Social network model generation device, method, and program

Abstract

One embodiment of the present invention, on the basis of a communication log which includes text messages conveying a plurality of conversations that took place among a plurality of users, estimates at least one conversation function for each of the messages on the basis of information representing a preset conversation instigation viewpoint, and generates a relationship structure representing the relationship between the conversation functions on the basis of information representing the sequencing of the messages. Thereafter, an action relationship between the conversation functions in the estimated relationship structure is estimated, and a functional model representing a functional structure of a conversation among the users is generated using the relationship structure and the action relationship of the conversation functions.

Description

Social network model generation device, method, and program

[0001] One aspect of the present invention relates to a social network model generation device, method, and program for generating a model that analyzes the functional structure of communication between users, for example.

[0002] For example, users engaged in business activities are required to not only perform routine tasks, but also to handle advanced tasks based on worker judgment, such as taking on new tasks and flexibly responding to highly non-routine tasks.To achieve this, it is important to form teams that can improve their learning and predictive abilities based on actual experience.Currently, team leaders manually observe the situation of the entire team based on their experience and intuition, and support users' work by communicating with them.

[0003] Meanwhile, in the field of behaviormetrics, research is being conducted on social network analysis, which quantitatively measures and quantifies the relationships between multiple users from communication logs, for example, on social networking services (SNS), emails, and chats, and analyzes the behavior of individual users based on these quantified relationships and their characteristics (see, for example, Non-Patent Document 1).

[0004] Yasuda Yuki, "Social Network Analysis: Its Theoretical Background and Scales," Behavioral Metrics, Vol. 21, No. 2, pp. 32-39, 1994.

[0005] By applying the above-mentioned social network analysis techniques to corporate activities in which work is carried out in teams, it becomes possible to quantitatively measure the relationships between each user within the team, and the results can be used to optimize work allocation, for example.

[0006] However, the social network analysis method described in Non-Patent Document 1 analyzes the relationships between users, i.e., who has spoken to whom. Therefore, it is not possible to analyze the functional structure of communication, such as what kind of influence a conversation has, i.e., what function it has, the relationships between conversations that show the connections and flow of conversations, or even the process by which a certain utterance is elicited when focusing on a certain conversation.

[0007] The present invention has been made in light of the above circumstances, and aims to provide a technique for generating a social network model that enables analysis of the functional structure of messages that represent conversations.

[0008] In order to solve the above problem, one aspect of a social network model generation device or method according to the present invention is to estimate at least one conversation function for each of a plurality of messages based on a communication log containing messages in which a plurality of conversations held between a plurality of users have been converted into text, based on information representing a predetermined perspective of the conversational influence, and generate a relational structure representing the relationships between the conversational functions based on information representing the order of the messages, estimate action relationships between the conversational functions in the estimated relational structure, and generate a functional model representing the functional structure of the conversations between the users using the relational structure of the conversational functions and the action relationships.

[0009] According to one aspect of the present invention, at least one conversation function is estimated for each conversation message included in a communication log based on the communication log, a relational structure representing the relationship between the estimated conversation functions is generated, and action relationships between the conversation functions are estimated. Then, a functional model representing the functional structure of a conversation held within a team, for example, is generated using the relational structure and action relationships.

[0010] Therefore, the above functional model makes it possible to know not only the relationship between which user a conversation took place from to which other user within a team, but also at least one conversation function that indicates what kind of influence the conversation had on other users, and the interaction relationship between conversation functions, which makes it possible to analyze the process and structure of how team formation and activities are promoted through a chain of influences in conversation.

[0011] That is, according to one aspect of the present invention, it is possible to provide a technology for generating a social network model that enables analysis of the functional structure in conversations.

[0012] FIG. 1 is a diagram illustrating an example of a communication system including a social network model generation device according to an embodiment of the present invention. FIG. 2 is a block diagram illustrating an example of the hardware configuration of a social network model generation device according to an embodiment of the present invention. FIG. 3 is a block diagram illustrating an example of the software configuration of a social network model generation device according to an embodiment of the present invention. FIG. 4 is a flowchart illustrating an example of the processing procedure and processing content of a conversation function model generation process executed by a control unit of the social network model generation device shown in FIG. 3. FIG. 5 is a flowchart illustrating an example of the processing procedure and processing content of a process for generating a relational structure between conversations, which is part of the functional model generation process shown in FIG. 4. FIG. 6 is a flowchart illustrating an example of the processing procedure and processing content of a process for estimating action relationships between conversations, which is part of the functional model generation process shown in FIG. 4. FIG. 7 is a diagram illustrating an example of the operation of the functional model generation process shown in FIG. 4. FIG. 8 is a diagram illustrating an example of a communication log. FIG. 9 is a diagram illustrating the estimation results of the functions of each conversation represented by the communication log shown in FIG. 8. FIG. 10 is a diagram illustrating an example of the generation results of a relational structure between conversations.

[0013] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0014] [One Embodiment] (Configuration Example) (1) System FIG. 1 is a diagram showing an example of a communication system including a social network model generating device CS according to one embodiment of the present invention.

[0015] A communication system according to one embodiment enables communication between a plurality of user terminals UT1 to UTn via a network NW, and also enables transmission of information data between the user terminals UT1 to UTn and an administrator terminal MT and a social network model generation device CS via the network NW.

[0016] The user terminals UT1 to UTn are used by a plurality of users engaged in business in a company, for example, and are made up of, for example, personal computers. The administrator terminal MT is used by an administrator of the company, and is made up of, for example, a personal computer.

[0017] The network NW is composed of a wide area network such as the Internet and an access network for accessing this wide area network. The access network may be a wired or wireless local area network (LAN), an optical transmission network, or a mobile communication network that adopts the 5G standard.

[0018] (2) Social Network Model Generating Device CS FIGS. 2 and 3 are block diagrams showing an example of the hardware configuration and software configuration of the social network model generating device CS, respectively.

[0019] The social network model generating device CS is provided as one of the functions of a server computer located on the web or in the cloud, for example. The social network model generating device CS may also be provided in a terminal such as a personal computer used by a user or a system administrator.

[0020] The social network model generation device CS comprises a control unit 1 having a hardware processor, and a storage unit having a program storage unit 2 and a data storage unit 3, and a communication I / F unit 4, which are connected to the control unit 1 via a bus 5.

[0021] The communication I / F unit 4 transmits and receives information data to and from the user terminals UT1 to UTn and the administrator terminal MT in accordance with a communication protocol such as TCP / IP (Transmission Control Protocol / Internet Protocol) defined in the network NW.

[0022] The program storage unit 2 is, for example, a combination of a non-volatile memory such as a HDD (Hard Disk Drive) or SSD (Solid State Drive) as a storage medium that can be written to and read from at any time, and a non-volatile memory such as a ROM (Read Only Memory), and stores application programs necessary to execute various processes related to one embodiment of the present invention, in addition to middleware such as an OS (Operating System).

[0023] The data storage unit 3 is a combination of a non-volatile memory such as an HDD or SSD that can be written to and read from at any time, and a volatile memory such as a RAM (Random Access Memory), and its storage area includes a communication log storage unit 31, a viewpoint information storage unit 32, a conversation function label storage unit 33, an inter-conversation relation structure storage unit 34, a relationship information storage unit 35, and a function model storage unit 36.

[0024] The communication log storage unit 31 is used to store communication logs that are structured in text format as historical information about conversations between multiple users that took place between the user terminals UT1 to UTn. Each communication log contains a message in which the content of the conversation has been converted into text, and information indicating, for example, "who" and "when" is added to the message. An example of a communication log will be described in the operation example.

[0025] Information representing a viewpoint of conversational approach, which is defined in advance for estimating conversational function, is stored in the viewpoint information storage unit 32. An example of viewpoint information will be described in the operation example.

[0026] The conversation function label storage unit 33 is used to store information indicating the conversation functions estimated for the conversation messages included in the communication log, such as conversation function labels.

[0027] The conversation relation structure storage unit 34 is used to store information representing the relation structure of conversation functions.

[0028] The relationship information storage unit 35 stores information representing elements of predefined action relationships in order to estimate action relationships between conversation functions. An example of the elements of action relationships will be described in the operation example.

[0029] The functional model storage unit 36 ​​is used to store a model representing the functional structure of communication within a team, obtained by the control unit 1, which will be described later.

[0030] The control unit 1 includes, as processing functions required to realize one embodiment of the present invention, a communication log acquisition processing unit 11, a conversation function estimation processing unit 12, an inter-conversation relation structure generation processing unit 13, an inter-conversation action relation estimation processing unit 14, and a function model output processing unit 15.

[0031] Each of the processing units 11 to 15 is realized by causing a hardware processor of the control unit 1 to execute an application program stored in the program storage unit 2. Note that some or all of the processing units 11 to 15 may be realized using hardware such as an LSI (Large Scale Integration) or an ASIC (Application Specific Integrated Circuit).

[0032] The communication log acquisition processing unit 11 acquires a communication log representing the history of conversations between multiple users belonging to any business team within a company, for example, from a communication log management server (not shown), and stores the acquired communication log in the communication log storage unit 31.

[0033] The conversation function estimation processing unit 12 estimates conversation functions for conversation messages included in each communication log stored in the communication log storage unit 31 by referring to the viewpoints stored in the viewpoint information storage unit 32, and stores function labels representing the estimated conversation functions in the conversation function label storage unit 33.

[0034] The conversation relational structure generation processing unit 13 generates information representing the relational structure of the conversation functions stored in the conversation function label storage unit 33, based on information representing the utterance time of the conversation included in each communication log stored in the communication log storage unit 31. At this time, the relational structure is represented, for example, by a network structure in which conversation messages are nodes and relationships between messages are edges. The conversation relational structure generation processing unit 13 stores the generated information representing the relational structure between the conversation functions in the conversation relational structure storage unit 34.

[0035] The inter-conversation effect relation estimation processing unit 14 estimates the effect relations between the conversation functions in the above-mentioned relational structure by referring to the relational information stored in the relational information storage unit 35. Then, the inter-conversation effect relation estimation processing unit 14 evaluates the relational structure between the conversation functions based on the estimated effect relations between the conversation functions, and stores the conversation functions that satisfy predetermined conditions and the relational structure having the effect relations between the conversation functions in the functional model storage unit 36 ​​as a model representing the functional structure of communication within the team.

[0036] The conversational action relation estimation processing unit 14 also feeds back the evaluation results to the conversational relation structure generation processing unit 13, and performs processing to correct and reconstruct the relation structure between conversational functions.

[0037] The functional model output processing unit 15 reads out a model representing the functional structure of communication within the team from the functional model storage unit 36, and transmits the read out model from the communication I / F unit 4 to, for example, the manager terminal MT.

[0038] (Example of Operation) Next, an example of operation of the social network model generating device CS configured as above will be described.

[0039] 4 to 6 are flowcharts showing an example of the processing procedure and processing contents of the conversation function model generation processing executed by the control unit 1 of the social network model generation device CS.

[0040] (1) Viewpoint Information The viewpoint information storage unit 32 pre-stores viewpoint information defined for estimating the function of a conversation. A viewpoint represents a form of conversational engagement with other users, and can be broadly classified into a plurality of first functional viewpoints that lead to the sharing of information related to work and the learning and awareness of members, and a plurality of second functional viewpoints that encourage the promotion of speech.

[0041] Examples of "predictive / learning utterances" that correspond to the first functional perspective include K1: "request", K2: "approval / understanding", K3: "report / consultation", K4: "guidance", K5: "sharing success stories", K6: "sharing failure stories", K7: "prediction of possible events", K8: "confirmation of purpose / background", K9: "raising a problem", K10: "analysis of causes", K11: "proposing ideas for solutions", K12: "deciding on actions", and K13: "sharing the content of actions".

[0042] Furthermore, examples of "facilitative speech" that corresponds to the second functional perspective include L1: "sensory sharing," L2: "empathy," L3: "active listening," L4: "positive feedback," L5: "open questioning," L6: "summarizing the discussion," and L7: "negation."

[0043] The types of functional perspectives corresponding to "predictive / learning speech" and "facilitative speech" are not limited to those described above, but can be set arbitrarily.

[0044] (2) Relationship Information Between Conversation Functions The relationship information storage unit 35 stores beforehand labels between conversation functions that represent elements of action relationships defined to estimate action relationships between conversation functions.

[0045] Examples of labels between conversational functions that represent functional relationships include "strengthening," "weakening," "promotion," "suppression," "presupposition," "control," "resource," "cooperation," "complementation," "conflict," "disagreement," and "neutralization."

[0046] "Enhancement" refers to when one conversation function strengthens the function of another conversation function (strengthens its effect). "Weakening" refers to when one conversation function weakens the function of another conversation function (weakening its effect). "Facilitation" refers to when one conversation function encourages (speeds up) the function of another conversation function. "Inhibition" refers to when, on the other hand, one conversation function suppresses (slows down) the function of another conversation function.

[0047] "Prerequisite" refers to the case where a conversational function is a prerequisite for another conversational function to begin functioning. "Control" refers to the case where a conversational function controls and manipulates the way another conversational function works. "Resource" refers to the case where a conversational function is a resource necessary for another conversational function to continue functioning.

[0048] "Cooperation" refers to when one conversational function cooperates with another to create a function. "Complementary" refers to when one conversational function complements the function of another. "Conflict" refers to when one conversational function competes with another. "Disagreement" refers to when one conversational function depends on another. "Neutralization" refers to when one conversational function neutralizes and cancels out the function of another.

[0049] The labels between conversation functions that represent the action relationships are not limited to the above examples, and can be defined arbitrarily as needed.

[0050] (3) Acquisition of Communication Logs For example, suppose that an instruction to generate a functional model for a certain business team within a company is sent from the manager terminal MT. In this case, upon receiving the instruction to generate a functional model in step S1, the control unit 1 of the social network model generation device CS acquires a communication log representing the conversation history of multiple users belonging to the business team as follows:

[0051] That is, in step S2, the control unit 1 of the social network model generating device CS accesses the communication log management server based on, for example, the business team ID included in the functional model generation instruction under the control of the communication log acquisition processing unit 11. Then, the control unit 1 acquires a communication log that records conversations between multiple users belonging to the business team from the communication log management server, and stores the acquired communication log in the communication log storage unit 31.

[0052] A communication log is, for example, a structured text record of conversations exchanged using a specific channel of a business chat tool, or a history of conversations exchanged through a web conference or on-site, with the content of each conversation being converted into text and included as a message. The conversation data is not limited to text data and may be audio data. Furthermore, from the perspective of privacy of personal information, etc., the conversation data may be anonymized or partially masked with respect to "who" and "the content of the conversation." The acquisition range may also be specified as a conversation period.

[0053] 8 shows an example of an acquired communication log. In this example, the communication log includes, for each conversation, information made up of a number [num] indicating the order of the conversation, the start time [time] of the conversation, the names of users A, B, C, ... [member] who participated in the conversation, and messages M1, M2, M3, ... [message] indicating the content of the conversation.

[0054] Examples of conversation messages M1, M2, M3, ... are shown below: Note that "@person's name" indicates the other party (destination) of the conversation.

[0055] M1: @B Can we please introduce an RPA tool to the ○○ goods purchasing process? M2: We'd like to start operations within this month, so first, we'd like you to create a prototype RPA scenario this week. M3: Understood. M4: That's a good idea, could you tell me more about it? M5: @C I'm thinking of creating an RPA scenario that will input data into the ◆◆ system and submit approval requests in a nightly batch. M6: @A That's a good idea. Why did you consider introducing an RPA tool to goods purchasing in the first place? M7: Mr. H told me that he was worried about the hassle of purchasing procedures in the ◆◆ system. M8: Goods purchasing is a routine task, so we thought an RPA tool would be a good fit. M9: That's great! We'd love to hear from you! M10: I think it would be a good idea to prepare manuals and forms along with the RPA scenario to make it easier to use. M11: I see, I understand. M12: Thank you for your advice. I would definitely like to take that into consideration when creating it. M13: @A I would like to report that I have completed the scenario for RPA implementation for the ◆◆ system that you requested the other day. We will now begin operational verification. Note that if the communication log has been stored in advance in the communication log storage unit 31, the process of retrieving the communication log will be omitted.

[0056] (4) Estimation of Conversational Functions Next, in step S3, the control unit 1 of the social network model generation device CS, under the control of the conversational function estimation processing unit 12, executes a process of estimating the conversational function of each of the conversational messages included in the communication log as follows.

[0057] That is, the conversation function estimation processing unit 12 reads out the communication logs one by one from the communication log storage unit 31, and for the conversation messages M1, M2, M3, ... included in the read communication logs, estimates the conversation function, which is the function that the conversation message has on other users, based on the viewpoint information stored in the viewpoint information storage unit 32.

[0058] The estimation of conversational functions may be performed, for example, by a method based on rules maintained for each perspective, or by a method of determining the likelihood of each perspective using an agent that has mechanically learned rules in generative artificial intelligence (hereinafter referred to as generative AI), etc. Furthermore, if the communication log contains information that indicates the structural relationship between conversational messages (for example, context or thread identity), the estimation may be performed based on context information between messages associated by this information.

[0059] As an example of a method of making inference based on rules, for example, from the perspective of "sharing cases," a rule can be set such that "the text of the conversation contains the string "experienced," or the string "it worked well," or the string "it did not work well," and an inference method can be used by determining from the text whether or not it is based on the above rule.

[0060] On the other hand, as an estimation method using a mechanically trained agent, a method that uses the results of commands given to the agent by prompts from the perspective of "sharing cases" can be used. For example, when using an estimation device based on generative AI, prompts are used that process communication logs by dividing them into specific number of days, or prompts that process each individual conversation.

[0061] FIG. 9 shows an example of the estimation results of conversation functions for conversation messages M1, M2, M3, . . . included in the communication log shown in FIG.

[0062] In this example, the results of determining whether each conversation message M1, M2, M3, ... corresponds to K1: "Request", K2: "Acknowledgement / Consent", K3: "Report / Consultation", K4: "Instruction", K5: "Sharing of success stories", K6: "Sharing of failure stories", K7: "Prediction of possible events", K8: "Confirmation of purpose / background", K9: "Raising a problem", K10: "Analysis of causes", K11: "Proposal of solution ideas", K12: "Decision on action", K13: "Sharing of action content", or L1: "Shared sense", L2: "Empathy", L3: "Active listening", L4: "Positive feedback", L5: "Open question", L6: "Summarization of discussion", L7: "Negative", are displayed as "1" or "0".

[0063] The conversation function estimation processing unit 12 stores the function labels indicating the estimated conversation functions in the conversation function label storage unit 33 in association with the conversation numbers.

[0064] (5) Generation of relational structures between conversation functions Next, in step S4, the control unit 1 of the social network model generation device CS, under the control of the inter-conversation relational structure generation processing unit 13, executes a process of generating a relational structure for each conversation function stored in the conversation function label storage unit 33 as follows.

[0065] FIG. 5 is a flowchart showing an example of the processing procedure and processing content of the inter-conversation function relation structure generation processing executed by the inter-conversation relation structure generation processing unit 13.

[0066] That is, first, in step S41, the conversation relation structure generation processing unit 13 acquires timestamp information (e.g., start time) as information indicating the order of conversations from each communication log stored in the communication log storage unit 31, and determines the chronological order of each conversation in step S42. Then, in step S43, the conversation relation structure generation processing unit 13 generates, as a relationship structure between the conversation functions, a network structure in which the conversation functions stored in the conversation function label storage unit 33 are arranged in the chronological order, with each conversation function as a node and the relationships between the conversation functions as edges.

[0067] Figure 10 shows an example of the relationship structure between the generated conversation functions. In this example, a network structure is shown in which nodes N1 to N14 are arranged in conversation order. Note that the conversations corresponding to nodes N1, N2, and N7, N8 are shown as occurring simultaneously.

[0068] The relationship structure between the conversation functions can also be generated based on additional information included in the communication log. For example, in some communications, the context between conversations is one-dimensional, as in voice data, while in other cases, the relationships between conversations structured for each thread are defined, as in chat logs. In such cases, the sequential relationships are included in the communication log as additional information. Therefore, it is possible to generate a network structure by referencing the additional information included in the communication log.

[0069] It is also possible to use an agent that has mechanically learned rules such as generative AI to infer the conversation process from the content of the conversation, and use this as information on the order of the conversation to generate a network structure.

[0070] The conversation relation structure generation processing unit 13 stores information representing the relation structure between conversation functions generated as a network structure in the conversation relation structure storage unit 34 .

[0071] (6) Estimation of Interaction Relationships Between Conversation Functions Next, in step S5, the control unit 1 of the social network model generation device CS, under the control of the inter-conversation action relationship estimation processing unit 14, executes a process of estimating the interaction relationships between conversation functions in the generated relational structure as follows.

[0072] FIG. 6 is a flowchart showing an example of the processing procedure and processing contents of the conversational action relation estimation processing executed by the conversational action relation estimation processing unit 14.

[0073] That is, in step S51, the conversation action relation estimation processing unit 14 selects a pair of conversations that have a relationship from the information representing the network structure stored in the conversation relation structure storage unit 34, and in step S52, estimates the action relations between the conversation functions in the selected pair of conversations based on their respective functions. This estimation of the action relations uses information representing the relationships between the conversation functions stored in the relationship information storage unit 35.

[0074] For example, the inter-conversation effect relation estimation processing unit 14 selects a label corresponding to the effect relation between the selected conversation function and another conversation function connected thereto from among a plurality of inter-conversation labels representing effect relations between conversation functions stored in the relation information storage unit 35. More specifically, the inter-conversation label corresponding to the effect relation between the selected conversation function and another conversation function connected thereto is selected from among "strengthening," "weakening," "promotion," "suppression," "premise," "control," "resource," "cooperation," "complement," "conflict," "disagreement," and "neutralization."

[0075] The estimation of the interaction relationships between conversation functions may be performed using a method based on rules set for each conversation function, or a method may be adopted in which an agent that mechanically learns rules such as generative AI is used to select interaction relationships between conversation functions whose likelihood is equal to or exceeds a predetermined threshold.Furthermore, as a method for estimating the interaction relationships between conversation functions, a method may be used in which the interaction relationships between conversation functions are estimated from context information between conversation messages that are associated by structural relationships (such as context or thread identity) of the conversation messages included in the communication log.

[0076] (7) Evaluation and reconstruction of relational structure between conversation functions The inter-conversation action relation estimation processing unit 14 feeds back the estimation result of the action relation to the inter-conversation relation structure generation processing unit 13 in step S53. In response to this, in step S44, the inter-conversation relation structure generation processing unit 13 evaluates the initial solution of the relational structure between conversation functions stored in the inter-conversation relation structure storage unit 34 based on the estimation result of the action relation that has been fed back.

[0077] For example, the conversation relation structure generation processing unit 13 determines whether or not there are edges to which conversation labels indicating the above-mentioned action relationships are not assigned in the network structure stored in the conversation relation structure storage unit 34 as an initial solution of the relation structure between conversation functions, and then modifies and reconstructs the network structure that is the initial solution based on the above-mentioned determination result.

[0078] For example, the conversation relation structure generation processing unit 13 deletes any edges to which no conversation function label is assigned, and then updates the initial solution of the network structure stored in the conversation relation structure storage unit 34 to the reconstructed network structure.

[0079] The conversation effect relation estimation processing unit 14 reads the reconstructed network structure from the conversation effect relation structure storage unit 34, and assigns conversation function labels and conversation labels indicating effect relations to the read reconstructed network structure.Then, the network structure to which the conversation function labels and conversation labels have been assigned is stored in the functional model storage unit 36 ​​as a functional model representing the functional structure of conversation functions between users within the team.

[0080] Figure 7 shows an example of a functional model in which the functions of each conversation, the relationship structure between the conversation functions, labels indicating the conversation functions, and labels indicating the functional relationships between the conversation functions are assigned to the relationship structure, which are estimated or generated by the processes (3) to (6) described above.

[0081] (7) Output of Functional Model Finally, in step S6, the control unit 1 of the social network model generation device CS, under the control of the functional model output processing unit 15, reads out the functional model from the functional model storage unit 36 ​​and transmits the read-out functional model from the communication I / F unit 4 to, for example, the administrator terminal MT.

[0082] Therefore, based on the above functional model sent from the social network model generation device CS, the administrator can recognize not only the functional structure of the conversation function within the team, that is, the relationships between speakers, i.e., which user has spoken to which user within the team, but also the function of the conversation itself, which indicates what kind of influence the conversation has had on others, and the interaction relationships between conversations, and furthermore, the process by which team formation and activities are promoted through a chain of influences in the conversation.

[0083] (Effect) As described above, in one embodiment, based on a communication log containing messages in which conversations held by multiple users within a team have been converted into text, at least one conversation function possessed by each conversation message is estimated from the perspective of influence, and information representing the relational structure of the conversation functions is generated based on information representing the order of the conversation messages, and action relationships between the conversation functions are estimated on the generated relational structure of the conversation functions, and the relational structure is modified and reconstructed based on the estimated results of this action relationship, thereby generating a functional model representing the functional structure of the conversation within the team.

[0084] Therefore, with the above functional model, the administrator can understand not only the speaker relationships within the team, i.e., which user initiated the conversation, but also the function of the conversation itself, which indicates what kind of influence the conversation had on others, and the interaction relationships between conversations.Furthermore, it becomes possible to recognize the process and structure of how team formation and activities are promoted through a chain of influences in the conversation.

[0085] Other Embodiments (1) In one embodiment, a functional model of a conversation function is generated based on the flow of text data exchanged within a team via chat, for example, regarding the same task. However, the present invention is not limited to this. For example, common conversation message patterns may be extracted from the flow of daily text data exchanged dispersedly throughout the day, and a functional model of conversation within the team may be generated based on the extracted conversation message patterns.

[0086] (2) In one embodiment, the social network model generation device CS according to the present invention is described as being installed on a server computer on the Web or cloud. However, the present invention is not limited to this. The social network model generation device CS according to the present invention may be installed on a personal computer such as a user terminal or an administrator terminal, or may be installed on a dedicated analysis device used by a social network analyst. Furthermore, the processing functions of the social network model generation device CS may be distributed across multiple information processing devices, such as server computers and personal computers, that are capable of communicating with each other via a network.

[0087] (3) In one embodiment, an example was given in which communication logs are acquired in bulk from a communication log management server, but communication logs may also be collected and stored from each of the user terminals UT1 to UTn of the user who spoke.

[0088] (4) The organizations to be analyzed are not limited to business teams of a company, but may also be countries, local governments, schools, sports teams, clubs, and various other organizations. Furthermore, the type of social network model generation device CS, the configuration of each processing function provided by this device, the processing procedure, and the processing content can be modified in various ways without departing from the spirit of the present invention.

[0089] Although the embodiments of the present invention have been described in detail above, the above description is merely an example of the present invention in every respect. It goes without saying that various improvements and modifications can be made without departing from the scope of the present invention. In other words, when implementing the present invention, specific configurations according to the embodiments may be appropriately adopted.

[0090] In short, this invention is not limited to the above-described embodiments, and in the implementation stage, the components can be modified and embodied without departing from the spirit of the invention. Furthermore, various inventions can be formed by appropriately combining multiple components disclosed in the above-described embodiments. For example, some components may be omitted from all the components shown in the embodiments. Furthermore, components from different embodiments may be appropriately combined.

[0091] CS...social network model generating device MT...administrator terminal UT1 to UTn...user terminal NW...network 1...control unit 2...program storage unit 3...data storage unit 4...communication I / F unit 5...bus 11...communication log acquisition processing unit 12...conversation function estimation processing unit 13...inter-conversation relation structure generation processing unit 14...inter-conversation action relation estimation processing unit 15...functional model output processing unit 31...communication log storage unit 32...viewpoint information storage unit 33...conversation function label storage unit 34...inter-conversation relation structure storage unit 35...relation information storage unit 36...functional model storage unit

Claims

1. A social network model generation device comprising: a first processing unit that, based on a communication log containing messages that are text versions of multiple conversations held between multiple users, estimates at least one conversation function for each of the messages based on information that represents a predetermined perspective of the conversation's influence; a second processing unit that generates a relational structure that represents the relationships between the conversational functions based on information that represents the order of the messages included in the communication log; a third processing unit that estimates the action relationships between the conversational functions in the relational structure; and a fourth processing unit that generates a functional model that represents the functional structure of the conversation between the users using the relational structure between the conversational functions and the action relationships.

2. The social network model generating device of claim 1, wherein the first processing unit prepares in advance, as information representing the perspective of the interaction, a plurality of first functional perspectives that induce prediction or learning regarding the user's activities and a plurality of second functional perspectives that encourage the user to speak, and estimates the conversation function by comparing the message with the first functional perspectives and the second functional perspectives.

3. The social network model generating device of claim 1, wherein the third processing unit pre-stores information representing the multiple expected interaction relationships between the conversation functions, and estimates the interaction relationships between the conversation functions by comparing the conversation functions with the stored information representing the interaction relationships.

4. The social network model generating device of claim 1, wherein the fourth processing unit modifies and reconstructs the relationship structure based on the action relationship estimated by the third processing unit, and generates a functional model representing the functional structure of the conversation between the users based on the reconstructed relationship structure and information representing the conversation function and the action relationship.

5. The social network model generating device according to claim 1, wherein the second processing unit generates, as the relationship structure, a network structure in which each of the plurality of messages is represented as a node and the relationships between the messages are represented as edges.

6. A social network model generation method executed by an information processing device, comprising the steps of: estimating at least one conversation function for each of a communication log containing messages in which a plurality of conversations held between a plurality of users have been converted into text; estimating a relationship structure representing the relationships between the conversation functions based on information representing the order of the messages contained in the communication log; estimating action relationships between the conversation functions in the relationship structure; and generating a functional model representing the functional structure of the conversation between the users using the relationship structure and the action relationships between the conversation functions.

7. A program that causes a processor provided in a social network model generation device to execute at least one of the processes performed by each of the first to fourth processing units provided in the social network model generation device according to any one of claims 1 to 5.

Images