JP7096277B2 - Prediction systems, prediction methods and programs - Google Patents

Description

The present invention relates to a prediction system, a prediction method and a program.

A call center (or contact center) that accepts inquiries about products and services sold or provided by companies and the like is known. In such a call center, for example, an operator responds to an inquiry from a customer via a telephone by referring to a manual or the like.

By the way, there are cases where the operator who has received an inquiry from a customer has little knowledge and cannot give an appropriate answer. In such a case, a technique capable of transferring an inquiry to an operator who can handle a specialized inquiry or a complicated inquiry has been conventionally known (see, for example, Patent Document 1).

Japanese Unexamined Patent Publication No. 2010-109895

However, in the above-mentioned prior art, for example, the operator who receives the transfer of the inquiry needs to ask the customer for more detailed inquiry contents. For this reason, the customer has to wait until an answer is obtained, and for example, there is a risk that customer satisfaction may decrease.

On the other hand, if the content of an inquiry from a customer can be predicted in advance, it will be possible to obtain an appropriate answer in advance, and it will be possible to answer without making the customer wait.

One embodiment of the present invention has been made in view of the above points, and an object thereof is to predict the content of an inquiry.

In order to achieve the above object, the prediction system according to the embodiment is a prediction system including a terminal used by a customer and a prediction device for predicting the content of an inquiry from the customer, and the prediction device is the above-mentioned prediction device. A model for creating a prediction model for predicting the content of an inquiry from a customer by using the first behavior history information indicating the customer's behavior history and the inquiry information associated with the first behavior history information. When the customer makes a call to the call center with the terminal, the creation means, the second action history information indicating the action history of the customer before the call is made, and the prediction model, the customer on the phone. It is characterized by having a prediction means for predicting the contents of the inquiry.

The content of the inquiry can be predicted.

It is a figure which shows an example of the whole structure of the inquiry prediction system which concerns on this embodiment. It is a figure for demonstrating an example of the screen transition of a Web page. It is a figure which shows an example of the functional structure of the inquiry prediction system which concerns on this embodiment. It is a figure which shows an example of the action history information. It is a figure which shows an example of inquiry information. It is a flowchart which shows an example of the creation process of the prediction template which concerns on this embodiment. It is a figure for demonstrating an example of making a prediction template. It is a sequence diagram which shows an example of the inquiry prediction processing which concerns on this embodiment. It is a figure for demonstrating an example of inquiry prediction. It is a figure which shows an example of an operator screen.

Hereinafter, an embodiment of the present invention will be described. In the present embodiment, when a customer makes an inquiry about a product, a service, or the like into a call center, the inquiry prediction system 1 capable of predicting the contents of the inquiry will be described. Here, in the present embodiment, the customer makes a telephone call to the call center by selecting a link for calling a telephone number on a Web page related to a product or service using a smartphone or the like. It should be noted that such a link for calling a telephone number can be realized by an anchor tag (<a> tag) of HTML (HyperText Markup Language).

<Overall configuration>
First, the overall configuration of the inquiry prediction system 1 according to the present embodiment will be described with reference to FIG. FIG. 1 is a diagram showing an example of the overall configuration of the inquiry prediction system 1 according to the present embodiment.

As shown in FIG. 1, the inquiry prediction system 1 according to the present embodiment includes an inquiry prediction device 10, one or more customer terminals 20, one or more operator terminals 30, one or more exchanges 40, and one or more. Web server 50 and the above are included. Of these, the inquiry prediction device 10, each operator terminal 30, and the exchange 40 are installed in the system environment in the call center.

Further, each customer terminal 20, the inquiry prediction device 10, and each Web server 50 are connected to the Internet, and each customer terminal 20 and the exchange 40 are connected to a telephone line network. Further, each operator terminal 30 and the exchange 40 are connected to the premises telephone network in the call center, and each operator terminal 30 and the inquiry prediction device 10 are connected to a communication network (for example, LAN (Local Area Network)) in the call center. ing.

The inquiry prediction device 10 is a computer or a computer system that predicts the contents of a customer's inquiry by using the customer's behavior history (for example, the access history of a Web page).

The customer terminal 20 is a terminal used by the customer. As the customer terminal 20, a computer having at least a telephone function for making a call with the operator terminal 30 and a Web page browsing function for browsing a Web page provided by the Web server 50 can be used. Specifically, as the customer terminal 20, for example, a smartphone, a tablet terminal, a wearable device, a PC (personal computer), or the like can be used.

The operator terminal 30 is a terminal used by the operator. As the operator terminal 30, a computer having a telephone function for making a call with the customer terminal 20 and a display function for displaying the inquiry content predicted by the inquiry prediction device 10 can be used. Specifically, as the operator terminal 30, for example, a PC (personal computer), a tablet terminal, a smartphone, a wearable device, or the like can be used.

The exchange 40 is a PBX (Private Branch Exchange) installed in the call center, and connects an incoming call from the customer terminal 20 to the operator terminal 30. The exchange 40 may be a PBX having VoIP (Voice over Internet Protocol).

The Web server 50 is a computer or a computer system that provides a Web page related to a product or service to be inquired. In addition, this Web page also includes a link for calling a telephone number for making a call to the call center to be inquired about.

Here, the web page in which a company or the like introduces some product or service has a hierarchical structure, and the web page related to the desired product or service can be browsed by following the links in order. In many cases. For example, the example shown in FIG. 2 is a case where the Web page has a two-layer hierarchical structure, and the viewer can access the Web page related to the desired product (“loan” in the example shown in FIG. 2) on the TOP screen. By pressing the link, you can browse the Web page related to the product. For example, when a viewer wants to know about a mortgage, the viewer can move to the mortgage screen that provides information about the mortgage by clicking the link to the mortgage screen on the TOP screen. Similarly, for example, when a viewer wants to know about an education loan, the viewer can move to the education loan screen that provides information about the education loan by clicking the link to the education loan screen on the TOP screen.

Further, in the example shown in FIG. 2, a telephone number calling link "0120-xxxxxx-xxxxx" is included in each of the mortgage screen, the education loan screen, and the private car loan screen. By pressing this telephone number calling link, the viewer can make a call to the telephone number specified for the link (that is, the telephone number of the call center). Therefore, for example, when there is an unclear point about the product, the viewer can make an inquiry to the call center by pressing the telephone number calling link.

The Web page and its screen transition shown in FIG. 2 are an example, and a company or the like that sells or provides a product, a service, or the like can arbitrarily configure a Web page and its screen transition. For example, instead of providing a link for calling a telephone number on the mortgage screen, the education loan screen, and the private car loan screen, a link to the inquiry screen may be provided and a link for calling a telephone number may be provided on this inquiry screen. ..

<Functional configuration>
Next, the functional configuration of the inquiry prediction system 1 according to the present embodiment will be described with reference to FIG. FIG. 3 is a diagram showing an example of the functional configuration of the inquiry prediction system 1 according to the present embodiment.

≪Customer terminal 20≫
As shown in FIG. 3, the customer terminal 20 according to the present embodiment has a Web browsing unit 201 and a telephone function unit 202. Each of these parts is realized, for example, by a process in which one or more programs installed in the customer terminal 20 cause the processor to execute. In particular, the Web browsing unit 201 is realized by a process that the Web browser causes the processor to execute.

The Web browsing unit 201 displays the Web page provided by the Web server 50 and accepts user operations and the like on the Web page. Further, the Web browsing unit 201 transmits Cookie and Key information to the inquiry prediction device 10 when the telephone function unit 202 makes a telephone call in response to the pressing of the telephone transmission link. Here, the Key information is arbitrary information that can uniquely identify the customer who made the telephone call, and is generated by the Web browsing unit 201. It should be noted that the cookie can identify the customer who uses the customer terminal 20 (more accurately, the Web browser used by the customer of the customer terminal 20). However, Cookie is an example, and any information other than Cookie can be used as long as it can identify the customer who accessed the Web page and the access history of the Web page accessed by this customer can be acquired from the Web server 50. Information may be used.

The telephone function unit 202 makes a call to the telephone number specified in this link in response to pressing the telephone call link on the Web page.

≪Web server 50≫
As shown in FIG. 3, the Web server 50 according to the present embodiment has a history providing unit 501. The history providing unit 501 is realized, for example, by a process of causing a processor to execute one or more programs installed in the Web server 50.

The history providing unit 501 provides the access history of the Web page of the corresponding customer in response to the request from the inquiry prediction device 10. Here, the request from the inquiry prediction device 10 includes a cookie ID, and the history providing unit 501 transmits the access history corresponding to this cookie ID to the inquiry prediction device 10. The access history includes at least a Cookie ID, a URL (Uniform Resource Locator) of a Web page accessed by a customer corresponding to the Cookie ID, and the access date and time. The cookie ID is identification information included in the cookie.

<< Inquiry prediction device 10 >>
As shown in FIG. 3, the inquiry prediction device 10 according to the present embodiment includes a collection unit 101, an analysis unit 102, a response unit 103, and a screen providing unit 104. Each of these parts is realized, for example, by a process of causing a processor to execute one or more programs installed in the query prediction device 10.

Further, the inquiry prediction device 10 according to the present embodiment has an action history information storage unit 105 and an inquiry information storage unit 106. Each of these parts can be realized by using, for example, a storage device included in the query prediction device 10. At least one of the action history information storage unit 105 and the inquiry information storage unit 106 may be realized by using, for example, a storage device connected to the inquiry prediction device 10 via a communication network or the like.

When the collecting unit 101 receives the cookie and the key information from the customer terminal 20, it transmits an access history acquisition request including the cookie ID of the cookie to the Web server 50. As a result, the access history is acquired from the Web server 50.

Then, the collecting unit 101 creates the action history information using the access history and the key information, and stores the action history information in the action history information storage unit 105. The action history information is, for example, information calculated from the access history of the customer's Web page, and includes Key information, the URL of the Web page, page information, the total staying time of this Web page, the number of visits, and the like. It is information. The details of the action history information will be described later.

The analysis unit 102 predicts the contents of the customer's inquiry by using the behavior history information created by the collection unit 101 and the prediction template created in advance. Here, the prediction template is a model for predicting the inquiry content from the action history information, and is created by using the action history information and the inquiry information corresponding to the action history information as learning data. The inquiry information is information representing the inquiry content, and includes Key information and information representing the inquiry content. The details of the method of creating the prediction template will be described later.

Further, the analysis unit 102 has the action history information created by the collection unit 101 and, for example, inquiry information indicating the customer's response result to the voice response guidance by the response unit 103 or inquiry information indicating the inquiry content actually received by the operator. Update the prediction template using and.

The response unit 103 uses the inquiry content predicted by the analysis unit 102 to provide voice response guidance to the customer terminal 20 that made the call. That is, the response unit 103 provides, for example, voice response guidance for the customer who uses the customer terminal 20 to confirm whether the inquiry content predicted by the analysis unit 102 is correct.

The screen providing unit 104 provides the operator terminal 30 that makes a call with the customer terminal 20 with information necessary for displaying a predetermined operator screen or the operator screen. Here, the operator screen includes at least the inquiry content predicted by the analysis unit 102. As a result, the operator can know the content of the inquiry from the customer in advance.

The action history information storage unit 105 stores the action history information. The details of the action history information will be described later.

The inquiry information storage unit 106 stores inquiry information. The details of the inquiry information will be described later.

<< Operator terminal 30 >>
As shown in FIG. 3, the operator terminal 30 according to the present embodiment has a display function unit 301 and a telephone function unit 302. Each of these parts is realized, for example, by a process in which one or more programs installed in the operator terminal 30 cause the processor to execute.

The display function unit 301 displays the operator screen. The telephone function unit 302 makes a call with the customer terminal 20 via the exchange 40. In addition to displaying the operator screen, the display function unit 301 may accept various operations of the operator on the operator screen.

≪Behavior history information≫
Here, the behavior history information stored in the behavior history information storage unit 105 will be described with reference to FIG. FIG. 4 is a diagram showing an example of behavior history information.

As shown in FIG. 4, the action history information includes Key information, Cookie ID, URL, transition source URL, page information, staying time, number of visits, and the like.

The key information is the key information transmitted from the customer terminal 20. The action history information and the inquiry information are associated with each other by this key information.

The cookie ID is a cookie ID included in the cookie transmitted from the customer terminal 20. The URL is the URL of the Web page accessed by the customer corresponding to this Cookie ID. The transition source URL is the URL of the transition source Web page of the Web page of the URL. The page information is the page information of the Web page accessed by the customer corresponding to the Cookie ID (for example, the title of the Web page). The staying time is the total time that the customer stayed on the Web page. The number of visits is the total number of visits by the customer to the Web page.

As described above, the action history information is information calculated (for example, aggregation, etc.) from the access history acquired from the Web server 50 for each customer (that is, for each Cookie ID). Further, in order to associate with the inquiry information described later, the action history information also includes Key information for identifying the customer. The behavior history information shown in FIG. 4 is an example, and even if the behavior history information includes various information (for example, visit frequency) that can be calculated from the access history in addition to the above-mentioned information. good.

≪Inquiry information≫
Next, the inquiry information stored in the inquiry information storage unit 106 will be described with reference to FIG. FIG. 5 is a diagram showing an example of inquiry information.

As shown in FIG. 5, the inquiry information includes Key information, inquiry contents, and the like. As described above, the key information is the key information transmitted from the customer terminal 20. The inquiry content is, for example, the inquiry content representing the response result of the customer to the voice response guidance by the response unit 103, or the inquiry content actually received by the operator.

As described above, the inquiry information is information including Key information and the actual inquiry contents of the customer. In the example shown in FIG. 5, "new mortgage loan" is set as the content of the inquiry. This indicates that the inquiry is for a new application for a mortgage. Other inquiries regarding mortgages include, for example, "mortgage cancellation" indicating that the inquiry is related to mortgage cancellation, and "mortgage renewal" indicating that the inquiry is related to refinancing the mortgage. it is conceivable that.

<Process executed by query prediction system 1>
Next, the process executed by the inquiry prediction system 1 according to the present embodiment will be described.

≪Process to create prediction template≫
First, assuming that a certain amount of action history information and inquiry information are stored in the action history information storage unit 105 and the inquiry information storage unit 106, these action history information and inquiry information are used as learning data. A case of creating a prediction template will be described with reference to FIG. FIG. 6 is a flowchart showing an example of a prediction template creation process according to the present embodiment.

First, the analysis unit 102 reads the behavior history information and the inquiry information as learning data from the behavior history information storage unit 105 and the inquiry information storage unit 106 (step S101). At this time, the analysis unit 102 reads the action history information and the inquiry information including the same Key information as one learning data. That is, the learning data is a set of action history information and inquiry information including the same Key information.

Next, the analysis unit 102 creates a prediction template for each inquiry content using the learning data read in step S101 above (step S102). Here, when creating a prediction template for a certain inquiry content, the analysis unit 102 creates a prediction template according to the following Step 1-1 to Step 1-3. In the following, as an example, a case of creating a prediction template for the inquiry content “new mortgage” will be described.

Step1-1) First, the analysis unit 102 extracts the learning data whose inquiry content is "new housing loan" from the read learning data. In the following, as an example, in Step _1-1 , learning data a1 to a3 including key information indicating Mr. _A , learning data b1 to b2 including key information indicating Mr. _{B ,} and Mr. C are used. It is assumed that a total of 7 learning data including the learning data c ₁ to c ₂ including the key information shown are extracted.

Step1-2) Next, the analysis unit 102 plots each learning data extracted in Step1-1 above for each URL of the same Web page in a space centered on the staying time and the number of visits, and plots these. Calculate the center of gravity and weight of the points.

For example, when the learning data including the URL of the TOP screen is a ₁ , b ₁ , c ₁ , the analysis unit 102 has the learning data a ₁ , b ₁ , c as shown in FIG. 7 (a). ₁ is plotted in a space centered on the staying time and the number of visits (that is, the staying time-visiting number plane). Then, the analysis unit 102 calculates the center of gravity g ₁ of these learning data a ₁ , b ₁ , and c ₁ and the weight w ₁ thereof. The weight w ₁ is the percentage of people who have visited the TOP screen. In the case of the example shown in FIG. 7A, since all of Mr. A, Mr. B, and Mr. C are visiting the TOP screen (accessing the TOP screen), the weight w ₁ is 1.0.

Similarly, for example, when the learning data including the URL of the mortgage screen is _a2 , _b2 , c2, the analysis unit 102 uses the learning data _a2 , as shown in _FIG . 7B. b ₂ and c ₂ are plotted in a space centered on the length of stay and the number of visits. Then, the analysis unit 102 calculates the center of gravity g ₂ of these learning data a ₂ , b ₂ , and c ₂ and the weight w ₂ thereof. In the case of the example shown in FIG. 7B, since all of Mr. A, Mr. B, and Mr. C are visiting the mortgage screen, the weight w ₂ is 1.0.

Similarly, for example, when the learning data including the URL of the education loan screen is a3 _, the analysis unit 102 uses the learning data _a3 for the staying time and the number of visits as shown in FIG. 7C. Plot in space as the axis. Then, the analysis unit 102 calculates the center of gravity g ₃ of the learning data a ₃ and its weight w ₃ . In the case of the example shown in FIG. 7 (c), since only Mr. A, Mr. A, Mr. B, and Mr. C are visiting the education loan screen, the weight w ₃ is 1/3 ≈ 0.3. Will be.

In the above, the axes of the space for plotting the learning data are the staying time and the number of visits, but these axes can be arbitrarily selected. In particular, three or more axes may be selected. Further, the space for plotting the training data may be normalized so that the values that can be taken in each axis are 0 or more and 1 or less, for example. In this case, when plotting the training data in the space, it is necessary to normalize the values included in the training data before plotting.

Step1-3) Then, for each inquiry content, the analysis unit 102 uses a set of a set of the center of gravity and weight calculated in Step1-2 above and the URL included in the learning data used to calculate the center of gravity as a prediction template. do. For example, the URL of the TOP screen is m ₁ , the URL of the mortgage screen is m ₂ , and the URL of the education loan screen is m _3. Regarding the inquiry content "new mortgage", the center of gravity g ₁ to g shown in FIG. 7 (d). When ₃ and these weights w ₁ to w ₃ are obtained, the analysis unit 102 determines {(g ₁ , w ₁ , m ₁ ), (g ₂ , w ₂ , m ₂ ), (g ₃ , w). ₃ , m ₃ )} is used as the prediction template for the inquiry content "new mortgage". In addition, m ₁ to m ₃ may be an ID or the like that can identify the screen of the Web page instead of the URL.

As described above, the inquiry prediction device 10 according to the present embodiment creates a prediction template for the inquiry content for each inquiry content. These prediction templates are stored in, for example, a storage device of the inquiry prediction device 10.

≪Inquiry prediction processing≫
Next, a case of predicting the content of an inquiry from a customer using a prediction template will be described with reference to FIG. FIG. 8 is a sequence diagram showing an example of the query prediction process according to the present embodiment.

The Web browsing unit 201 of the customer terminal 20 browses the Web page provided by the Web server 50 (step S201). Here, it is assumed that the customer who uses the customer terminal 20 presses the telephone call link on the Web page and performs the call call operation to the telephone number specified in the link.

The telephone function unit 202 of the customer terminal 20 adds key information generated by the Web browsing unit 201 to the end of the telephone number specified in the telephone number calling link (for example, adding a comma after the telephone number). After that, Key information is added) to make a telephone call (step S202). As the key information, any character string can be adopted as long as it is a character string that can be added to the end of the telephone number. For example, the Cookie ID, the login ID of the Web page, or the date and time for these information. It is possible to use information or the like to which information is added. Key information can be generated, for example, by executing Javascript (registered trademark) acquired from the Web server 50 by the Web browsing unit 201 .

Further, when the telephone transmission in step S202 is performed, the Web browsing unit 201 of the customer terminal 20 transmits Cookie and Key information to the inquiry prediction device 10 (step S203).

When the collecting unit 101 of the inquiry prediction device 10 receives the cookie and the key information from the customer terminal 20, it transmits an access history acquisition request including the cookie ID of the cookie to the Web server 50 (step S204). If the collection unit 101 uses, for example, the Web server 50 corresponding to the domain or path included in the cookie (that is, the domain or path to which the cookie is issued) as the Web server 50 to which the access history acquisition request is transmitted. good.

When the history providing unit 501 of the Web server 50 receives the access history acquisition request from the inquiry prediction device 10, the history providing unit 501 acquires the access history corresponding to the Cookie ID included in the access history acquisition request, and the acquired access history is used as the inquiry prediction device. It is transmitted to 10 (step S205). As described above, the access history includes at least the Cookie ID, the URL of the Web page accessed by the customer corresponding to the Cookie ID, and the access date and time.

When the collection unit 101 of the inquiry prediction device 10 receives the access history from the Web server 50, the collection unit 101 uses the access history and the Key information transmitted from the customer terminal 20 in step S203 above for each Web page URL. The action history information is created, and the created action history information is stored in the action history information storage unit 105 (step S206). Here, for example, the collecting unit 101 analyzes the access history to calculate the staying time, the number of visits, and the like, and then creates the action history information including the staying time, the number of visits, and the key information. good. The collection unit 101 creates action history information for each URL of a Web page having substantially the same content, for example, by appropriately deleting the query parameters of the URL included in the access history.

Next, the analysis unit 102 of the inquiry prediction device 10 predicts the customer's inquiry content by using the action history information created in step S206 and the prediction template created in advance (step S207). Here, the analysis unit 102 predicts the inquiry content by the following Step2-1 to Step2-3.

Step2-1) First, the analysis unit 102 plots each action history information created in step S206 above and the prediction template in a space centered on the staying time and the number of visits for each prediction template. Hereinafter, as an example, the action history information created in step S206 is the action history information r ₁ including the URL of the TOP screen and the action history information r ₂ including the URL of the mortgage screen. And.

Step2-2) Next, the analysis unit 102 calculates the weighted distance between the point indicating each action history information and the center of gravity corresponding to this action history information for each prediction template, and then these weighted distances. Is calculated as the evaluation distance. Here, the center of gravity corresponding to the action history information means a case where the URL included in the action history information and the URL corresponding to the center of gravity are the same or these URLs point to the same Web page.

For example, let the prediction template of the inquiry content "new mortgage" be {(g ₁ , w ₁ , m ₁ ), (g ₂ , w ₂ , m ₂ ), (g ₃ , w ₃ , m ₃ )} and m. It is assumed that ₁ is the URL of the TOP screen, m ₂ is the URL of the mortgage screen, and m ₃ is the URL of the education loan screen. In this case, as shown in FIG. 9A, the analysis unit 102 has a weighted distance between the point indicating the action history information r ₁ and the center of gravity g ₁ , and a point indicating the action history information r ₂ and the center of gravity g ₂ . The sum with the weighted distance of is calculated as the evaluation distance. The weighted distance between the point indicating the action history information r ₁ and the center of gravity g ₁ is, for example, the distance obtained by dividing the Euclidean distance between the point indicating the action history information r ₁ and the center of gravity g ₁ by the weight w ₁ . Is. Similarly, the weighted distance between the point indicating the action history information r ₂ and the center of gravity g ₂ is, for example, the distance obtained by dividing the Euclidean distance between the point indicating the action history information r ₂ and the center of gravity g ₂ by the weight w ₂ . That is.

Similarly, for example, when the prediction template of the inquiry content "new education loan" is {(g ₁ ^' , w ₁ ^' , m ₁ ), (g ₃ ^' , w ₃ ^' , m ₃ )}, FIG. 9 ( As shown in b), the analysis unit 102 sets a weighted distance between the point indicating the action history information r ₁ and the center of gravity g ₁ ^′ . On the other hand, since the center of gravity corresponding to this action history information does not exist at the point indicating the action history information r ₂ , a predetermined maximum value is set as the weighted distance. Then, the analysis unit 102 uses the sum of the weighted distance between the point indicating the action history information r ₁ and the center of gravity g ₁ ^′ and the weighted distance (maximum value) of the point indicating the action history information r ₂ as the evaluation distance. calculate.

Step2-3) Then, the analysis unit 102 identifies the prediction template of the smallest evaluation distance among the evaluation distances calculated in Step2-2 above, and sets the inquiry content corresponding to the specified prediction template as the prediction result. .. As a result, the content of the inquiry is predicted from the behavior history of the customer.

In the above, the axis of the space for plotting the action history information and the prediction template uses the axis used when the prediction template was created. In addition, when the prediction template is created, if the normalization of each axis is performed, the value included in the action history information is also normalized in the same manner.

Subsequently, the response unit 103 of the inquiry prediction device 10 confirms the inquiry content predicted in the above step S207 to the customer terminal 20 that made a telephone call in the above step S202 via the exchange 40. (Step S208 to Step S209). As such a voice guidance response, for example, if the inquiry content predicted in step S207 above is "new mortgage loan", "inquiry about new application for mortgage loan. If you like, 0 is different. If so, please press 1. " In addition to this, for example, the inquiry content may be confirmed by "Yes" or "No" using voice recognition technology.

The telephone function unit 202 of the customer terminal 20 responds to the voice guidance response in step S209 (step S210). That is, the telephone function unit 202 of the customer terminal 20 has a response indicating that the inquiry content predicted in step S207 is correct (for example, when 0 is pressed) or a response indicating that the inquiry content is incorrect (for example, 1). When is pressed, etc.). As a result, the customer terminal 20 is in a state of waiting for connection with the operator terminal 30, for example.

On the other hand, the screen providing unit 104 of the inquiry prediction device 10 displays the operator screen including the inquiry content predicted in step S207 above on the corresponding operator terminal 30 (for example, the operator terminal 30 of the operator who can answer the inquiry content). ) (Step S211). Here, the screen providing unit 104 of the inquiry prediction device 10 may transmit the operator screen to the operator terminal 30, and the operator may send information necessary for displaying the operator screen (that is, a prediction result of the inquiry content, etc.). It may be transmitted to the terminal 30. As will be described later, the operator screen contains various information necessary or useful for answering the inquiry content in addition to the prediction result of the inquiry content.

When the display function unit 301 of the operator terminal 30 receives the information necessary for displaying the operator screen or the operator screen from the inquiry prediction device 10, the display function unit 301 displays the operator screen (step S212). The details of the operator screen will be described later.

Then, the telephone function unit 302 of the operator terminal 30 responds to the customer terminal 20 that responded in the above step S210 via the exchange 40, for example, in response to the operator's response operation (step S213). As a result, a voice call is started between the customer terminal 20 and the operator terminal 30.

Finally, the analysis unit 102 of the inquiry prediction device 10 indicates the action history information created in step S206, the inquiry information indicating that the predicted inquiry content in step S210 is correct, or the customer. As a result of the voice call between the terminal 20 and the operator terminal 30, the prediction template is updated using the inquiry information representing the inquiry content actually received by the operator (step S214). The inquiry information includes Key information added to the destination telephone number when making a telephone call in step S202. Here, as the method of updating the prediction template, the center of gravity and the weight may be updated by the same method as the method of creating the prediction template. The prediction template does not need to be updated in real time. For example, the inquiry information may be stored in the inquiry information storage unit 106 and may be performed every predetermined period or when a predetermined trigger occurs. ..

Here, an example of the operator screen displayed on the operator terminal 30 in step S212 will be described with reference to FIG. 10. FIG. 10 is a diagram showing an example of an operator screen.

The operator screen 1000 shown in FIG. 10 includes the inquiry content 1100 predicted in step S207 above. In the example shown in FIG. 10, "new mortgage loan" is displayed as the inquiry content 1100. As a result, the operator can know the contents of the inquiry of the customer who will make a voice call from now on.

Further, the operator screen 1000 shown in FIG. 10 includes, for example, an action history list 1200 in which a Web browsing history and the like are displayed, a Web browsing history detail 1300 in which a more detailed Web browsing history is displayed, and the like. This allows the operator to know what kind of Web page the customer has viewed before calling the call center. The action history list 1200 and the Web browsing history details 1300 can be displayed using the action history information.

In addition to the above, the operator screen 1000 may include various information. For example, when the browsing history of the Web page can be acquired as the behavior history information of the customer, various information that can be displayed from the behavior history information may be included. For example, if the history of sending / receiving e-mails, the history of questionnaires, the history of using smart speakers, the history of answering telephone calls in the past, and the like can be acquired, such information may be displayed.

As described above, when the customer makes an inquiry to the call center, the inquiry prediction system 1 according to the present embodiment predicts the inquiry content using the customer's behavior history and presents the prediction result to the operator. As a result, in the inquiry prediction system 1 according to the present embodiment, the operator can know the contents of the customer's inquiry in advance, so that an appropriate answer can be promptly given when a voice call is actually connected to the customer. It becomes possible to convey. Therefore, it is possible to reduce the waiting time for customers when answering a call at a call center, which can contribute to the improvement of customer satisfaction.

In the present embodiment, as an example, the case of predicting the inquiry content from the customer's behavior history information has been described, but the present invention is not limited to this, and can be applied to the case of predicting various information. For example, by using the purchase information of the product instead of the inquiry information, it is possible to predict the product to be purchased by the customer from the behavior history information of the customer.

The present invention is not limited to the above-described embodiment disclosed specifically, and various modifications and modifications, combinations with existing techniques, and the like are possible without departing from the description of the scope of claims. be.

1 Inquiry prediction system 10 Inquiry prediction device 20 Customer terminal 30 Operator terminal 40 Switch 50 Web server 101 Collection unit 102 Analysis unit 103 Response unit 104 Screen provision unit 105 Action history information storage unit 106 Inquiry information storage unit 201 Web browsing unit 202 Telephone function Department 301 Display function unit 302 Telephone function unit 501 History provision unit

Claims (6)

A prediction system that includes a terminal used by a customer and a prediction device that predicts the content of an inquiry from the customer.
The prediction device is
The first behavior history information indicating the behavior history of the customer and the first behavior history information via the Key information which is the identification information for identifying the customer when the behavior represented by the behavior history is performed. A model creation means for creating a prediction model for predicting the content of an inquiry from the customer using the associated inquiry information, and a model creation means.
When the customer makes a call to the call center with the terminal, the customer's inquiry content on the phone is obtained by using the second action history information indicating the action history of the customer before the call is made and the prediction model. Predictive means to predict and
Have,
The terminal is
When making a call to the call center via a Web page, the calling means for making the call by adding the Key information generated when displaying the Web page to the telephone number of the call center.
In response to the transmission, the Key information and predetermined information generated by the Web server that provides the Web page and for acquiring the behavior history of the customer are transmitted to the prediction device. And the means of transmission
A prediction system characterized by having. The prediction system according to claim 1, wherein the action history is a browsing history of a Web page. The prediction device is
It has an inquiry creating means for creating inquiry information including the Key information and the result of confirming the correctness of the inquiry content predicted by the prediction means or the actual inquiry content of the customer in the call center with the terminal. ,
The model creation means is
The prediction according to claim 2, wherein the prediction model is updated by using the second action history information and the inquiry information including the same Key information as the second action history information. system. The prediction system according to claim 2 or 3, wherein the predetermined information is a cookie. A prediction method used in a prediction system including a terminal used by a customer and a prediction device for predicting the content of an inquiry from the customer.
The predictor
The first behavior history information indicating the behavior history of the customer and the first behavior history information via the Key information which is the identification information for identifying the customer when the behavior represented by the behavior history is performed. A model creation procedure for creating a prediction model for predicting the content of an inquiry from the customer using the associated inquiry information, and a model creation procedure.
When the customer makes a call to the call center with the terminal, the customer's inquiry content on the phone is obtained by using the second action history information indicating the action history of the customer before the call is made and the prediction model. Prediction procedure to predict and
And run
The terminal
When making a call to the call center via a web page, the call procedure for making the call by adding the Key information generated when displaying the web page to the telephone number of the call center and the call center.
In response to the transmission, the Key information and predetermined information generated by the Web server that provides the Web page and for acquiring the behavior history of the customer are transmitted to the prediction device. Sending procedure and
A prediction method characterized by performing. A program for making a computer function as each means in the prediction system according to any one of claims 1 to 4.

Images