Estimation device and estimation method

Abstract

An estimation device (10) has a question generation unit (133), a requirement extraction unit (134), and a presentation unit (135). The question generation unit (133) generates, on the basis of a message of a question input by a first user and a chat log and attribute information of the first user, a refined question of the question. The requirement extraction unit (134) extracts, from the question generated by the question generation unit (133), requirements to be satisfied by a user serving as a question destination. The presentation unit (135) presents, on the basis of chat logs and attribute information of the plurality of users, a user satisfying the requirements as the question destination among the plurality of users.

Description

Estimation device and estimation method

[0001] The present invention relates to an estimation device and an estimation method.

[0002] Conventionally, technologies have been developed to make tacit knowledge explicit and to improve searchability so that necessary knowledge can be easily found.

[0003] Examples of such technologies include cross-document searches of documents scattered throughout a company and the accumulation of FAQs (Frequently Asked Questions). By collecting and accumulating various knowledge and further improving searchability, even those without business knowledge can access the necessary information with minimal effort. Furthermore, in recent years, interactive document search AI that utilizes large-scale language models known as LLMs (Large Language Models) has emerged (see, for example, Non-Patent Document 1).

[0004] There is also technology that estimates the knowledge (keyword centrality) and human relationships (intimacy network) held by chat speakers from chat data, etc., and recommends close people who are connected to the desired knowledge (see, for example, non-patent document 2).

[0005] Llamaindex, [online], [Retrieved July 29, 2024], Internet, <URL: https: / / www.llamaindex.ai / > Pham Thanh Quang, Junichi Yamamoto, Study on Knowledge Management Systems Considering Human Relationships, 7F-02, 82nd National Convention of the Information Processing Society of Japan, 2020, <URL: https: / / ipsj.ixsq.nii.ac.jp / ej / ?action=pages_view_main&active_action=repository_view_main_item_detail&item_id=205763&item_no=1&page_id=13&block_id=8>

[0006] However, the conventional techniques have a problem in that they are unable to accurately introduce respondents to questions.

[0007] If automatic prediction of who to ask a question to only takes into account the respondent's proficiency, willingness to answer, and relationship (department, conversation history), there is a risk that the person will be connected to someone at a different level than the person desired.

[0008] Note that connecting means introducing the questioner to an appropriate person to answer the question (answer). For example, the system can connect the questioner and the answerer by outputting information for identifying an appropriate person to answer the question.

[0009] When a question is asked about a certain topic, it may be difficult to give an appropriate answer if the background to the question is unknown. For example, when someone asks for more information about a certain technology, a company executive may be looking for information about trends and future prospects in the world, a researcher may be looking for information at the level of a paper, and a developer may be looking for information on whether the technology meets development requirements. When these are conversations between people, both parties share the same background, so misunderstandings are unlikely to occur even if they are not explicitly stated.

[0010] Additionally, depending on the question, a complex range of knowledge may be required. For example, a question about whether a certain technology can be incorporated into development involves not only technical feasibility but also various other aspects such as cost, availability, and maintainability. Generally, when such complex requirements are met, multiple members come together to discuss the matter. In this case, simply introducing one member may only be able to meet part of the requirements.

[0011] Conventional techniques are unable to introduce respondents who share backgrounds and take into consideration the need for complex knowledge.

[0012] For example, the system can introduce a suitable answerer to the questioner by outputting information to identify the appropriate person to answer the question.

[0013] In order to solve the above-mentioned problems, the estimation device of the present invention is characterized by having: a question generation unit that generates a question that details the question input by a first user, based on the question message, the chat log of the first user, and attribute information of the first user; a requirements extraction unit that extracts requirements required of the user to whom the question is being asked, from the question generated by the question generation unit; and a presentation unit that presents users who satisfy the requirements from among the plurality of users as users to whom the question should be asked, based on the chat logs and attribute information of the plurality of users.

[0014] According to the present invention, it is possible to introduce an answerer to a question with high accuracy.

[0015] FIG. 1 is a diagram illustrating the difference between the level desired by a questioner and the level of a question recipient. FIG. 2 is a diagram illustrating an outline of the processing of an estimation device. FIG. 3 is a diagram illustrating an example configuration of an estimation device. FIG. 4 is a diagram illustrating processing for generating a question. FIG. 5 is a diagram illustrating processing for extracting requirements. FIG. 6 is a diagram illustrating processing for presenting a question recipient. FIG. 7 is a diagram illustrating processing for generating a question using a chat log. FIG. 8 is a diagram illustrating an example of a dialogue. FIG. 9 is a diagram illustrating matching processing based on weights. FIG. 10 is a flowchart illustrating an example of a procedure for updating attribute information executed by an estimation device. FIG. 11 is a flowchart illustrating an example of a procedure for presenting information executed by an estimation device. FIG. 12 is a diagram illustrating a computer that executes a program.

[0016] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a description will be given of an embodiment of the present invention with reference to the drawings, but the present invention is not limited to the embodiment.

[0017] First, an outline of the estimating device of this embodiment will be described. The estimating device estimates appropriate question recipients in response to a question from a questioner and presents the estimation results (candidates for question recipients).

[0018] For example, for a user who has information they want that is not documented (tacit knowledge) and does not know who to ask when they are in trouble (tacit knowledge of the knowledge owner), the estimation device estimates and presents to them who they should ask to find an answer.

[0019] In the following description, the estimation device will be described taking as an example a case where it estimates the destination of a user's question based on a post in a chat system, but estimation may also be performed based on a post other than a chat system. Note that a post refers to the input of a message to a chat system or the like.

[0020] If the estimation device only considers the proficiency level, proactiveness in answering, and relationships (department and conversation history) of the asker when estimating who to ask, the device may connect the asker with someone who does not meet the level of the asker's needs. This is due to a lack of understanding of the asker's job title, industry, question background, etc.

[0021] Figure 1 is a diagram that explains the difference between the level desired by the questioner and the level of the person asking the question. As shown in Figure 1, UserA, UserB, and UserC are potential candidates for the person asking a question about LLM. UserA is an evangelist and advisor to the sales department. While not an expert on machine learning, he is somewhat knowledgeable and often gives lectures related to LLM. UserB is a researcher and machine learning professional who loves and is knowledgeable about LLM. UserC is a development manager who has experience developing using LLM, but does not particularly like LLM.

[0022] Consider a case where a salesperson asks the question "I'm looking for someone who can explain about LLM" with the intention of finding someone knowledgeable about LLM to sell to. In this case, User A is not an appropriate person to ask because he / she is at a higher level than the salesperson is looking for.

[0023] Also, consider the case where a developer asks the question "I would like to develop with LLM" with the intention of knowing whether LLM meets the following development requirements. In this case, User B does not know whether LLM can really be used for development, and is not an appropriate person to ask.

[0024] Therefore, the estimation device introduces a question destination that is appropriate for the level desired by the asker through a process including two steps, question generation (step S1) and requirement extraction (step S2), as shown in Fig. 2. Fig. 2 is a diagram for explaining an outline of the process of the estimation device.

[0025] The estimation device estimates each poster's skills (including, for example, knowledge, ability, interest, etc.), attributes, and level of motivation to answer questions from the chat logs in a chat log DB (database), and stores the estimation results in the attribute DB as attribute information of the poster (user). The information in the chat log DB and the attribute DB is used for question generation and requirement extraction.

[0026] As shown in FIG. 2, assume that UserX, the asker, inputs the question "I'd like to know more about LLM." The estimation device infers the details of the question in question generation (step S1). For example, the estimation device concretizes the input question by having a bot interact with the asker, and by searching the asker's past logs (chat logs) and attribute information. For example, from the question "I'd like to know more about LLM" input by the asker, the estimation device generates the question "Question: I'd like to implement a summary function in LLM. I'd like to gather information on whether the technical feasibility and cost / accuracy requirements can be met in the next development."

[0027] Furthermore, in the requirement extraction (step S2) from the generated question, the estimation device extracts two conditions: (1) knowledge of the technical details of LLM, and (2) operational knowledge of the cost and accuracy of LLM. The estimation device identifies a person to whom the question should be connected based on the degree of match of each extracted condition. At this time, the estimation device may refer to the experience and attributes of users of the candidate question recipients. For example, a requirement may be rephrased as a condition of one poster attribute or a combination of conditions of multiple poster attributes.

[0028] For example, User B's degree of match with condition (1) is 90%, and User C's degree of match with condition (2) is 80%, both of which are the highest values ​​for the degree of match for each condition, so the estimation device introduces User B and User C as people to ask questions to.

[0029] [Configuration Example] An example configuration of the estimation device 10 will be described with reference to Fig. 3. The estimation device 10 includes, for example, an input / output unit 11, a storage unit 12, and a control unit 13.

[0030] The input / output unit 11 is an interface that controls the input and output of various data. The input / output unit 11 receives, for example, a question input from a user. The input / output unit 11 also outputs an estimation result of who the user should ask the question to (candidates of appropriate people to whom the user should ask the question).

[0031] The storage unit 12 stores data, programs, etc. that are referenced when the control unit 13 executes various processes. The storage unit 12 is realized by a semiconductor memory element such as a random access memory (RAM) or a flash memory, or a storage device such as a hard disk or an optical disk.

[0032] For example, the storage unit 12 includes a chat log DB (posted information storage unit) 121. The storage unit 12 also includes an attribute DB (attribute information storage unit) 122 that stores attribute information of each user created by the control unit 13. The storage unit 12 also includes a work experience DB 123 that stores information related to the work experience of each user.

[0033] The control unit 13 is responsible for overall control of the estimation device 10. The functions of the control unit 13 are realized, for example, by a central processing unit (CPU) executing a program stored in the storage unit 12.

[0034] The control unit 13 includes, for example, an attribute analysis unit 131 , an input reception unit 132 , a question generation unit 133 , a requirement extraction unit 134 , and a presentation unit 135 .

[0035] The attribute analysis unit 131 analyzes the skills and attributes of each user from the posts of each user and stores the analysis results as attribute information for each user in the attribute DB 122. For example, the attribute analysis unit 131 creates attribute information for each user from the posts of each user in the chat log DB 121 and stores the information in the attribute DB 122.

[0036] For example, the attribute analysis unit 131 extracts major keywords from the content of chat posts in the chat log DB 121, classifies the nature of the post (whether it is a question or a general post) and the topic, and analyzes the relationship between the keywords and the poster.

[0037] Then, through the above analysis, the attribute analysis unit 131 obtains analysis results such as the nature of the chat log post, keywords, topics, technical level, questioner, answerer, the answerer's proactiveness, whether or not the problem was ultimately successfully solved (result), etc. Then, the attribute analysis unit 131 creates attribute information of the poster from the analysis results of the post content and stores it in the attribute DB 122.

[0038] For example, the estimation device 10 may convert each piece of text in the chat system into a multidimensional vector in advance and store the vector in the chat log DB 121. For example, a Generative Pretrained Transformer (GPT), a Bidirectional Encoder Representations from Transformers (BERT), or the like may be used to vectorize the text.

[0039] The semantic similarity of the vectorized text can be evaluated based on the vector distance, which allows the estimation device to search the chat log DB 121 for text that is semantically similar to the question.

[0040] The input accepting unit 132 accepts input of a question from a user. The question generating unit 133 generates a question by refining the question input by the user. The requirement extracting unit 134 extracts requirements from the specific question. The presenting unit 135 presents users who match the requirements as recipients of the question.

[0041] 4 is a diagram illustrating the process of generating a question. First, the question generation unit 133 refines the user's question (step S101). That is, the question generation unit 133 vectorizes the question input by the user. Then, the question generation unit 133 searches the chat log DB 121 for vectors similar to the vector of the question input by the user. The question generation unit 133 refines the question input by the user based on text corresponding to the vector obtained by the search.

[0042] For example, based on chat log 1 similar to "I want to know more about LLM" input by the questioner (UserX), the question generator 133 refines it to "I want to implement a summary function in LLM, but I don't know if it will work, so I would like someone who is knowledgeable about it to help me." For example, the question generator 133 refines the question by combining chat log 1 obtained by the search using LLM.

[0043] Next, the question generator 133 determines an ambiguous portion of the detailed question (step S102). For example, in a question such as "I want to implement a summary function in LLM, but I'm not sure if it will work, so I'd like someone knowledgeable to tell me," the question generator 133 determines the part "I don't know if it will work" as an ambiguous portion. For example, the question generator 133 inputs the detailed question into LLM to determine the ambiguous portion.

[0044] Furthermore, the question generation unit 133 performs question concretization (step S103). The question generation unit 133 concretizes the ambiguous part. The question generation unit 133 concretizes the ambiguous part based on a dialogue between the Bot and the questioner. For example, the question generation unit 133 causes the LLM to generate a question for concretizing the ambiguous part in advance.

[0045] For example, the question generation unit 133 causes the Bot to send a message to UserX asking, "What do you prioritize in development?" If UserX answers, "Mainly cost, accuracy, and feasibility," the question generation unit 133 clarifies the question as, "I want to implement a summary function in LLM. I want to collect information on whether the technical feasibility and cost / accuracy requirements can be met in the next development."

[0046] FIG. 5 is a diagram illustrating the process of extracting requirements. The requirements extraction unit 134 extracts requirements from a specific question. Specifically, the requirements extraction unit 134 extracts conditions for the question recipient that satisfy the requirements of the question. For example, the requirements extraction unit 134 inputs the question into the LLM to extract conditions. Furthermore, the requirements extraction unit 134 may extract conditions based on domain knowledge (e.g., details of the system being developed) as needed.

[0047] For example, from the question, "I would like to implement a summary function using LLM. I would like to collect information on whether the technical feasibility and cost / accuracy requirements can be met in the next development," the requirements extraction unit 134 extracts two conditions: "(1) I know the technical details of LLM" and "(2) I have operational knowledge of the cost and accuracy of LLM."

[0048] 6 is a diagram illustrating the process of presenting a question recipient. The presenting unit 135 calculates the degree of match with the conditions of each user. For example, the presenting unit 135 calculates the degree of match based on each user's experience, personality, degree of support for others, and relationship with the questioner, which are stored in the attribute DB 122. Also, for example, the presenting unit 135 calculates the degree of match based on each user's past experience, which is stored in the work experience DB 123.

[0049] For example, the presentation unit 135 searches the work experience DB 123 for the keyword "technical details" in the condition "(1) know the technical details of LLM." Here, the work experience DB 123 stores that User B is a "researcher." As a result of the search, User B, who corresponds to the keyword "researcher," which is semantically close to "technical details," is matched. The presentation unit 135 presents User B as a question recipient along with the degree of match with the condition. Note that FIG. 6 shows that User B's degree of match with condition (1) is 72%. The degree of match is the similarity between keywords.

[0050] Furthermore, for example, the presentation unit 135 searches the work experience DB 123 for the keyword "operation" in the condition "(2) has operational knowledge of LLM cost and accuracy." Here, the work experience DB 123 stores that User C is a "development manager." As a result of the search, User C, which corresponds to the keyword "development," which is semantically close to "operation," is found to match. The presentation unit 135 presents User C as the person to whom the question should be posed, along with the degree of match with the condition. Note that FIG. 6 shows that User C's degree of match with condition (2) is 80%. The degree of match is the similarity between the keywords.

[0051] For example, User B's degree of match with condition (1) is 72%, and User C's degree of match with condition (2) is 80%, both of which are the highest values ​​for the degree of match for each condition, so the estimation device introduces User B and User C as question recipients. That is, the presentation unit 135 presents information about User B and User C.

[0052] [Example 1] The estimation device 10 may generate a question by extracting a history of recent chat comments from a user that are similar to a question and analyzing the extracted history. Note that LLM may be used to identify the background, problem, and purpose, search the chat log DB 121, and specify ambiguous parts.

[0053] First, as shown in Fig. 7, the question generator 133 identifies the background, problem, and purpose from the problem (question) in question refinement (step S111). Fig. 7 is a diagram illustrating the question generation process using a chat log. For example, from a question such as "I want to know more about LLM," the question generator 133 identifies the background, "I want to implement a summary function using LLM in my next development," the problem, "Will it work?", and the purpose, "I want to ask someone who is knowledgeable about it."

[0054] The question generation unit 133 searches the chat log DB 121 for the text of the message entered by the questioner (UserX) to identify the background, problem, and purpose. The question generation unit 133 assigns a score to the text searched for on a thread-by-thread basis, and identifies the background, problem, and purpose based on the thread with the highest score. The score is higher for messages that have been entered less recently.

[0055] Furthermore, the question generator 133 determines an ambiguous part (step S112) and specifies the question (step S113). For example, the question generator 133 specifies the background, problem, and purpose based on the dialogue between the Bot and the questioner. For example, the question generator 133 specifies the question "Will it work?" as "Can the technical feasibility, cost, and accuracy requirements be met in the next development?"

[0056] The question generator 133 also repeatedly asks the questioner additional questions through dialogue with the Bot until there are no more ambiguous points, and supplements the information through dialogue.The question generator 133 further clarifies the question based on the supplemented information.

[0057] [Example 2] The estimation device 10 may perform the processes of generating questions, extracting requirements, and presenting a destination for questions through dialogue with a Bot. The Bot's utterances may be generated using an LLM. Note that dialogue here refers to text exchange in a chat format. Furthermore, utterance refers to inputting text or input text. Note that the utterances of the estimation device 10 are performed via the Bot.

[0058] First, as shown in FIG. 8, in response to a question from a questioner (UserX) saying "I'd like to know more about LLM," the question generator 133 makes the following utterance: "From past logs, I interpreted your question as 'I want to implement a summary function in LLM, but I'm not sure if it will work, so I'd like someone with more knowledge to tell me.' Please tell me your specific concerns." This utterance is intended to detail the question entered by the questioner and to obtain information necessary for concretization. FIG. 8 is a diagram showing an example of a dialogue.

[0059] Furthermore, in response to the questioner's utterance of "feasibility," the question generation unit 133 shows the process of concretization, such as, "I updated the question to 'I want to implement a summary function in LLM. I want to collect information about the feasibility in the next development.'" Furthermore, the presentation unit 135 presents the person to ask, such as, "In this case, it may be a good idea to ask User C, who has development experience. Are there any other concerns?", and makes a speech to elicit further requirements.

[0060] When the questioner utters "speed and cost," the presentation unit 135 updates the question to "I would like to implement a summary function in LLM. I would like to gather information on whether the technical feasibility and cost / accuracy requirements can be met in the next development." In this case, since user B is knowledgeable about LLM technology, it would be a good idea to ask both user C and user B." The presentation unit 135 presents a more specific question and suggests the person to whom the question should be sent after the change.

[0061] Third Embodiment The estimation device 10 may determine (match) a question recipient based on the weights calculated for each item of the requirements and the weights calculated for each user who is a candidate for question recipient.

[0062] 9, it is assumed that the requirements extraction unit 134 extracts two conditions, "(1) knowledge of the technical details of LLM" and "(2) operational knowledge of the cost and accuracy of LLM" using the method described above (step S131). Fig. 9 is a diagram illustrating the matching process based on weights.

[0063] Next, the presenting unit 135 extracts information related to the conditions from the chat log DB 121 (step S231), and assigns a weight to each item of the conditions (step S232).

[0064] For example, for the condition "(1) knowledge of LLM technical details," the presentation unit 135 assigns weights to each of the three items, "research," "development," and "operations." For example, the presentation unit 135 may assign weights according to the similarity between "technical details" included in the condition (1) and each of the item names, "research," "development," and "operations."

[0065] Furthermore, in the work experience DB 123, a weight is assigned to each item for each user. For example, in the work experience DB 123, a weight is assigned to each of the three items of "research," "development," and "operations" for each user. The weights may be assigned dynamically or statically. The weight assigned to each user is called an expertise score.

[0066] The presenting unit 135 compares the weights assigned to the conditions with the expertise scores and performs matching (step S233). For example, the presenting unit 135 calculates a matching score that increases as the distance between the vector of the weights assigned to the conditions and the vector of the expertise score decreases, and presents the results as questions to be queried in descending order of matching score.

[0067] 10 is a flowchart showing an example of an attribute information updating procedure executed by the estimation device. First, when the attribute analysis unit 131 of the estimation device 10 acquires a new chat post (S11), it extracts changes in the user's (poster's) knowledge and skills from the post (S12), and updates the user's attribute information in the attribute DB 122 based on the extracted changes (S13). The attribute analysis unit 131 executes the above process every time a new chat post is acquired or at predetermined intervals.

[0068] 11 is a flowchart showing an example of a presentation procedure executed by the estimation device 10. The input receiving unit 132 of the estimation device 10 receives a question input from a questioner (step S21).

[0069] The question generator 133 extracts the chat log and attribute data of the questioner (step S22). The question generator 133 analyzes the background of the question and refines the question (step S23). If the question is specific (Yes in step S24), the question generator 133 proceeds to step S26.

[0070] If the question is not specific enough (No in step S24), the question generator 133 asks the questioner a follow-up question (step S25). The question generator 133 further refines the question based on the answer to the follow-up question (step S23).

[0071] The requirement extraction unit 134 analyzes the requirements of the question and extracts conditions for satisfying the requirements (step S26). The presentation unit 135 then extracts candidates who have the potential to answer the question (step S27). The presentation unit 135 then extracts a group of people who can satisfy the conditions from among the candidates (step S28). The presentation unit 135 then presents the candidates included in the extracted group to the questioner as candidates to whom the question should be posed (step S29).

[0072] Effect of the embodiment The question generation unit 133 generates a detailed question based on the question message input by the first user, the chat log and attribute information of the first user. The requirement extraction unit 134 extracts requirements that are required of the user to whom the question is to be asked from the question generated by the question generation unit 133. The presentation unit 135 presents users who meet the requirements from among the multiple users as users to whom the question should be asked, based on the chat logs and attribute information of the multiple users.

[0073] In this way, the estimation device 10 does not infer the question recipient by simply referring to the question entered by the user, but rather infers the question recipient after detailing the question. Therefore, the estimation device 10 can prevent a discrepancy between the level of the answerer desired by the asker and the level of the answerer presented as the question recipient. As a result, according to this embodiment, it is possible to accurately introduce an answerer to a question.

[0074] The question generator 133 outputs a question regarding an ambiguous part included in the question message input by the first user, and refines the question based on the first user's answer to the output question. This allows the estimation device 10 to refine the question efficiently.

[0075] The process for presenting the query destination by the estimation device 10 may include the process described in International Application PCT / JP2023 / 044515.

[0076] [System Configuration, etc.] The components of each unit shown in the figure are conceptual functional units and do not necessarily have to be physically configured as shown. In other words, the specific form of distribution and integration of each device is not limited to that shown, and all or part of them can be functionally or physically distributed and integrated in any unit depending on various loads, usage conditions, etc. Furthermore, all or any part of the processing functions performed by each device can be realized by a CPU and a program executed by the CPU, or can be realized as hardware using wired logic.

[0077] Furthermore, among the processes described in the above embodiments, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically using a known method.In addition, the information including the processing procedures, control procedures, specific names, various data and parameters shown in the above documents and drawings can be changed as desired unless otherwise specified.

[0078] [Program] The above-described estimation device 10 can be implemented by installing a program (estimation program) as package software or online software on a desired computer. For example, by executing the above-described program on an information processing device, the information processing device can function as the estimation device 10. The information processing device referred to here includes mobile communication terminals such as smartphones, mobile phones, and PHS (Personal Handyphone Systems), as well as terminals such as PDAs (Personal Digital Assistants).

[0079] 12 is a diagram showing an example of a computer that executes an estimation program. The computer 1000 includes, for example, a memory 1010 and a CPU 1020. The computer 1000 also includes a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. These components are connected by a bus 1080.

[0080] The memory 1010 includes a read-only memory (ROM) 1011 and a random access memory (RAM) 1012. The ROM 1011 stores a boot program such as a basic input / output system (BIOS). The hard disk drive interface 1030 is connected to a hard disk drive 1090. The disk drive interface 1040 is connected to a disk drive 1100. A removable storage medium such as a magnetic disk or optical disk is inserted into the disk drive 1100. The serial port interface 1050 is connected to a mouse 1110 and a keyboard 1120, for example. The video adapter 1060 is connected to a display 1130, for example.

[0081] The hard disk drive 1090 stores, for example, an OS 1091, an application program 1092, a program module 1093, and program data 1094. That is, the programs that define the processes executed by the above-described estimation device 10 are implemented as program modules 1093 in which computer-executable code is written. The program modules 1093 are stored, for example, in the hard disk drive 1090. For example, the program modules 1093 for executing processes similar to those of the functional configuration of the estimation device 10 are stored in the hard disk drive 1090. Note that the hard disk drive 1090 may be replaced with an SSD (Solid State Drive).

[0082] Data used in the processing of the above-described embodiment is stored as program data 1094, for example, in the memory 1010 or the hard disk drive 1090. The CPU 1020 then reads the program module 1093 or the program data 1094 stored in the memory 1010 or the hard disk drive 1090 into the RAM 1012 as necessary and executes them.

[0083] The program module 1093 and program data 1094 may not necessarily be stored in the hard disk drive 1090, but may also be stored in a removable storage medium and read by the CPU 1020 via the disk drive 1100 or the like. Alternatively, the program module 1093 and program data 1094 may be stored in another computer connected via a network (such as a local area network (LAN) or a wide area network (WAN)). The program module 1093 and program data 1094 may then be read by the CPU 1020 from the other computer via the network interface 1070.

[0084] REFERENCE SIGNS LIST 10 Estimation device 11 Input / output unit 12 Storage unit 13 Control unit 121 Chat log DB 122 Attribute DB 131 Attribute analysis unit 132 Input reception unit 133 Question generation unit 134 Requirement extraction unit 135 Presentation unit

Claims

1. An estimation device comprising: a question generation unit that generates a more detailed question based on a question message input by a first user, the chat log and attribute information of the first user; a requirement extraction unit that extracts requirements required of the user to whom the question is being asked from the question generated by the question generation unit; and a presentation unit that presents users who meet the requirements from among the multiple users as users to whom the question should be asked, based on the chat logs and attribute information of the multiple users.

2. The estimation device described in claim 1, characterized in that the question generation unit outputs a question regarding an ambiguous part contained in the question message input by the first user, and refines the question based on the first user's answer to the output question.

3. An estimation method executed by an estimation device, comprising: a question generation step of generating a more detailed question based on a question message input by a first user, the chat log and attribute information of the first user; a requirements extraction step of extracting requirements required of the user to whom the question is to be asked from the question generated by the question generation step; and a presentation step of presenting users who satisfy the requirements from among the plurality of users as users to whom the question should be asked, based on the chat logs and attribute information of the plurality of users.

Images