Processing device, processing method, and program

Abstract

To facilitate evaluation on response operation of an operator in a contact center.SOLUTION: An AI server 100 includes: a response evaluation unit 10 which calculates a response speed of an operator U10, using the number of characters in a chat from a customer U20 and a time interval between reception of the chat from the customer U20 and transmission of a response chat transmitted by the operator U10; and an answer evaluation unit 20 which creates an optimal answer candidate for the chat from the customer U20, calculates a matching degree between the optimal answer candidate and the response chat from the operator U10, and calculates difficulty in creating the optimal answer candidate.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a processing device, a processing method, and a program. [Background technology]

[0002] In the past, in contact centers, which are business divisions that specialize in responding to customers via text chat, managers evaluated the operators who handled the response work, and used the evaluation results to provide operator training and improve operations, thereby improving response quality. Operator evaluations were conducted by comparing the time it took for an operator to respond to a chat message received from a customer, the accuracy of the operator's response, and other factors with indicators prepared in advance by the contact center. Non-Patent Document 1 discloses monitoring the "accuracy of the operator's response" in contact center operations. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] “Glossary of Monitoring”, [online], KDDI Evolva Corporation, [searched January 27, 2022], Internet,<URL:https: / / www.k-evolva.com / glossary / monitoring / > Summary of the Invention [Problem to be solved by the invention]

[0004] However, the evaluation of operators as described above requires the manager to review the results after the fact, which is a cumbersome process and requires a great deal of time.

[0005] In view of the above circumstances, an object of the present invention is to make it easier to evaluate the customer service work of contact center operators. [Means for solving the problem]

[0006] The present invention, which solves the above problem, provides a processing device including: a first processing unit, in an operator device used by an operator, inputting a first chat message received from a customer device used by a customer, acquiring first character count information in the first chat message; inputting, in the operator device, first time information when the first chat message is received from the customer device and second time information when a second chat message is sent to the customer device as a reply to the first chat message, acquiring first time information from the first time information to the second time information, and outputting a first response speed calculated based on the first time information and the first character count information; and a second processing unit, which, upon inputting the first chat message, creates a first reply message based on the message content of the first chat message, compares the first reply message with the message content of the second chat message, outputs degree of agreement information based on the comparison result, calculates a difficulty level for creating the first reply message, and outputs difficulty level information based on the calculation result.

[0007] The present invention also provides a processing method, in which a processing device executes the following steps: a first step of inputting, into an operator device used by an operator, a first chat message received from a customer device used by a customer, and acquiring first character count information in the first chat message; inputting, into the operator device, first time information when the first chat message is received from the customer device and second time information when a second chat message is sent to the customer device as a reply to the first chat message, and acquiring first time information from the first time information to the second time information; and outputting a first response speed calculated based on the first time information and the first character count information; and a second step of inputting, upon inputting the first chat message, creating a first reply based on the message content of the first chat message, comparing the first reply content with the message content of the second chat message, outputting coincidence information based on the comparison result, calculating a difficulty level for creating the first reply content, and outputting difficulty level information based on the calculation result.

[0008] The present invention also provides a program for causing a computer of a processing device to function as a first processing unit that inputs, into an operator device used by an operator, a first chat message received from a customer device used by a customer, and acquires first character count information in the first chat message; inputs, into the operator device, first time information when the first chat message is received from the customer device and second time information when a second chat message is sent to the customer device as a reply to the first chat message, and acquires first time information from the first time information to the second time information; and outputs a first response speed calculated based on the first time information and the first character count information; and a second processing unit that, upon input of the first chat message, creates a first reply based on the message content of the first chat message, compares the first reply with the message content of the second chat message, and outputs similarity information based on the comparison result; calculates a difficulty level for creating the first reply message, and outputs difficulty level information based on the calculation result. [Effects of the Invention]

[0009] According to the present invention, it is possible to easily evaluate the response work of an operator at a contact center. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 is a functional configuration diagram of a processing system according to an embodiment of the present invention. [Figure 2] This is a flowchart of the processing performed by the AI ​​server. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. Each drawing is merely a schematic illustration to allow a sufficient understanding of the present invention. Therefore, the present invention is not limited to the illustrated examples. Furthermore, in the drawings, the dimensions of components constituting the present invention may be exaggerated for clarity. In addition, in each drawing, common or similar components are designated by the same reference numerals, and redundant explanations thereof will be omitted.

[0012] [overview] The present invention relates to a method for evaluating the responses provided by an operator and suggesting desirable responses by utilizing artificial intelligence (AI) functions in a contact center that handles chats. As mentioned above, conventionally, operators have been evaluated by comparing the time it takes for the operator to respond to a chat received from a customer, the accuracy of the content of the operator's response, and other factors with indicators prepared in advance by the contact center.

[0013] However, with the above method, the time it takes for an operator to respond via chat depends on, for example, the length of the message from the customer, whether an answer to the customer's question has been prepared in advance, and whether the customer's question is a frequently asked question or a rare question. Furthermore, the accuracy of the operator's response depends on whether the customer's message is logical, whether an answer has been prepared in advance, and so on. Because there are various factors to consider when evaluating an operator's response performance, the manager's review of the evaluations becomes cumbersome and requires a great deal of time.

[0014] [composition] FIG. 1 is a functional configuration diagram of a processing system according to this embodiment. The processing system 1 includes a server system 2 configured with an AI server 100, a business information DB 200, and a contact center server 300. The processing system 1 includes one or more operator communication devices 400. The AI ​​server 100, the business information DB 200, the contact center server 300, and the operator communication devices 400 are communicatively connected via an in-house communication network 600. The processing system 1 includes one or more customer communication devices 500. The customer communication devices 500 are communicatively connected to the contact center server 300 via a public network 700. An operator U10 operates the operator communication device 400 to handle customer U20's calls. The customer U20 owns and operates the customer communication device 500.

[0015] (Operator communication device 400) The operator communication device 400 is a device that provides the operator U10 with chat communication with the customer on the customer communication device 500 connected by the contact center server 300. The operator communication device 400 has at least a display device that displays text from the customer on the customer communication device 500 via the contact center server 300, an input device for the operator U10 to input text, and a control function for conducting text chat communication (the communication control function is omitted).

[0016] (Contact Center Server 300) The contact center server 300 is a server device that controls the outgoing and incoming calls at the contact center and executes the processing of the contact center services provided to customers (users). The contact center server 300 is configured as a computer having a CPU (Central Processing Unit), memory, storage means (storage unit) such as a hard disk, and a network interface. This computer realizes various functions by the CPU executing programs loaded into the memory.

[0017] (AI Server 100) The AI ​​server 100 is a server that provides so-called artificial intelligence (AI) functions, and has functions such as sentence understanding, semantic understanding, sentiment analysis, text entailment recognition, and machine learning based on natural language processing. The AI ​​server 100 is configured as a computer having a CPU, memory, storage means (storage unit) such as a hard disk, and a network interface. This computer realizes various functions by having the CPU execute programs loaded into the memory.

[0018] The AI ​​server 100 can acquire the contents of communication between the customer communication device 500 and the operator communication device 400 in real time, along with information identifying the customer U20 and information identifying the operator U10. The AI ​​server 100 also stores processing results and has an interface that allows the administrator or operator U10 to refer to them.

[0019] The AI ​​server 100 comprises a response evaluation unit 10 and an answer evaluation unit 20. The response evaluation unit 10 evaluates the response of the operator U10 to the chat from the customer U20. The reply evaluation unit 20 evaluates the content of the reply from the operator U10 to the chat from the customer U20.

[0020] (Business Information DB200) The business information DB 200 is a database that stores various information related to contact center operations. For example, the business information DB 200 stores, as text information, so-called "frequently asked questions" provided by the contact center, a collection of answers to FAQs (Frequently Asked Questions), manuals related to contact center operations, manuals related to services and products offered by companies that use the contact center services, catalogs of the services and products, etc. The business information DB 200 can be designed to be implemented in a computer (not shown) that is communicatively connected to the local area network 600.

[0021] (Local area network 600, public network 700) The local area network 600 may be, for example, a LAN (Local Area Network). The public network 700 can be, for example, an IP (Internet Protocol) network.

[0022] [Evaluation of operator response] When the contact center server 300 receives a call from the customer communication device 500, the contact center server 300 selects an appropriate operator using ACD (Automatic Call Distribution) and connects the customer communication device 500 to the operator communication device 400 of the selected operator U10. The operator U10 then checks the message (chat) from the customer U20 on the operator communication device 400, and enters a response (chat) into the operator communication device 400. The contact center server 300 returns the response to the customer communication device 500, thereby carrying out communication between the customer U20 and the operator U10.

[0023] Regarding this exchange, the AI ​​server 100 can obtain messages (chat) from the customer communication device 500 from the contact center server 300. The response evaluation unit 10 can analyze the messages from the customer communication device 500 and obtain and store the content of the message from the customer U20 (MC001) and the number of characters in the message (MC001N). The response evaluation unit 10 can also obtain and store the timing (MC001T) at which the operator communication device 400 received the message from the customer U20. The timing can indicate, for example, when the message from the customer U20 was displayed on the operator communication device 400 of the operator U10, but is not limited to this.

[0024] The response evaluation unit 10 can also obtain a response (chat) from the operator communication device 400 from the contact center server 300. The response evaluation unit 10 can analyze the response from the operator communication device 400 and obtain and store the content of the message (MO001) and the number of characters in the message (MO001N) that the operator U10 responded to. The response evaluation unit 10 can also obtain and store the timing (MO001T) at which the operator U10 sent a reply to the customer U20. The timing can be, for example, the timing at which the response from the operator communication device 400 is sent to the customer communication device 500, but is not limited to this.

[0025] The response evaluation unit 10 can calculate the response speed from the number of characters in the message from the customer U20 (MC001N) and the time interval (MO001T-MC001T) between when the operator communication device 400 receives the message from the customer U20 and when the operator U10 replies to the customer U20. The response speed is an index that quantitatively estimates the operator U10's response skills; the higher the response speed, the shorter the waiting time for a response on the customer U20's side, and the better the operator U10's response skills can be said to be.

[0026] For example, the response evaluation unit 10 can calculate the response speed as a value proportional to the number of characters (MC001N) and inversely proportional to the time interval (MO001T-MC001T). This is because the larger the number of characters (MC001N), the longer it takes the operator U10 to understand the content of the inquiry from the customer U20. Also, the shorter the time interval (MO001T-MC001T), the faster the reply from the operator U10.

[0027] [Evaluation of operator's answers] The response evaluation unit 20 performs sentence understanding, semantic understanding, sentiment analysis, and textual entailment recognition based on natural language processing on the content of the message (MC001) from the customer U20. The response evaluation unit 20 also performs sentence understanding, semantic understanding, sentiment analysis, and textual entailment recognition based on natural language processing on the content of the message (MO001) responded to by the operator U10.

[0028] The answer evaluation unit 20 also references the business information DB 200 to create an optimal answer candidate (MO002) for the content of the chat message (MC001) entered by the customer U20. The optimal answer candidate is a model of the content of a response to the content of the message (MC001) from the customer U20, and is based on AI functions. The answer evaluation unit 20 may also perform sentence understanding, semantic understanding, sentiment analysis, and textual entailment recognition based on natural language processing on the created answer candidate.

[0029] (degree of match) The answer evaluation unit 20 compares the content (MO001) of the message responded by the operator U10 with the answer candidate (MO002) to calculate the degree of match. For example, the calculation of the degree of match is performed as follows. That is, morphological analysis is performed to divide each of the content (MO001) and the answer candidate (MO002) into morphemes. Next, the number of hits is counted, which indicates the frequency with which the morphemes of the content (MO001) match the morphemes of the answer candidate (MO002), and also with the morphemes of the content (MO001) that follow the morphemes of the content (MO001) match the morphemes of the answer candidate (MO002). Finally, a value based on the number of hits is calculated as the degree of match. Note that the number of hits itself may also be used as the degree of match. The answer evaluation unit 20 determines that the greater the number of hits, the higher the degree of match. The answer evaluation unit 20 can evaluate that the higher the degree of match, the more appropriate the response the operator U10 provided to the customer U20.

[0030] Furthermore, when the operator U10 digs deeper to find an answer, the answer may be reached after multiple exchanges of messages (chat exchanges with the customer U20). For this reason, the answer evaluation unit 20 can calculate the degree of match not only for one exchange but also for the content leading up to the answer. In other words, the answer evaluation unit 20 calculates the degree of match for a series of exchanges between the customer U20 and the operator U10. The answer evaluation unit 20 also calculates the degree of match for all exchanges between the agents.

[0031] (Difficulty) Furthermore, when creating an optimal answer candidate (MO002), the answer evaluation unit 20 calculates the difficulty of creating it by referring to the business information DB 200. For example, the answer evaluation unit 20 can calculate the difficulty so that the smaller the degree of reuse of various information stored in the business information DB 200, the larger the value of the difficulty.

[0032] For example, suppose the answer evaluation unit 20 determines that a specific collection of answers to questions stored in the business information DB 200 can be used as is (i.e., without editing) as the optimal answer candidate (MO002) for the content of a message (MC001) from customer U20. In this case, creating the optimal answer candidate (MO002) is easy because it requires only extracting the specific collection of answers to questions, and the answer evaluation unit 20 can calculate the difficulty level as a small value.

[0033] When the operator communication device 400 refers to the business information DB 200, extracts the above-mentioned specific question and answer collection in response to the content of the chat message (MC001) entered by the customer U20, and the operator U10 responds with the extracted question and answer collection as the message content (MO001), the response speed becomes extremely fast, and the response evaluation unit 20 can evaluate that the operator U10's service was good. On the other hand, when the operator U10 does not (cannot) extract the above-mentioned specific question and answer collection, and instead creates the message content (MO001) himself and responds, the response speed becomes extremely slow, and the response evaluation unit 20 can evaluate that the operator U10's service was not good.

[0034] Also, for example, suppose that the answer evaluation unit 20 determines that a specific collection of answers to questions stored in the business information DB 200 can be edited a little to create an optimal answer candidate (MO002) for the content of the message (MC001) from customer U20. In this case, creating the optimal answer candidate (MO002) requires not only extracting a specific collection of answers to questions but also some editing, so the answer evaluation unit 20 can calculate the difficulty level as a medium value according to the degree of editing.

[0035] In response to the chat message content (MC001) entered by customer U20, the operator communication device 400 refers to the business information DB 200, extracts and edits the above-mentioned specific question and answer collection, and then when operator U10 responds with the edited question and answer collection as the message content (MO001), the response speed is faster than when the operator U10 creates the message content (MO001) himself, and the response evaluation unit 20 can evaluate that the operator U10's service performance was good.On the other hand, if operator U10 does not (cannot) extract the above-mentioned specific question and answer collection and instead creates and responds with the message content (MO001) himself, the response speed is extremely slow, and the response evaluation unit 20 can evaluate that the operator U10's service performance was poor.

[0036] Also, for example, suppose that the answer evaluation unit 20 determines that a question and answer collection that can be used to create an optimal answer candidate (MO002) for the content (MC001) of the message from customer U20 is not stored in the business information DB 200. In this case, the answer evaluation unit 20 needs to create the optimal answer candidate (MO002) without using a question and answer collection, and therefore the answer evaluation unit 20 can calculate the difficulty as a large value.

[0037] Suppose the operator communication device 400 references the business information DB 200 in response to the content of a chat message (MC001) entered by customer U20, but is unable to extract the above-mentioned collection of questions and answers. If the operator U10 responds with a collection of questions and answers that he or she created himself or herself as the content of the message (MO001), the response speed will inevitably be slow because it takes a lot of time to create it. However, the response evaluation unit 20 can evaluate that the operator U10's service was good.

[0038] The AI ​​server 100 accumulates the results of the degree of match and difficulty. The contact center manager can refer to the results via the interface and use them for operational evaluation. In addition, the operator U10 can refer to examples of accurate answers, which can be useful for future customer service work.

[0039] [process] The processing performed by the AI ​​server 100 will now be described. Figure 2 is a flowchart of the processing performed by the AI ​​server. Assume that the contact center server 300 receives a chat message regarding an inquiry from a customer U20, distributes the call to a selected operator U10 using ACD, and then receives a chat message regarding the operator U10's response and sends it to the customer communication device 500.

[0040] 2, the response evaluation unit 10 acquires the content and number of characters in the chat of the customer U20 (step S1). Next, the response evaluation unit 10 acquires the response speed of the operator U10 using the number of characters in the chat of the customer U20, the timing of receiving the chat, and the timing of sending the chat of the operator U10 (step S2).

[0041] Next, the answer evaluation unit 20 refers to the business information DB 200 and creates an optimal answer candidate for the content of the chat of the customer U20 (step S3). Next, the answer evaluation unit 20 compares the chat of the operator U10 with the optimal answer candidate and calculates the degree of match between them (step S4). Finally, the answer evaluation unit 20 refers to the business information DB 200 and calculates the difficulty of creating an optimal answer candidate (step S5). This completes the processing performed by the AI ​​server 100.

[0042] [effect] According to this embodiment, AI calculates in real time the response speed, accuracy, and difficulty of the response of the operator U10 in response to the length of the message from the customer U20, thereby reducing the time required for evaluation and creating optimal response examples. This makes it easier to evaluate the customer service work of the contact center operator U10. Furthermore, the contact center manager can use the optimal response examples as feedback to train the operators. In addition, the degree of match is calculated using the results of morphological analysis of the content of the message (MO001) to which operator U10 responded and the results of morphological analysis of the answer candidate (MO002), so the content of the message (MO001) to which operator U10 responded can be objectively evaluated.

[0043] [Correspondence to claims] AI server 100 is an example of a "processing device" in the claims. The response evaluation unit 10 is an example of the "first processing unit" in the claims. The response evaluation unit 20 is an example of the "second processing unit" in the claims. The operator communication device 400 is an example of an "operator device" in the claims. Customer communication device 500 is an example of a "customer device" in the claims. The message (chat about an inquiry) from customer U20 is an example of the "first chat" in the claims. The number of characters in the message from customer U20 (MC001N) is an example of the "first character number information" in the claims. The timing (MC001T) at which the operator communication device 400 receives a message from the customer U20 is an example of the "first time information" in the claims. The reply chat from operator U10 is an example of a "second chat" as claimed. The timing (MO001T) when the reply is sent from the operator U10 to the customer U20 is an example of the "second time information" in the claims. The time interval (MO001T-MC001T) from when the operator communication device 400 receives a message from the customer U20 until when the operator U10 sends a reply to the customer U20 is an example of the "first time information" in the claims. The response speed is an example of the "first response speed" in the claims. The optimal answer candidate (MO002) is an example of the "first answer content" in the claims. The degree of coincidence is an example of "degree of coincidence information" in the claims. The difficulty level is an example of "difficulty level information" in the claims.

[0044] [Variations] (a) The AI ​​server 100 may be configured to additionally include a GPU (Graphics Processing Unit) for the purpose of achieving efficient parallel computation. (b) The business information DB 200 may be designed to be implemented in part or in whole on the AI ​​server 100.

[0045] (c): When the response evaluation unit 10 evaluates the response speed of an agent U10, it may take into account a situation in which the agent U10 simultaneously responds to chats from multiple customers U20. In such a situation, it is inevitable that the agent U10 will take time to respond to multiple responses. Therefore, even if the response speed of each response is slow, the response evaluation unit 10 may evaluate the agent U10's customer service performance as excellent. For example, the response evaluation unit 10 may calculate the response speed as a value proportional to the number of customers who will simultaneously respond. (d): After the response is completed, if the customer U20 answers a questionnaire regarding the operator U10's response, the response evaluation unit 10 may evaluate the operator U10's response based on the content of the questionnaire. For example, the contact center server 300 may send the questionnaire to the customer communication device 500, and the response to the questionnaire may be received from the customer communication device 500.

[0046] (e): The above-mentioned invention-specific features can be combined as appropriate. (f): Means that can be realized in software can be realized in hardware, and means that can be realized in hardware can be realized in software. [Explanation of symbols]

[0047] 1 Processing System 2. Server System 10 Response Evaluation Unit 20. Answer Evaluation Section 100 AI servers 200 Business information DB 300 Contact Center Servers 400 Operator Communication Device 500 Customer Communication Device 600 In-house communication network 700 Public Network U10 Operator U20 customer

Claims

1. In an operator device used by an operator, a first chat message received from a customer device used by a customer is input, and first character number information in the first chat message is acquired; inputting, into the operator device, first time information when the first chat message is received from the customer device and second time information when a second chat message is sent to the customer device as a reply to the first chat message, and acquiring first time information from the first time information to the second time information; a first processing unit that outputs a first response speed calculated based on the first time information and the first character number information; When the first chat message is input, a first reply content is created based on the message content of the first chat message; comparing the first reply content with the second chat message content, and outputting information on the degree of agreement based on the comparison result; a second processing unit that calculates a degree of difficulty for creating the first answer content and outputs difficulty level information based on the calculation result.

2. The processing device according to claim 1 , wherein the second processing unit outputs the degree of coincidence information based on a result of morphological analysis of the content of the first reply and a result of morphological analysis of the content of the message in the second chat.

3. The processing device In an operator device used by an operator, a first chat message received from a customer device used by a customer is input, and first character number information in the first chat message is acquired; inputting, into the operator device, first time information when the first chat message is received from the customer device and second time information when a second chat message is sent to the customer device as a reply to the first chat message, and acquiring first time information from the first time information to the second time information; a first step of outputting a first response speed calculated based on the first time information and the first character number information; When the first chat message is input, a first reply content is created based on the message content of the first chat message; comparing the first reply content with the second chat message content, and outputting information on the degree of agreement based on the comparison result; a second step of calculating a difficulty level for creating the first answer content and outputting difficulty level information based on the calculation result.

4. The processing unit computer In an operator device used by an operator, a first chat message received from a customer device used by a customer is input, and first character number information in the first chat message is acquired; inputting, into the operator device, first time information when the first chat message is received from the customer device and second time information when a second chat message is sent to the customer device as a reply to the first chat message, and acquiring first time information from the first time information to the second time information; a first processing unit that outputs a first response speed calculated based on the first time information and the first character number information; When the first chat message is input, a first reply content is created based on the message content of the first chat message; comparing the first reply content with the second chat message content, and outputting information on the degree of agreement based on the comparison result; A program for causing the computer to function as a second processing unit that calculates the degree of difficulty for creating the first answer content and outputs difficulty level information based on the calculation result.

Images