and other actions in companies and organizations by considering not only causal relationships with high validity but also causal relationships that are important to the user.

[0046] Furthermore, in this first embodiment, the validity of causal relationships and the user's level of interest in those causal relationships are calculated using an LLM (Large-Scale Language Model). This allows for accurate and rapid calculation of validity and level of interest. Note that the method for calculating validity and level of interest is not limited to the above; other analytical models and methods may also be used.

[0047] Furthermore, in this first embodiment, the persona creation unit 14 generates a user persona based on attribute information entered by the user, so the user persona can be automatically and easily generated.

[0048] Furthermore, in this first embodiment, when calculating the level of interest in a certain causal relationship, the interest level calculation unit 18 evaluates the relationship between that causal relationship and each of the multiple pieces of information included in the persona using LLM. The interest level calculation unit 18 then uses the average of the multiple relationship values ​​obtained through the evaluation as the user's level of interest in that causal relationship. This makes it possible to accurately determine the user's level of interest in the causal relationship.

[0049] In the first embodiment described above, as shown in Figure 6, a case was described in which causal relationships with high validity and causal relationships of high user interest are not distinguished, but the invention is not limited to this. For example, as shown in Figure 10, arrows indicating causal relationships with high validity and arrows indicating causal relationships of high user interest may be displayed in different ways (arrow color, line type, line thickness, etc., or combinations thereof). In Figure 10, arrows indicating causal relationships with high validity are solid lines, and arrows indicating causal relationships of high user interest are dashed lines.

[0050] In Figures 6 and 10, causal relationships are shown only with arrows, but this is not the only option. Causal coefficients, validity values, and interest values ​​may also be displayed near the arrows. Furthermore, the display style of the arrows (e.g., thickness) may be varied based on the causal coefficients, validity values, and interest values.

[0051] In the first embodiment described above, when displaying the causal relationships of (1), (2), and (4) in Figure 9 in a causal graph, the display patterns of the arrows indicating each of the causal relationships of (1), (2), and (4) may be different.

[0052] In the above embodiment, the causal relationships of (1), (2), and (4) in Figure 9 are shown in the causal graph. However, it is also possible to show only the causal relationships of (1) and (4) in Figure 9 in the causal graph (without showing the causal relationship of (2)).

[0053] 《Second Embodiment》 The second embodiment will now be described. In this second embodiment, the processing content of the extraction unit 20 differs from that of the first embodiment, so this point will be the main focus of the explanation.

[0054] Figure 11 shows a flowchart illustrating the processing of the server 10 according to the second embodiment. The processing indicated by the thick outline in Figure 11 (processing of the extraction unit 20) differs from that of the first embodiment (Figure 7).

[0055] Once the process shown in Figure 11 is started, steps S12 to S18 are carried out in the same manner as in the first embodiment. That is, it is assumed that the calculation result DB38 shown in Figure 5(a) is obtained when the process in step S18 is completed.

[0056] Next, in step S120, the extraction unit 20 plots each data from the calculation result DB38 on a two-axis graph with validity and interest. Specifically, as shown in Figure 12(a), each data stored in the calculation result DB38 is plotted on a two-axis graph with validity on the vertical axis and interest on the horizontal axis.

[0057] Next, in step S122, the extraction unit 20 calculates statistically derived values ​​(mean and median) from the validity values. For example, the extraction unit 20 calculates the mean of the validity values ​​for each plotted causal relationship.

[0058] Next, in step S124, the extraction unit 20 calculates statistically derived values ​​(mean and median) from the interest level values. For example, the extraction unit 20 calculates the mean of the interest level values ​​for each plotted causal relationship.

[0059] Next, in step S126, the extraction unit 20 divides the data into four regions using statistically obtained values ​​(e.g., mean values) for validity and level of interest. In Figure 12(b), the mean values ​​are shown by dashed lines, and the data is divided into four regions (1) to (4), similar to Figure 9.

[0060] Next, in step S128, the extraction unit 20 extracts the causal relationships included in the four regions (1), (2), and (4). In the example in Figure 12(b), the extraction unit 20 extracts the causal relationships indicated by black circles and stores them in the extraction result DB 40, while not extracting the causal relationships indicated by white circles.

[0061] Subsequently, when the process moves to step S30, the output unit 22 generates a causal graph based on the data stored in the extracted results DB 40 and outputs it to the user terminal 70 (causal graph display unit 76).

[0062] With this, all the processes shown in Figure 11 are completed.

[0063] In this second embodiment, the first threshold is a value statistically obtained from the validity values ​​of each causal relationship (mean or median), and the second threshold is a value statistically obtained from the degree of interest values ​​of each causal relationship (mean or median). The statistically obtained values ​​do not have to be the mean or median; other values ​​(for example, the mean of the minimum and maximum values) may also be used.

[0064] As explained above, in this second embodiment, since the first and second thresholds are statistically obtained values ​​from actual validity and interest values, it becomes possible to set appropriate thresholds based on actual validity and interest values. This makes it possible to appropriately extract causal relationships to be included in the causal graph.

[0065] In the second embodiment described above, the causal relationships included in areas (1), (2), and (4) of Figure 12(b) are displayed in the causal graph, but this is not the only option. For example, only the causal relationships included in areas (1) and (4) of Figure 12(b) may be displayed in the causal graph.

[0066] The above processing functions can be implemented by a computer. In this case, a program describing the processing content of the functions that the processing unit should have is provided. By executing this program on a computer, the above processing functions are implemented on the computer. The program describing the processing content can be recorded on a storage medium that can be read by a computer (except for carrier waves).

[0067] When distributing a program, it may be sold in the form of a portable storage medium such as a DVD (Digital Versatile Disc) or CD-ROM (Compact Disc Read Only Memory) on which the program is recorded. Alternatively, the program can be stored in the storage device of a server computer and transferred from the server computer to other computers via a network.

[0068] A computer executing a program stores programs, for example, those recorded on a portable storage medium or transferred from a server computer, in its own memory. The computer then reads the program from its memory and executes the processing according to the program. Alternatively, the computer can directly read the program from the portable storage medium and execute the processing according to that program. Furthermore, the computer can sequentially execute the processing according to the programs received as they are transferred from the server computer.

[0069] The embodiments described above are preferred examples of the present invention. However, the invention is not limited thereto, and various modifications are possible without departing from the spirit of the invention.

[0070] Furthermore, the following additional information is disclosed regarding the above description of the embodiments. (Note 1) Obtain information that shows the user's profile, Extract information about the causal relationships of the data from the storage unit that stores the analysis data. By inputting information regarding each of the aforementioned causal relationships into the analysis model, a first indicator value showing the validity of each of the aforementioned causal relationships is calculated. By inputting information regarding each of the aforementioned causal relationships and information representing the user's profile into the analysis model, a second indicator value is calculated that shows the degree of interest the user has in each of the aforementioned causal relationships. When outputting causal relationships in which the first indicator value is equal to or greater than the first threshold, causal relationships in which the second indicator value is equal to or greater than the second threshold, among those in which the first indicator value is less than the first threshold, are also output. A method for outputting causal relationships, characterized in that the processing is performed by a computer. (Note 2) The causal relationship output method according to Note 1, characterized in that the analysis model is a large-scale language model. (Note 3) The first threshold is a value statistically obtained from the first index value of each of the causal relationships. The causal relationship output method according to Appendix 1 or 2, characterized in that the second threshold is a value statistically obtained from the second index value of each of the causal relationships. (Note 4) The causal relationship output method according to Note 3, characterized in that the statistically obtained value is the mean or median. (Note 5) A causal relationship output method according to any one of Notes 1 to 4, characterized in that it generates information indicating the user's profile based on attribute information entered by the user. (Note 6) The causal relationship output method according to any one of Notes 1 to 5, characterized in that the method for outputting causal relationships in which the first indicator value is equal to or greater than the first threshold is different from the method for outputting causal relationships in which the second indicator value is equal to or greater than the second threshold among causal relationships in which the first indicator value is less than the first threshold. (Appendix 7) The causal relationship output method according to any one of Appendix 1 to 6, characterized in that, in the process of calculating the first indicator value, when calculating the first indicator value of the first causal relationship, an explanatory text of the first causal relationship and a question for evaluating the validity of the first causal relationship are input to the analysis model. (Appendix 8) The causal relationship output method according to any one of Appendix 1 to 7, characterized in that, in the process of calculating the second index value, when calculating the second index value of the first causal relationship, a question sentence for evaluating the relationship between the first causal relationship and each of the multiple elements included in the information representing the user's profile is input to the analysis model, and the average value of the values ​​indicating the relationship between each of the multiple elements and the first causal relationship is set as the second index value of the first causal relationship. (Note 9) Obtain information that shows the user's profile, Extract information about the causal relationships of the data from the storage unit that stores the analysis data. By inputting information regarding each of the aforementioned causal relationships into the analysis model, a first indicator value showing the validity of each of the aforementioned causal relationships is calculated. By inputting information regarding each of the aforementioned causal relationships and information representing the user's profile into the analysis model, a second indicator value is calculated that shows the degree of interest the user has in each of the aforementioned causal relationships. When outputting causal relationships in which the first indicator value is equal to or greater than the first threshold, causal relationships in which the second indicator value is equal to or greater than the second threshold, among those in which the first indicator value is less than the first threshold, are also output. A causal relationship output program characterized by having a computer perform the processing. [Explanation of symbols]

[0071] 10 servers 12. Causal Analysis Processing Unit 14. Persona Creation Department 16. Validity Calculation Unit 18. Interest Level Calculation Unit 20 Extraction part 22 Output section 70 User terminals 100 Causal Relationship Output System 190 CPU (Computer)

Claims

1. We obtain information that shows the user's profile, Extract information about the causal relationships of the data from the storage unit that stores the analysis data. By inputting information regarding each of the aforementioned causal relationships into the analysis model, a first indicator value showing the validity of each of the aforementioned causal relationships is calculated. By inputting information regarding each of the aforementioned causal relationships and information representing the user's profile into the analysis model, a second index value indicating the degree of interest the user has in each of the aforementioned causal relationships is calculated. When outputting causal relationships in which the first indicator value is equal to or greater than the first threshold, causal relationships in which the second indicator value is equal to or greater than the second threshold, among those in which the first indicator value is less than the first threshold, are also output. A method for outputting causal relationships, characterized in that the processing is performed by a computer.

2. The causal relationship output method according to claim 1, characterized in that the analysis model is a large-scale language model.

3. The first threshold is a value statistically obtained from the first index values ​​of each of the causal relationships, The causal relationship output method according to claim 1, characterized in that the second threshold is a value statistically obtained from the second index value of each of the causal relationships.

4. The causal relationship output method according to claim 3, characterized in that the statistically obtained value is the mean or median.

5. The causal relationship output method according to claim 1, characterized in that it generates information representing the user's profile based on attribute information entered by the user.

6. The causal relationship output method according to claim 1, characterized in that there is a mode in which a causal relationship in which the first indicator value is equal to or greater than a first threshold is output, and a mode in which a causal relationship in which the first indicator value is less than the first threshold is output in which the second indicator value is equal to or greater than a second threshold.

7. We obtain information that shows the user's profile, Extract information about the causal relationships of the data from the storage unit that stores the analysis data. By inputting information regarding each of the aforementioned causal relationships into the analysis model, a first indicator value showing the validity of each of the aforementioned causal relationships is calculated. By inputting information regarding each of the aforementioned causal relationships and information representing the user's profile into the analysis model, a second index value indicating the degree of interest the user has in each of the aforementioned causal relationships is calculated. When outputting causal relationships in which the first indicator value is equal to or greater than the first threshold, causal relationships in which the second indicator value is equal to or greater than the second threshold, among those in which the first indicator value is less than the first threshold, are also output. A causal relationship output program characterized by having a computer perform the processing.

Images