WO2023144949A1 - Dispositif d'aide à la contre-mesure de risque, dispositif d'apprentissage, procédé d'aide à la contre-mesure de risque, procédé d'apprentissage et programme - Google Patents

Dispositif d'aide à la contre-mesure de risque, dispositif d'apprentissage, procédé d'aide à la contre-mesure de risque, procédé d'apprentissage et programme Download PDF

Info

Publication number
WO2023144949A1
WO2023144949A1 PCT/JP2022/002987 JP2022002987W WO2023144949A1 WO 2023144949 A1 WO2023144949 A1 WO 2023144949A1 JP 2022002987 W JP2022002987 W JP 2022002987W WO 2023144949 A1 WO2023144949 A1 WO 2023144949A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
risk
transmission
transmission information
learning
Prior art date
Application number
PCT/JP2022/002987
Other languages
English (en)
Japanese (ja)
Inventor
ナットナリー サミッティメティーン
Original Assignee
日本電気株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 日本電気株式会社 filed Critical 日本電気株式会社
Priority to PCT/JP2022/002987 priority Critical patent/WO2023144949A1/fr
Publication of WO2023144949A1 publication Critical patent/WO2023144949A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services

Definitions

  • the present invention relates to a risk coping support device, a learning device, a risk coping support method, a learning method and a program.
  • Patent Literature 1 discloses a conversation analysis device that aims to efficiently discover important opinions of conversation speakers.
  • Patent Literature 1 As described above, in situations where users have the opportunity to express their opinions in various formats, there are cases in which priority should be given to responding according to the risk level of such transmission. Regarding this point, even if the technique of Patent Literature 1 is used, a problem remains from the viewpoint of handling according to the risk level.
  • One aspect of the present invention has been made in view of the above problems, and an example of its purpose is to preferably identify a case that should be treated preferentially according to the risk information of the content sent by the user. It is to provide technology that can
  • a risk handling support device includes acquisition means for acquiring first transmission information related to transmission of a target, and a degree to which the second transmission information is to be handled with the second transmission information as input.
  • Risk information calculation means for calculating risk information of the transmission content of the target based on the first transmission information acquired by the acquisition means, using a learned model that outputs risk information indicating
  • a prioritizing means for prioritizing the first transmission information in the order to be dealt with based on the risk information of the one transmission information and the risk information of the transmission information other than the first transmission information.
  • a learning device includes acquisition means for acquiring teacher data including a plurality of pairs of transmitted information and labels representing risk information of the transmitted information; and a learning means for learning a model for outputting risk information used for attaching, using the teacher data.
  • a risk countermeasure support method acquires first transmission information related to transmission of a target, uses second transmission information as input, and indicates the degree to which countermeasures should be taken for the second transmission information.
  • the risk information of the transmission content of the target is calculated, and the risk information of the first transmission information and the first prioritizing the first transmission information in the order in which it should be dealt with, based on the risk information of the transmission information other than the transmission information.
  • a learning method acquires teacher data including a plurality of pairs of transmitted information and labels representing risk information of the transmitted information, receives the transmitted information, and ranks the transmitted information.
  • a model that outputs risk information used for is learned using the teacher data.
  • a program provides a computer with an acquisition process for acquiring first transmission information related to transmission of a target, and a degree to which the second transmission information is to be handled with the second transmission information as input.
  • a risk information calculation process for calculating risk information of the transmission content of the target based on the first transmission information acquired in the acquisition process using a learned model that outputs risk information indicating the a ranking process for prioritizing the first transmission information in the order in which it should be dealt with based on the risk information of one transmission information and the risk information of the transmission information other than the first transmission information; .
  • a program provides a computer with an acquisition process for acquiring teacher data including a plurality of sets of transmission information and a label representing risk information of the transmission information, and inputting the transmission information, the transmission information and a learning process of learning a model that outputs risk information used for ranking using the teacher data.
  • FIG. 1 is a block diagram showing the configuration of a risk countermeasure support device according to Exemplary Embodiment 1;
  • FIG. FIG. 2 is a flow chart showing the flow of a risk coping support method according to exemplary Embodiment 1;
  • FIG. 1 is a block diagram showing the configuration of a learning device according to Exemplary Embodiment 1;
  • FIG. 3 is a flow diagram showing the flow of a learning method according to exemplary embodiment 1;
  • FIG. 11 is a block diagram showing the configuration of a risk coping support system according to Exemplary Embodiment 2;
  • FIG. FIG. 11 is a flow chart showing the flow of a risk coping support method according to exemplary embodiment 2;
  • FIG. 10 is a diagram showing an example of a screen according to exemplary embodiment 2;
  • FIG. 11 is a block diagram showing the configuration of a risk coping support system according to exemplary Embodiment 3;
  • FIG. 11 is a block diagram showing the configuration of a risk coping support system according to exemplary embodiment 4; It is a block diagram showing the configuration of a computer that functions as a risk coping support device or a learning device according to each exemplary embodiment.
  • FIG. 1 is a block diagram showing the configuration of the risk handling support device 1.
  • the risk coping support device 1 is a device that assists coping with risks related to transmission information by prioritizing transmission information.
  • the risk countermeasure support device 1 is, for example, a device that supports the work of an inquiry window for an EC site.
  • the risk management support device 1 is, for example, a device that supports risk management related to information transmitted by SNS (Social Networking Service).
  • the transmission information is information related to transmission by the caller, and includes, for example, at least one of text data, voice data, and image data representing the content of the transmission. More specifically, the transmission information includes, for example, data representing the contents of inquiries from consumers in commercial transactions or data representing the contents of transmissions in SNS.
  • the transmission information is, for example, the purchase information of the sender in the commercial transaction (for example, the name, category, purchase price, purchase date, etc.) or the exhibition information (for example, the name, category, purchase price, purchase price, etc. of the exhibition product). date and time) may be included.
  • the transmission information may include, for example, information regarding the exchange of messages (chat, etc.) between the seller and the purchaser in commercial transactions.
  • the transmission information includes, for example, information related to the caller who made the call, or transaction information indicating past commercial transactions by the caller who made the call. More specifically, the transmission information may include, for example, information indicating the seller or purchaser of the commercial transaction (eg, the seller's or purchaser's ID, account name, phone number, email address, etc.). However, the transmission information is not limited to the above example, and may be other information.
  • Acquisition unit 11 acquires first transmission information relating to a target transmission.
  • the first transmission information is transmission information for which risk information is calculated or prioritized.
  • the acquisition unit 11 may receive the first transmission information from another device connected via a communication line.
  • the acquisition unit 11 may acquire the first transmission information from an input device connected via an input interface.
  • the acquisition unit 11 may acquire the first transmission information by reading the first transmission information from a storage device built into the risk management support device 1 or an external storage device.
  • the risk information calculation unit 12 uses a learned model (hereinafter also referred to as a “risk evaluation model”) to calculate the risk information of the target transmission content based on the first transmission information acquired by the acquisition unit 11. Calculate As an example, the risk information calculation unit 12 calculates risk information by inputting the first transmission information acquired by the acquisition unit 11 or the feature amount of the first transmission information into a learned model.
  • the feature amount of the first transmission information is, for example, a vector representing the feature of the first transmission information.
  • the first transmission information may be, for example, information indicating whether or not a specific keyword is included in the first transmission information.
  • Standard Scaler, One-Hot-Encoding, and word2vec are examples of algorithms for conversion to data representing features. However, the algorithms for conversion to feature data are not limited to these, and other algorithms may be used.
  • the characteristic amount of transmission information is simply referred to as “transmission information”. In other words, in this specification, the transmission information may include the feature amount of the transmission information.
  • a learned model is a model constructed by machine learning.
  • the input to the model is the second outgoing information, and the output from the model is risk information indicating the degree to which the second outgoing information should be dealt with. That is, the learned model is generated by learning the relationship between the transmitted information and the risk information of the transmitted information.
  • the method of machine learning of the model is not limited, and as an example, a decision tree-based, linear regression, or neural network method may be used, or two or more of these methods may be used.
  • Decision tree bases include, for example, LightGBM (Light Gradient Boosting Machine), Random Forest, and XGBoost.
  • Linear regression includes, for example, Bayesian regression, support vector regression, Ridge regression, Lasso regression, and ElasticNet.
  • Neural networks include, for example, deep learning.
  • the risk information is information regarding the risk of the contents of transmission by the user.
  • the risk information is information indicating the level of risk. More specifically, the risk information is, for example, information that indicates the degree to which inquiries or troubles are likely to occur in commercial transactions related to transmission. The degree may be represented by a level of "high”, “medium”, or “low”, or may be represented by a numerical value (for example, a value from 0 to 1 or a value from 1 to 10 in 10 steps).
  • the inquiry includes, for example, an inquiry about the product, a complaint about the product, a request for return of goods, and a request for some sort of response regarding the transaction.
  • the ranking unit 13 prioritizes the first transmission information in the order to be dealt with based on the risk information of the first transmission information and the risk information of the transmission information other than the first transmission information.
  • the transmission information other than the first transmission information is, for example, information representing an inquiry that has not yet been answered at the inquiry window of the EC site.
  • the transmission information other than the first transmission information is, for example, transmission information by a sender other than the sender of the first transmission information on the SNS.
  • Risk information is attached in advance to transmission information other than the first transmission information.
  • the risk information attached to transmission information other than the first transmission information is risk information calculated by the risk information calculation unit 12 as an example.
  • the ranking unit 13 prioritizes the first transmission information by comparing the risk information of the first transmission information and the risk information of transmission information other than the first transmission information. Further, as an example, the ranking unit 13 calculates the priority of the first transmission information based on the risk information of the first transmission information, and ranks the first transmission information based on the calculated priority. may In addition, the ranking unit 13 may rank by sorting the risk information values of each transmission information, as an example.
  • the first transmission information regarding the target transmission is acquired, the second transmission information is input, and the second transmission information is handled.
  • a learned model that outputs risk information indicating the degree to which it should be performed, based on the acquired first transmission information, calculates the risk information of the target transmission content, and calculates the risk information of the first transmission information.
  • a configuration is adopted in which the first transmission information is prioritized in the order in which it should be dealt with based on the risk information of the transmission information other than the first transmission information. Therefore, according to the risk handling support device 1 according to the present exemplary embodiment, it is possible to obtain the effect of being able to preferably specify the case to be dealt with preferentially according to the risk information of the contents of the transmission by the user.
  • FIG. 2 is a flowchart showing the flow of the risk coping support method S1.
  • the acquisition unit 11 acquires the first transmission information regarding the target transmission.
  • the risk information calculation unit 12 uses a trained model that inputs the second transmission information and outputs risk information indicating the degree to which the second transmission information should be dealt with, and acquires in step S11 Based on the first transmission information obtained, the risk information of the target transmission content is calculated.
  • the ranking unit 13 prioritizes the first transmission information in the order in which it should be handled, based on the risk information of the first transmission information and the risk information of the transmission information other than the first transmission information. attach.
  • the first transmission information regarding the target transmission is acquired, the second transmission information is input, and the second transmission information is handled.
  • a learned model that outputs risk information indicating the degree to which it should be performed, based on the acquired first transmission information, calculates the risk information of the target transmission content, and calculates the risk information of the first transmission information.
  • a configuration is adopted in which the first transmission information is prioritized in the order in which it should be dealt with based on the risk information of the transmission information other than the first transmission information. Therefore, according to the risk countermeasure support method S1 according to the present exemplary embodiment, it is possible to obtain the effect of being able to preferably identify the case to be dealt with preferentially according to the risk information of the contents of the transmission by the user.
  • FIG. 3 is a block diagram showing the configuration of the learning device 2.
  • the learning device 2 is a device for learning a model that outputs risk information used for ranking transmission information.
  • the learning device 2 includes an acquisition unit 21 and a learning unit 22 .
  • the acquiring unit 21 acquires teacher data including a plurality of sets of transmission information and labels representing risk information of the transmission information.
  • the acquisition unit 21 may receive teacher data from another device connected via a communication line, or may acquire teacher data from an input device connected via an input interface.
  • the acquisition unit 21 may acquire the teacher data by reading the teacher data from a storage device built into the learning device 2 or an external storage device, for example.
  • the teacher data is, for example, data including a plurality of pairs of past transmission information and labels representing risk information of the transmission information.
  • the label representing the risk information indicates the risk level of the transmitted information.
  • the risk level is "0" when no problem occurs with the transmitted information, It may be "1".
  • the label is input by a commercial transaction user (seller, purchaser, etc.), an EC site operator, or the like.
  • the acquiring unit 21 associates the transmission information with the label input by the user for the transmission information and stores them in a predetermined storage device.
  • the learning unit 22 uses training data to learn a model that receives transmission information and outputs risk information used for ranking the transmission information. That is, the learning unit 22 learns the relationship between the transmission information and the risk information of the transmission information, and generates a risk evaluation model.
  • the method of machine learning of the model is not limited, and as an example, a decision tree-based, linear regression, or neural network method may be used, or two or more of these methods may be used.
  • Decision tree bases include, for example, LightGBM, Random Forest, and XGBoost.
  • Linear regression includes, for example, Bayesian regression, support vector regression, Ridge regression, Lasso regression, and ElasticNet.
  • Neural networks include, for example, deep learning.
  • ⁇ Effect of learning device 2> teacher data including a plurality of pairs of transmission information and labels representing risk information of the transmission information is acquired, the transmission information is input, and the A configuration is adopted in which a model for outputting risk information used for ranking transmitted information is learned using teacher data. For this reason, according to the learning device 2 according to the present exemplary embodiment, it is possible to generate a model for suitably identifying a case to be treated preferentially according to the risk information of the content sent by the user. be done.
  • FIG. 4 is a flow diagram showing the flow of the learning method S2.
  • the acquisition unit 21 acquires teacher data including a plurality of sets of transmission information and labels representing risk information of the transmission information.
  • the learning unit 22 uses training data to learn a model that receives transmission information and outputs risk information that is used to rank the transmission information.
  • FIG. 5 is a block diagram showing the configuration of a risk handling support system 100A according to this exemplary embodiment.
  • the risk coping support system 100A is a system that assists risk coping in an EC site. More specifically, as an example, the risk handling support system 100 provides services such as detecting in advance transactions that are likely to cause problems, and presenting inquiries to the person in charge in the order in which they should be handled.
  • the risk handling support system 100 includes a risk handling support device 1A, an operator terminal 3, and a user terminal 4.
  • the risk management support device 1A is a device that provides a service for supporting risk management in an EC site, and is a cloud server as an example.
  • the operator terminal 3 is a terminal used by the operator of the EC site.
  • the user terminal 4 is a terminal used by users (sellers, purchasers, etc.) of the EC site.
  • the risk management support device 1A communicates with the operator terminal 3 and the user terminal 4 via the communication line N.
  • FIG. Although the specific configuration of the communication line N does not limit this exemplary embodiment, examples of the communication line include wireless LAN (Local Area Network), wired LAN, WAN (Wide Area Network), public line network, mobile data communication network, or a combination thereof.
  • the risk handling support device 1A includes a control section 10A, a storage section 20A, a communication section 30A and an input/output section 40A.
  • the communication unit 30A communicates with a device external to the risk management support device 1A via a communication line N.
  • the communication unit 30A transmits data supplied from the control unit 10A to other devices, and supplies data received from other devices to the control unit 10A.
  • Input/output unit 40A Input/output devices such as a keyboard, mouse, display, printer, and touch panel are connected to the input/output unit 40A.
  • the input/output unit 40A receives input of various kinds of information from the connected input device to the risk countermeasure support device 1A. Also, the input/output unit 40A outputs various kinds of information to the connected output device under the control of the control unit 10A.
  • an interface such as a USB (Universal Serial Bus) can be used as the input/output unit 40A.
  • control section 10A As shown in FIG. 5, the control section 10A includes an acquisition section 11, a risk information calculation section 12, a ranking section 13, a priority analysis section 14A, and an output section 15A.
  • the acquisition unit 11 receives the first transmission information regarding the target transmission from the user terminal 4 .
  • the first transmission information includes text data, audio data, and/or image data representing the content of the target transmission.
  • the data representing the content of the transmission includes, for example, purchase information or exhibition information in transactions.
  • the purchase information and the exhibition information include, for example, data (text data, voice data, still image data, moving image data, etc.) representing the product to be traded, data related to the purchaser, or data related to the seller.
  • the data representing the content of the transmission includes, for example, text data representing message exchanges (chat, etc.) between the seller and the purchaser.
  • the data representing the content of the call may include, for example, data representing the voice of the caller uttering the content of the inquiry.
  • the data representing the content of the transmission may include, for example, data representing a document image regarding the transaction.
  • the first transmission information includes information about the caller who made the target call.
  • Information about the caller includes, by way of example, information indicating the seller or purchaser of the transaction.
  • the information indicating the seller or purchaser includes, for example, the seller's or purchaser's name, contact information, identification information, and the like.
  • the first transmission information includes commercial transaction information indicating past commercial transactions by the sender who made the target transmission.
  • Transaction information includes, for example, information indicating the product purchased/sold by the sender, the date and time of purchase/sale, presence or absence of inquiries to business partners (sellers/buyers), frequency of inquiries to business partners, business partner This includes information indicating the content of inquiries to, whether or not troubles have occurred in transactions, or the frequency of troubles in commercial transactions.
  • the transaction information is not limited to these, and may include other information indicating past transactions by the caller.
  • the risk information calculation unit 12 uses the learned first model M1 to calculate the risk information of the target transmission content.
  • the risk information calculation unit 12 calculates risk information by inputting the feature amount of the first transmission information into the first model M1.
  • (Priority analysis unit 14A) 14 A of priority analysis parts analyze the priority of 1st transmission information from the risk information which the risk information calculation part 12 calculated, and the 1st transmission information which the acquisition part 11 acquired.
  • the priority analyzer 14A identifies the priority of the first outgoing information by inputting the risk information and the first outgoing information into the second model M2.
  • Priority is information for prioritizing the order in which a plurality of outgoing information should be dealt with.
  • Examples of high-priority transmission information include transmission information for transactions that are likely to result in returned goods, refunds, or cancellations, transmission information for transactions that fall under fraud, transmission information for transactions that were purchased with points for a limited period, and the like. be done.
  • the transmission information with high priority is not limited to these.
  • the ranking unit 13 prioritizes the first transmission information in the order in which it should be dealt with, based on the risk information and priority of the first transmission information and the risk information and priority of transmission information other than the first transmission information. Rank.
  • the transmission information other than the first transmission information is the transmission information accumulated in the transaction information database DB1, and is information representing an inquiry that has not yet been answered.
  • the ranking unit 13 ranks transmission information representing inquiries for which responses have not yet been completed based on risk information and priority. Below, the result of ranking by the ranking part 13 is also called "priority.”
  • the ranking result includes, for example, a list in which a plurality of transmission information including the first transmission information are arranged in order of priority.
  • the ranking results are output to other devices (operator terminal 3, etc.) via the communication unit 30A, and to output devices (display, printer, projector, speaker, etc.) via the input/output unit 40A. ) and storage in a storage device (storage unit 20A, external storage device, etc.).
  • the storage unit 20A stores a transaction information database DB1, a first model M1, and a second model M2.
  • the transaction information database DB1 is a database in which information related to transactions on EC sites is accumulated. A plurality of transmission information is accumulated in the transaction information database DB1.
  • the transaction information database DB1 contains the name of the user (seller/buyer), the user's contact information, User information such as user identification information is accumulated, along with user purchase information or exhibition information, information indicating the history of message exchanges between sellers and purchasers, and the like.
  • the transaction information database DB1 stores information such as presence/absence of inquiries from users to business partners (sellers/buyers), frequency of inquiries, content of inquiries, presence/absence of troubles in transactions, frequency of troubles in transactions, etc. Information indicating is accumulated.
  • risk information indicating the risk level of past transactions is accumulated in the transaction information database DB1.
  • one piece of risk information may be associated with a plurality of transmission information, or transmission information and risk information may be associated on a one-to-one basis.
  • Risk information may be associated with a user on a one-to-one basis, or risk information may be associated with each of a plurality of past transactions for one user.
  • transaction information database DB1 stores the priority calculated by the priority analysis unit 14A and the ranking result (priority order ) is stored.
  • the first model M1 is a model used when the risk information calculator 12 calculates risk information.
  • the first model M1 receives as input the feature amount of the first transmission information, and outputs risk information indicating the degree to which the input first transmission information should be dealt with.
  • the risk information is, for example, a value from "0" to "3", and the higher the value, the higher the risk.
  • the first model M1 may be a learned model constructed by machine learning, or may be a model for identifying risk information on a rule basis.
  • the method of machine learning of the model is not limited, and as an example, a decision tree-based, linear regression, or neural network method may be used. Two or more of the methods may be used.
  • the first model M1 may calculate risk information based on a character string included in the input transmission information. For example, if the first transmission information includes the character string "difficult to understand”, add “1" to the risk information in the first model M1, and the character string “dirty” is included. If yes, add “2" to the risk information and include the character string “broken”, add “2" to the risk information and include the character string "I want to refund” In this case, "3" may be added to the risk information.
  • the first model M1 will set “1" as the risk information. Output. Further, for example, if (ii) the first transmission information to be input is a character string "the product that arrived was dirty", the first model M1 outputs "2" as the risk information. Further, for example, if (iii) the first transmission information to be input is a character string "the product that arrived is dirty and I want to refund", the first model M1 uses "2" and "3" as the risk information. " is added to output "5". In this case, among the above (i) to (iii), (iii) has the highest complaint risk, and (i) has the lowest complaint risk.
  • the second model M2 is a model used when the priority analysis unit 14A calculates the priority.
  • the second model M2 receives as input the feature quantity and risk information of the transmitted information, and outputs the priority of the transmitted information.
  • the second model M2 may be a learned model constructed by machine learning, or may be a model for specifying priorities on a rule basis.
  • the method of machine learning of the model is not limited, as an example, a decision tree-based, linear regression, or neural network method may be used, and Two or more of the methods may be used.
  • the operator terminal 3 and the user terminal 4 are, for example, a control unit, a storage unit, a communication unit, and an input/output unit. Prepare. An input device such as a keyboard and a mouse and an output device such as a display are connected to the input/output unit.
  • FIG. 6 is a flowchart showing the flow of the risk coping support method S1A executed by the risk coping support system 100A. Note that some steps may be performed in parallel or out of order. Also, the description of the already described contents will not be repeated.
  • Step S101 A user of the EC site uses an input device connected to the input/output unit of the user terminal 4 to perform an operation for transmitting the first transmission information.
  • the user terminal 4 transmits the first transmission information to the risk management support device 1A according to the user's operation.
  • the first transmission information is, for example, purchase information or exhibition information on the EC site, or information indicating the contents of chats between users, as described above.
  • step S102 the acquisition unit 11 receives the first transmission information transmitted by the user terminal 4, and stores the received first transmission information in the transaction information database DB1.
  • step S103 the acquisition unit 11 extracts feature amounts from the first transmission information.
  • Algorithms for extracting features include, for example, Standard Scaler, One-Hot-Encoding, and word2vec.
  • the acquiring unit 11 converts it into a standardized numerical value using, for example, a standard scaler.
  • the acquisition unit 11 converts it into a numerical value of “0” or “1” using One-Hot-Encoding, for example.
  • the acquisition unit 11 converts the first transmission information into a vector using word2vec, for example.
  • the acquisition unit 11 determines whether or not the specific keyword is included in the first transmission information. may be used as the feature amount.
  • step S104 the risk information calculation unit 12 calculates the risk information of the first transmission information by inputting the feature quantity of the first transmission information into the first model M1.
  • the risk information calculation unit 12 stores the calculated risk information in the transaction information database DB1 in association with the first transmission information received by the acquisition unit 11 .
  • the risk information calculation unit 12 may store the calculated risk information in the transaction information database DB1 in association with the user information corresponding to the first transmission information instead of in association with the first transmission information. good.
  • Step S105 the priority analysis unit 14A inputs the risk information calculated by the risk information calculation unit 12 and the feature amount of the first transmission information acquired by the acquisition unit 11 into the second model M2, Calculate the priority of the first transmission information.
  • the priority analysis unit 14A stores the calculated priority in the transaction information database DB1 in association with the first transmission information received by the acquisition unit 11 .
  • the priority analysis unit 14A may store the calculated priority in the transaction information database DB1 in association with the user information corresponding to the first transmission information instead of associating it with the first transmission information. good.
  • the ranking unit 13 refers to the risk information and priority of the first transmission information and the risk information and priority of transmission information other than the first transmission information, and selects the first transmission information. Prioritize. As an example, the ranking unit 13 sorts a plurality of transmission information including the first transmission information using priority, and sorts transmission information having the same priority using risk information. , may rank the first outgoing information. However, the ranking method performed by the ranking unit 13 is not limited to the example described above, and the ranking unit 13 may perform ranking by other methods.
  • step S107 In step S ⁇ b>107 , the output unit 15 ⁇ /b>A transmits information indicating the ranking result (priority order) by the ranking unit 13 to the operator terminal 3 .
  • the output unit 15A also stores information indicating the ranking result by the ranking unit 13 in the transaction information database DB1.
  • Step S108 the operator terminal 3 receives the information from the risk management support device 1A, and based on the received information, outputs the results of ranking the plurality of transmission information to an output device such as a display.
  • FIG. 7 is a diagram showing a screen example displayed on the display by the operator terminal 3 in step S108.
  • screen sc11 includes messages msg11, msg12, msg13, .
  • Inquiries in conventional e-commerce are generally handled by contact points classified by content, or on a first-come, first-served basis.
  • the person who works as a contact person checks and investigates the contents of all the chats.
  • speeding up the response to urgent cases As a result, even an urgent case will not be dealt with until the other cases are finished, as the number of people who can respond to inquiries is limited. As a result, the satisfaction level of the customer who used the inquiry counter is lowered.
  • the risk countermeasure support device 1A acquires the first transmission information regarding the target transmission, and uses the first model M1 to calculate the target transmission content based on the first transmission information. and prioritizes the first transmission information based on the risk information of the first transmission information and the risk information of transmission information other than the first transmission information. Therefore, according to this exemplary embodiment, it is possible to preferably identify a case that should be preferentially dealt with according to the risk information of the contents of the transmission by the user. As a result, for example, it is possible to shorten the time required for online-based inquiry service, and improve customer satisfaction.
  • the first transmission information according to this exemplary embodiment includes at least one of text data, voice data, and image data representing the content of the target transmission. Therefore, according to the risk management support device 1A according to the present exemplary embodiment, it is possible to preferably specify a case that should be preferentially dealt with for calls containing at least one of text data, voice data, and image data. You can get the effect that you can.
  • the first transmission information includes information about the caller who made the target transmission. Therefore, according to the present exemplary embodiment, in the ranking of outgoing information, it is possible to perform the ranking taking into account the information about the caller who made the outgoing call.
  • the first transmission information includes transaction information indicating past transactions by the originator who made the target transmission. Therefore, according to the present exemplary embodiment, in the ranking of outgoing information, it is possible to perform ranking in consideration of the past transactions by the sender who made the outgoing call.
  • the risk coping support device 1A includes an output unit 15A that outputs the ranking result by the ranking unit 13. Therefore, the user of the risk management support device 1A can preferably identify the cases that should be treated preferentially by referring to the output results.
  • the machine-learned first model M1 and second model M2 it is possible to determine the priority using, for example, a feature amount that is not noticed by humans. Attachment accuracy can be improved. Also, consistent prioritization can be achieved as compared to human discretionary prioritization.
  • FIG. 8 is a block diagram showing the configuration of the risk handling support system 100B according to this exemplary embodiment.
  • the risk handling support system 100B includes a risk handling support device 1B, an operator terminal 3, and a user terminal 4.
  • FIG. 1B is a block diagram showing the configuration of the risk handling support system 100B according to this exemplary embodiment.
  • the risk handling support system 100B includes a risk handling support device 1B, an operator terminal 3, and a user terminal 4.
  • the control unit 10A of the risk handling support device 1B includes an acquisition unit 11, a risk information calculation unit 12, a ranking unit 13, a priority analysis unit 14A, an output unit 15A, and a learning unit 16B.
  • the risk coping support device 1B is an example of a learning device according to the present specification.
  • the storage unit 20A of the risk coping support device 1B stores teacher data TD used for learning the first model M1.
  • the acquisition unit 11 acquires teacher data TD that includes a plurality of pairs of transmission information and labels representing risk information of the transmission information.
  • the learning unit 16B learns the first model M1 using teacher data TD including a plurality of pairs of transmission information and labels representing risk information of the transmission information.
  • the teacher data TD includes, for example, the feature quantity of the past transmission information accumulated in the transaction information database DB1 and the label representing the risk information.
  • the characteristic amount of the transmitted information includes, for example, a vector representing the characteristics of the transmitted information, information indicating whether or not the first transmitted information includes a specific keyword, and the like.
  • Standard Scaler, One-Hot-Encoding, and word2vec are examples of algorithms for conversion to data representing features.
  • a label included in the training data TD is a label that represents risk information, and as an example, a label that indicates whether or not an inquiry has actually been made by the user in the past.
  • the label may be input by an EC site operator or the like, or may indicate the result of categorization by classifying past transaction data by cluster analysis.
  • the operator of the EC site uses the operator terminal 3 to enter the label of the transmitted information based on the presence or absence of inquiries about past transactions, the frequency of inquiries, etc. do.
  • the operator terminal 3 transmits the label input by the operator to the risk handling support device 1B, and the risk handling support device 1B uses the set of the label and the transmission information received from the operator terminal 3 as teacher data TD.
  • the learning unit 16B classifies the past transmission information stored in the transaction information database DB1 by cluster analysis and categorizes. By cluster analysis, the transmitted information and the label indicating the classification result are linked.
  • the risk coping support device 1B re-learns the first model M1 based on the feedback information.
  • the acquisition unit 11 further acquires feedback information for the transmission information regarding the target transmission.
  • the feedback information is information about the content of the user's feedback on the transmitted information, and for example, information indicating the results of the user's questionnaire about transactions.
  • Feedback information is collected, for example, as follows.
  • the risk handling support device 1B transmits a request for a questionnaire regarding the response to the inquiry to the user terminal 4.
  • FIG. The user uses the user terminal 4 to answer the questionnaire, and the user terminal 4 transmits information indicating the results of the questionnaire to the risk countermeasure support device 1B.
  • the risk countermeasure support device 1B accumulates feedback information indicating the results of the questionnaire in the transaction information database DB1.
  • the risk coping support device 1B re-learns the first model M1 by referring to the transmission information and the feedback information regarding the target transmission. More specifically, as an example, when the feedback information is information indicating product return, the risk handling support device 1B displays transmission information corresponding to the feedback information and risk information corresponding to product return. Add pairs with labels to training data. Further, as an example, if the feedback information is information indicating dissatisfaction with the response to the inquiry, the risk handling support device 1B may send transmission information corresponding to the feedback information and a label indicating that the risk level is high. are added to the training data.
  • the risk coping support device 1B updates the teacher data based on the feedback information, and re-learns the first model M1 using the updated teacher data.
  • the method of updating teacher data is not limited to the example described above.
  • the risk coping support device 1B may update the teacher data by updating the labels included in the existing teacher data based on the feedback information.
  • the risk management support device 1B learns the first model M1 using training data including a plurality of sets of transmission information and labels representing risk information of the transmission information. As a result, according to this exemplary embodiment, it is possible to preferably identify the case that should be preferentially dealt with using the learned first model M1.
  • the risk coping support device 1B re-learns the first model M1 by referring to the transmission information and the feedback information regarding the target transmission. Since the training data is recreated based on the user's feedback, it is possible to build a more accurate model by repeating this process. Thus, according to the present exemplary embodiment, it is possible to rank outgoing information more accurately by using a re-learned model.
  • FIG. 9 is a block diagram showing the configuration of a risk handling support system 100C according to this exemplary embodiment.
  • the risk handling support system 100C includes a risk handling support device 1C, an operator terminal 3, and a user terminal 4.
  • FIG. 9 is a block diagram showing the configuration of a risk handling support system 100C according to this exemplary embodiment.
  • the risk handling support system 100C includes a risk handling support device 1C, an operator terminal 3, and a user terminal 4.
  • FIG. 9 is a block diagram showing the configuration of a risk handling support system 100C according to this exemplary embodiment.
  • the risk handling support system 100C includes a risk handling support device 1C, an operator terminal 3, and a user terminal 4.
  • FIG. 9 is a block diagram showing the configuration of a risk handling support system 100C according to this exemplary embodiment.
  • the risk handling support system 100C includes a risk handling support device 1C, an operator terminal 3, and a user terminal 4.
  • FIG. 9 is a block diagram showing the configuration of a risk handling support system 100C according to
  • FIG. 9 is a block diagram showing the configuration of a risk coping support device 1C according to this exemplary embodiment.
  • the risk handling support device 1C includes an acquisition unit 11, a risk information calculation unit 12, a ranking unit 13, a priority analysis unit 14A, and a response unit 17C.
  • the response unit 17C sequentially automatically responds to the target transmission information and the transmission information other than the target transmission information.
  • the response unit 17C automatically responds in the order according to the ranking result by the ranking unit 13 in the chat application that received the transmission information representing the inquiry.
  • the ranking unit 13 ranked the first outgoing information with reference to both risk information and priority.
  • the method of ranking the first transmission information by the ranking unit 13 is not limited to the above example, and the ranking unit 13 may rank the first transmission information by other methods.
  • the ranking unit 13 may rank the first transmission information using only the priority without referring to the risk information.
  • the ranking unit 13 may rank the first transmission information using only the risk information without referring to the priority.
  • the risk information calculation unit 12 calculated the risk information of the first transmission information when the acquisition unit 11 received the first transmission information.
  • the timing at which the risk information calculator 12 calculates risk information is not limited to that shown in the exemplary embodiment described above.
  • the risk information calculation unit 12 may calculate the risk information using the first model M1 for the first transmission information acquired at that time every time a certain period of time elapses.
  • the learning unit 16B may learn the second model M2.
  • the teacher data related to the learning of the second model M2 includes, as an example, a plurality of pairs of the feature amount of the transmitted information and the label representing the priority of the transmitted information.
  • a label representing a priority is input by an EC site operator or the like, for example.
  • the learning unit 16B constructs the second model M2 by machine learning using the teacher data.
  • the weighting of the second model M2 is carried out by setting targets for risk reduction by humans based on accumulated past data statistical information, etc., and considering the optimal balance. For example, to satisfy the goal of maximizing customer satisfaction with fast response times and the goal of avoiding potential litigation and avoiding increased costs, statistical information from historical data can be used to and customer satisfaction, the number of lawsuits per all cases and their costs, evaluate each with the same dimensional index, add them up, and apply the weight that maximizes the value to the second model M2. do.
  • ⁇ Modification 4> a system configuration in which the risk handling support device 1A, 1B or 1C (hereinafter referred to as "risk handling support device 1A etc.") and the user terminal 4 are separate entities has been described.
  • the system configuration is not limited to the example described above, and may be other configurations.
  • an application that realizes the functions of the above-described risk management support device 1A and the like may be installed in the user terminal 4, and the user terminal 4 may operate alone.
  • the user terminal 4 is an example of the risk handling support device according to the present specification, and the user terminal 4 includes the above-described acquisition unit 11, risk information calculation unit 12, ranking unit 13, priority analysis unit 14A, and the like. Prepare.
  • the transmission information is not limited to the information related to the EC site, and may be other information. good too.
  • the transmission information may be, for example, text data, audio data, or image data transmitted by the SNS user.
  • by ranking transmission information transmitted on the SNS based on the risk information it is possible to preferably specify a case to be preferentially handled by the operator of the SNS or the like.
  • risk handling support device 1, etc. Some or all of the functions of the risk handling support devices 1, 1A, 1B, 1C, and the learning device 2 (hereinafter referred to as "risk handling support device 1, etc.") are implemented by hardware such as integrated circuits (IC chips). may be implemented by software.
  • the risk management support device 1 and the like are implemented by a computer that executes program instructions, which are software that implements each function, for example.
  • An example of such a computer (hereinafter referred to as computer C) is shown in FIG.
  • Computer C comprises at least one processor C1 and at least one memory C2.
  • a program P for operating the computer C as the risk coping support device 1 or the like is recorded in the memory C2.
  • the processor C1 reads the program P from the memory C2 and executes it, thereby realizing each function of the risk coping support device 1 and the like.
  • processor C1 for example, CPU (Central Processing Unit), GPU (Graphic Processing Unit), DSP (Digital Signal Processor), MPU (Micro Processing Unit), FPU (Floating point number Processing Unit), PPU (Physics Processing Unit) , a microcontroller, or a combination thereof.
  • memory C2 for example, a flash memory, HDD (Hard Disk Drive), SSD (Solid State Drive), or a combination thereof can be used.
  • the computer C may further include a RAM (Random Access Memory) for expanding the program P during execution and temporarily storing various data.
  • Computer C may further include a communication interface for transmitting and receiving data to and from other devices.
  • Computer C may further include an input/output interface for connecting input/output devices such as a keyboard, mouse, display, and printer.
  • the program P can be recorded on a non-temporary tangible recording medium M that is readable by the computer C.
  • a recording medium M for example, a tape, disk, card, semiconductor memory, programmable logic circuit, or the like can be used.
  • the computer C can acquire the program P via such a recording medium M.
  • the program P can be transmitted via a transmission medium.
  • a transmission medium for example, a communication network or broadcast waves can be used.
  • Computer C can also obtain program P via such a transmission medium.
  • a risk countermeasure support device comprising a prioritizing means for prioritizing the first transmission information in the order in which countermeasures should be taken, based on the risk information of the transmission information other than the transmission information.
  • Appendix 2 The risk handling support device according to appendix 1, wherein the first transmission information includes at least one of text data, voice data, and image data representing the content of the transmission of the target.
  • appendix 6 Any one of appendices 1 to 4, further comprising response means for sequentially and automatically responding to the target transmitted information and to the transmitted information other than the target transmitted information according to the ranking result of the ranking means.
  • Appendix 7 The risk handling support device according to any one of Appendices 1 to 6, further comprising learning means for learning the model using teacher data including a plurality of sets of transmitted information and labels representing risk information of the transmitted information. .
  • the acquisition means further acquires feedback information for the transmission information regarding the transmission of the target, and the risk coping support device refers to the transmission information regarding the transmission of the target and the feedback information to re-learn the model. , supplementary note 7.
  • (Appendix 9) Acquisition means for acquiring teacher data including a plurality of pairs of transmitted information and labels representing risk information of the transmitted information; and receiving the transmitted information as input and outputting risk information used for ranking the transmitted information.
  • a learning device comprising learning means for learning a model using the teacher data.
  • Appendix 14 an acquisition means for acquiring first transmission information relating to a target transmission; Based on the first transmission information acquired by the acquisition means using a learned model that receives the second transmission information and outputs risk information indicating the degree to which the second transmission information should be dealt with a risk information calculation means for calculating risk information of the transmission content of the target; a prioritizing means for prioritizing the first transmission information in the order in which it should be dealt with, based on the risk information of the first transmission information and the risk information of transmission information other than the first transmission information; including risk coping support system.
  • At least one processor is provided, and the processor receives as input a first transmission information relating to a target transmission and a second transmission information, and indicates the degree to which the second transmission information should be dealt with.
  • a risk information calculation process for calculating risk information of the transmission content of the target based on the first transmission information acquired in the acquisition process using a learned model that outputs risk information;
  • Risk countermeasures for executing a ranking process for prioritizing the first transmission information in the order in which it should be handled, based on the risk information of the transmission information and the risk information of the transmission information other than the first transmission information. support equipment.
  • the risk management support device may further include a memory, and the memory stores a program for causing the processor to execute the acquisition process, the risk information calculation process, and the ranking process. may have been Also, this program may be recorded in a computer-readable non-temporary tangible recording medium.
  • At least one processor is provided, and the processor performs acquisition processing for acquiring teacher data including a plurality of pairs of transmission information and labels representing risk information of the transmission information, and inputting the transmission information and ranking the transmission information. and a learning process of learning a model that outputs risk information used for attaching the data using the teacher data.
  • the learning device may further include a memory, and the memory may store a program for causing the processor to execute the acquisition process and the learning process. Also, this program may be recorded in a computer-readable non-temporary tangible recording medium.
  • At least one of the functions of the risk coping support devices 1, 1A, 1B, and 1C and the learning device 2 described above can be performed by a plurality of different information processing devices installed and connected anywhere on the network. It may also be performed, ie on so-called cloud computing.

Landscapes

  • Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Health & Medical Sciences (AREA)
  • Economics (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

Afin de spécifier de manière appropriée un cas à traiter de manière préférentielle en fonction d'informations de risque de détails de transmission par un utilisateur, un dispositif d'aide à la contre-mesure de risque (1) comprend : un moyen d'acquisition (11) pour acquérir des premières informations de transmission concernant la transmission par une cible ; un moyen de calcul d'informations de risque (12) pour calculer, à l'aide d'un modèle formé pour recevoir une entrée de secondes informations de transmission et fournir en sortie des informations de risque indiquant le degré d'exigence de contre-mesure par rapport aux secondes informations de transmission, des informations de risque des détails de la transmission par la cible sur la base des premières informations de transmission acquises par le moyen d'acquisition (11) ; et un moyen de classement (13) pour fournir un classement de priorité aux premières informations de transmission dans un ordre d'exigence de contre-mesure sur la base d'informations de risque des premières informations de transmission et d'informations de risque d'informations de transmission autres que les premières informations de transmission.
PCT/JP2022/002987 2022-01-27 2022-01-27 Dispositif d'aide à la contre-mesure de risque, dispositif d'apprentissage, procédé d'aide à la contre-mesure de risque, procédé d'apprentissage et programme WO2023144949A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/JP2022/002987 WO2023144949A1 (fr) 2022-01-27 2022-01-27 Dispositif d'aide à la contre-mesure de risque, dispositif d'apprentissage, procédé d'aide à la contre-mesure de risque, procédé d'apprentissage et programme

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2022/002987 WO2023144949A1 (fr) 2022-01-27 2022-01-27 Dispositif d'aide à la contre-mesure de risque, dispositif d'apprentissage, procédé d'aide à la contre-mesure de risque, procédé d'apprentissage et programme

Publications (1)

Publication Number Publication Date
WO2023144949A1 true WO2023144949A1 (fr) 2023-08-03

Family

ID=87471208

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2022/002987 WO2023144949A1 (fr) 2022-01-27 2022-01-27 Dispositif d'aide à la contre-mesure de risque, dispositif d'apprentissage, procédé d'aide à la contre-mesure de risque, procédé d'apprentissage et programme

Country Status (1)

Country Link
WO (1) WO2023144949A1 (fr)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2008210367A (ja) * 2007-01-29 2008-09-11 Nec Corp リスク検知システム、リスク検知方法及びそのプログラム
JP2009043144A (ja) * 2007-08-10 2009-02-26 Internatl Business Mach Corp <Ibm> 電子メールメッセージの特性を検出する装置及び方法
JP2017199254A (ja) * 2016-04-28 2017-11-02 日本電気株式会社 会話分析装置、会話分析方法および会話分析プログラム
JP2018092578A (ja) * 2016-11-30 2018-06-14 富士通株式会社 サイバーブーリング防止のための方法、モバイル電子デバイス、及び非一時的なコンピュータ可読媒体

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2008210367A (ja) * 2007-01-29 2008-09-11 Nec Corp リスク検知システム、リスク検知方法及びそのプログラム
JP2009043144A (ja) * 2007-08-10 2009-02-26 Internatl Business Mach Corp <Ibm> 電子メールメッセージの特性を検出する装置及び方法
JP2017199254A (ja) * 2016-04-28 2017-11-02 日本電気株式会社 会話分析装置、会話分析方法および会話分析プログラム
JP2018092578A (ja) * 2016-11-30 2018-06-14 富士通株式会社 サイバーブーリング防止のための方法、モバイル電子デバイス、及び非一時的なコンピュータ可読媒体

Similar Documents

Publication Publication Date Title
US10268653B2 (en) Goal-oriented user matching among social networking environments
US10162884B2 (en) System and method for auto-suggesting responses based on social conversational contents in customer care services
CN108280670B (zh) 种子人群扩散方法、装置以及信息投放系统
CN110995459B (zh) 异常对象识别方法、装置、介质及电子设备
TWI793412B (zh) 消費預測系統及消費預測方法
WO2020110664A1 (fr) Procédé de génération d&#39;un modèle de prédiction de réception d&#39;ordre, modèle de prédiction de réception d&#39;ordre, dispositif de prédiction de réception d&#39;ordre, procédé de prédiction de réception d&#39;ordre et programme de prédiction de réception d&#39;ordre
US11842156B2 (en) Systems and methods of artificially intelligent sentiment analysis
CN112215448A (zh) 分配客服的方法和装置
JP6611068B1 (ja) 企業情報処理装置、企業のイベント予測方法及び予測プログラム
WO2020087828A1 (fr) Procédé et appareil d&#39;évaluation de risque d&#39;avant vente, appareil informatique et support de stockage lisible
CN112634062B (zh) 基于Hadoop的数据处理方法、装置、设备及存储介质
JPWO2014068745A1 (ja) メッセージ一元管理システム
WO2023144949A1 (fr) Dispositif d&#39;aide à la contre-mesure de risque, dispositif d&#39;apprentissage, procédé d&#39;aide à la contre-mesure de risque, procédé d&#39;apprentissage et programme
CN110929144A (zh) 一种业务数据管理方法、系统和可读存储介质
CN113987186B (zh) 一种基于知识图谱生成营销方案的方法和装置
CN115936748A (zh) 一种商业大数据分析方法及系统
CN115293291A (zh) 排序模型的训练方法、排序方法、装置、电子设备及介质
CN115689143A (zh) 工单分派方法、装置、电子设备及介质
US20200286104A1 (en) Platform for In-Memory Analysis of Network Data Applied to Profitability Modeling with Current Market Information
CN114298825A (zh) 还款积极度评估方法及装置
CN111882339A (zh) 预测模型训练及响应率预测方法、装置、设备及存储介质
Chu et al. Enhancing the customer service experience in call centers using preemptive solutions and queuing theory
WO2023119460A1 (fr) Système de prédiction de question, procédé de prédiction de question et support d&#39;enregistrement de programme
JP7537815B1 (ja) 情報処理システム
KR102666173B1 (ko) 금융 소비자 맞춤형 공시 분석 정보 제공 장치 및 방법

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22923815

Country of ref document: EP

Kind code of ref document: A1