Dialogue quality inspection method, electronic device, computer storage medium and program product

Abstract

Embodiments of the present application provide a dialogue quality inspection method, an electronic device, a computer storage medium, and a computer program product. Among them, the dialogue quality inspection method includes: extracting dialogue data to be processed from a dialogue segment to be quality inspected, where the dialogue segment contains multiple pieces of dialogue data for service interaction between a service provider and a service recipient; if the dialogue data to be processed is the problem data of the service recipient, then determining the target dialogue data of the service provider located after the problem data from the dialogue segment; splicing the problem data, the intermediate dialogue data between the problem data and the target dialogue data of the service provider, and the target dialogue data of the service provider in the dialogue segment; performing a reply detection on whether there is a reply to the problem data for the spliced dialogue data, and determining the service quality of the service provider according to the detection result. Through the embodiments of the present application, the quality inspection efficiency is improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of computer technologies, and in particular, to a method for quality inspection of conversations, an electronic device, a computer storage medium, and a computer program product. Background Art

[0002] With the development of Internet technologies, more and more affairs in real life can be realized through the Internet, especially interactive affairs, such as Internet medical treatment, Internet customer service, Internet knowledge Q&A, and so on.

[0003] Taking Internet medical treatment as an example, a patient can conduct an online consultation through an online consultation community or an APP application, and remotely interact with a doctor to obtain answers or suggestions for the patient's health problems. To ensure the rights and interests of patients, the online consultation community or the APP application will conduct quality inspection on the service quality provided by the doctor. Currently, such quality inspection is mostly achieved by quality inspectors manually reading the complete conversation between the doctor and the patient and manually scoring the service quality of the doctor.

[0004] However, in this way, if the volume of consultations is large, a large amount of quality inspection manpower is required, and the quality inspection cost is high while the efficiency is low. Similarly, Internet affairs with similar scenarios, such as Internet customer service and Internet knowledge Q&A, also have the same problem. Summary of the Invention

[0005] In view of this, the embodiments of the present application provide a conversation quality inspection solution to at least partially solve the above problems.

[0006] According to a first aspect of the embodiments of the present application, there is provided a method for quality inspection of conversations, including: extracting to-be-processed conversation data from a conversation segment to be quality-inspected, where the conversation segment contains multiple pieces of conversation data for service interaction between a service provider and a service recipient; if the to-be-processed conversation data is the problem data of the service recipient, determining target conversation data of the service provider located after the problem data from the conversation segment; splicing the problem data, intermediate conversation data between the problem data and the target conversation data of the service provider, and the target conversation data of the service provider in the conversation segment; performing a reply detection on whether there is a reply to the problem data for the spliced conversation data, and determining the service quality of the service provider according to the detection result.

[0007] According to a second aspect of the embodiments of the present application, another method for dialogue quality inspection is provided, including: sending request information for requesting dialogue quality inspection of a dialogue segment of a service interaction, where the request information includes information of the dialogue segment to be quality inspected; receiving feedback information for the request information, where the feedback information includes quality inspection result information of the service quality provided by the service provider to the service recipient determined according to the dialogue data in the dialogue segment; obtaining service adjustment information matching the quality inspection result information, and replacing the service provider or sending service behavior adjustment instruction information to the service provider according to the service adjustment information.

[0008] According to a third aspect of the embodiments of the present application, yet another method for dialogue quality inspection is provided, including: sending request information for requesting quality inspection of the service quality of a doctor for a dialogue segment of an online consultation, where the request information includes information of the dialogue segment; receiving feedback information for the request information, where the feedback information includes quality inspection result information of the service quality provided by the doctor to the patient determined according to the dialogue data in the dialogue segment; obtaining service adjustment information matching the quality inspection result information, and replacing the doctor serving the patient or sending service behavior adjustment instruction information to the doctor according to the service adjustment information.

[0009] According to a fourth aspect of the embodiments of the present application, still another method for dialogue quality inspection is provided, including: sending request information for requesting quality inspection of the service quality of a customer service, where the request information includes identification information of the customer service; receiving feedback information for the request information, where the feedback information includes quality inspection result information of the service quality of the customer service determined according to the identification information; obtaining service adjustment information matching the quality inspection result information, and instructing to replace the customer service staff or sending service behavior adjustment instruction information to the customer service staff according to the service adjustment information.

[0010] According to a fifth aspect of the embodiments of the present application, an electronic device is provided, including: a processor, a memory, a communication interface, and a communication bus, where the processor, the memory, and the communication interface complete communication with each other through the communication bus; the memory is used for storing at least one executable instruction, and the executable instruction causes the processor to perform operations corresponding to the dialogue quality inspection method described in the first aspect or the second aspect or the third aspect or the fourth aspect.

[0011] According to a sixth aspect of the embodiments of the present application, a computer storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, it implements the dialogue quality inspection method described in the first aspect or the second aspect or the third aspect or the fourth aspect.

[0012] According to a seventh aspect of the embodiments of the present application, there is provided a computer program product including computer instructions that direct a computing device to perform operations corresponding to the dialogue quality inspection method described in the first aspect, the second aspect, the third aspect, or the fourth aspect.

[0013] According to the dialogue quality inspection solution provided by the embodiments of the present application, in a scenario of providing interactive services through dialogue, a device for service quality inspection can automatically perform quality inspection based on a dialogue segment to be inspected. Since the interactive service scenario usually involves a questioner and an answerer, that is, the person being served and the service provider, service quality inspection mostly targets the answers given by the service provider to the questions of the person being served. Therefore, in the solution provided by the embodiments of the present application, if the dialogue data extracted from the dialogue segment is the question data of the person being served, several sentences of dialogue data after the question data will be obtained and spliced, including the question data of the person being served, the dialogue data of the service provider, and other possible dialogue data between the two. Based on this, a reply detection is then performed to check whether the service provider has replied to the question of the person being served, thereby realizing the service quality detection of the service provider. It can be seen that through the solution of the embodiments of the present application, on the one hand, there is no longer a need for manual service quality inspection, reducing the relatively high quality inspection cost and improving the quality inspection efficiency; on the other hand, for non-standard question-and-answer dialogues (such as dialogues that are not in the form of one question and one answer), through the method of splicing and then detecting dialogue data, the problem of inaccurate detection caused by detecting through question-and-answer dialogue pairs in the traditional method can be effectively avoided, improving the accuracy of service quality detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] To more clearly illustrate the technical solutions in the embodiments of the present application or in the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings described below are only some embodiments recorded in the embodiments of the present application, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0015] Figure 1 Schematic diagram of an exemplary system applicable to the dialogue quality inspection method of the embodiments of the present application;

[0016] Figure 2A Flowchart of the steps of a dialogue quality inspection method according to Embodiment 1 of the present application;

[0017] Figure 2B For Figure 2A Schematic diagram of a scenario example in the illustrated embodiment;

[0018] Figure 3 Flowchart of the steps of a dialogue quality inspection method according to Embodiment 2 of the present application;

[0019] Figure 4It is a flowchart of steps of a dialogue quality inspection method according to Embodiment 3 of the present application;

[0020] Figure 5 It is a flowchart of steps of a dialogue quality inspection method according to Embodiment 4 of the present application;

[0021] Figure 6A It is a flowchart of steps of a dialogue quality inspection method according to Embodiment 5 of the present application;

[0022] Figure 6B It is Figure 6A a schematic diagram of a scenario example in the illustrated embodiment;

[0023] Figure 7A It is a flowchart of steps of a dialogue quality inspection method according to Embodiment 6 of the present application;

[0024] Figure 7B It is Figure 7A a schematic diagram of a scenario example in the illustrated embodiment;

[0025] Figure 8 It is a schematic diagram of the structure of an electronic device according to Embodiment 7 of the present application. Detailed implementation manners

[0026] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art shall fall within the protection scope of the embodiments of the present application.

[0027] The following further illustrates the specific implementation of the embodiments of the present application in conjunction with the accompanying drawings of the embodiments of the present application.

[0028] Figure 1 An exemplary system of a dialogue quality inspection method applicable to the embodiments of the present application is shown. As Figure 1 shown, the system 100 may include a server 102, a communication network 104, and / or one or more user devices 106, Figure 1 wherein in the example, there are multiple user devices (in the embodiments of the present application, unless otherwise specified, the quantities related to "multiple", "multiple types", etc. related to "multiple" all mean two or more).

[0029] Server 102 can be any suitable server for storing information, data, programs, and / or any other suitable type of content. In some embodiments, server 102 can perform any suitable functions. For example, in some embodiments, server 102 can be used for service quality inspection based on interactive conversations. As an alternative example, in some embodiments, server 102 can be used to splice the question data of the served person in the conversation segment to be quality inspected, the conversation data after the question data up to the conversation data of the server, and perform service quality inspection based on the spliced conversation data. As another example, in some embodiments, server 102 can be used to send the results of service quality inspection to the user device.

[0030] In some embodiments, communication network 104 can be any suitable combination of one or more wired and / or wireless networks. For example, communication network 104 can include any one or more of the following: the Internet, intranet, wide area network (WAN), local area network (LAN), wireless network, digital subscriber line (DSL) network, frame relay network, asynchronous transfer mode (ATM) network, virtual private network (VPN), and / or any other suitable communication network. User device 106 can be connected to communication network 104 through one or more communication links (e.g., communication link 112), and communication network 104 can be linked to server 102 via one or more communication links (e.g., communication link 114). The communication link can be any communication link suitable for transmitting data between user device 106 and server 102, such as a network link, dial-up link, wireless link, hardwired link, any other suitable communication link, or any suitable combination of such links.

[0031] User device 106 can include any one or more user devices suitable for providing interaction between the served person and the server. In some embodiments, user device 106 can include any suitable type of device. For example, in some embodiments, user device 106 can include mobile devices, tablet computers, laptop computers, desktop computers, wearable computers, game consoles, media players, vehicle entertainment systems, and / or any other suitable type of user device. It should be noted that in some embodiments, if user device 106 has good software and hardware performance, it can additionally or alternatively be used to implement service quality inspection of conversation data based on interactive conversations.

[0032] Although server 102 is illustrated as one device, in some embodiments, any suitable number of devices can be used to perform the functions performed by server 102. For example, in some embodiments, multiple devices can be used to implement the functions performed by server 102. Alternatively, the functions of server 102 can be implemented using cloud services.

[0033] Based on the above system, an embodiment of the present application provides a method for dialogue quality inspection, which will be described below through multiple embodiments.

[0034] Embodiment 1

[0035] Referring to Figure 2A , a flowchart showing the steps of a method for dialogue quality inspection according to Embodiment 1 of the present application is shown.

[0036] The dialogue quality inspection method of this embodiment includes the following steps:

[0037] Step S202: Extract the dialogue data to be processed from the dialogue segments to be quality inspected.

[0038] Among them, the dialogue segments to be quality inspected can be a complete dialogue. For example, the dialogue segment between a doctor and a patient from the start to the end of the dialogue; but it can also be a part of a complete dialogue, and this part forms a complete semantic segmentation. For example, an original complete dialogue contains Question 1 raised by Student A and the reply 1 from the teacher, as well as Question 2 raised by Student B and the reply 2 from the teacher. All the dialogues corresponding to Question 1 and the reply 1 can be taken as one dialogue segment, and all the dialogues corresponding to Question 2 and the reply 2 can be taken as another dialogue segment. Or, the patient had the first dialogue interaction with the doctor regarding the first question, and then had the second dialogue interaction with the doctor regarding the second question. Then these two interaction processes can be taken as two dialogue segments, and of course, they can also be taken as a complete dialogue segment. However, it is not limited to this, and other forms of dialogue segments are also applicable to the embodiments of the present application.

[0039] As can be seen from the above, the dialogue segments in the embodiments of the present application usually contain multiple dialogue data of service interactions between a service provider (such as a doctor, teacher, or customer service) and a service recipient (such as a patient, student, or consulting customer). It should be noted that a piece of dialogue data does not necessarily represent a natural sentence, but a sentence output by the service provider or the service recipient once. For example, the service recipient: "Doctor, my foot is swollen. Please take a look. What's going on? Is there any good solution?" That is, if these multiple natural sentences are output by the service recipient once, they are considered as one piece of dialogue data. Of course, the situation of outputting one natural sentence at a time is also applicable to the solution of the embodiments of the present application and is within the protection scope of the embodiments of the present application.

[0040] In a feasible way, the interaction between the service recipient and the service provider can be distinguished by an identification ID, wherein each conversation between each party will correspond to an identification ID, and the conversation data can be extracted and processed based on the identification ID. For example, assuming that the identification ID of the service recipient is ID-1 and the identification ID of the service provider is ID-2, the two have the following interactive conversation: ID-1: "Doctor, my foot is swollen. Can you see what's going on? Is there any good way?"; ID-2: "Is it twisted? What caused it?"; ID-1: "I fell off my bike yesterday"; ID-1: "But I didn't feel any discomfort in my foot at the time"; ID-2: "Have you done anything?"; ID-1: "No"; ID-2 "You can apply cold compresses and take rest. If it doesn't ease, it is recommended to go to the hospital"... In the above interactive conversation, the service recipient and the service provider had 7 conversations respectively, so it can be considered that it contains 7 conversation data, one conversation corresponds to an ID number and one conversation data, and one conversation data may contain only one natural sentence or multiple natural sentences.

[0041] It should be noted that the dialogue data in the embodiment of the present application may be the original input text data, or the text data obtained by converting the original input voice data, or a mixture of the two.

[0042] Step S204: If the dialogue data to be processed is the question data of the served party, the target dialogue data of the server located after the question data is determined from the dialogue segment.

[0043] As mentioned above, in the interactive dialogue between the server and the service recipient, the service recipient usually asks questions and the server answers them. However, in actual applications, there may be various phenomena such as the server not answering or answering questions that are not targeted at the service recipient's questions or answering with a bad tone or attitude. For this reason, when it is determined that the currently extracted dialogue data is the service recipient's question data, it is necessary to determine the subsequent dialogue data of the server from the dialogue segment to provide a basis for subsequent judgment on whether the server has answered the service recipient's questions in a targeted manner.

[0044] It should be noted that when determining the server's conversation data, you can first select the server's conversation data that is most recently located after the question data. If the conversation data is not reply data, such as it is also a server's question data, continue to search downward until you find conversation data that may be a reply as the target conversation data or stop searching when you determine that there is no reply data in the entire conversation segment.

[0045] Step S206: splicing the question data, the intermediate conversation data between the question data and the server's target conversation data, and the server's target conversation data in the conversation segment.

[0046] As described above, there are often dialogue scenarios in interactive services other than the one-question-one-answer form, which cannot be processed by the analysis and quality detection based on such one-question-one-answer dialogue pairs in the traditional way. Therefore, in the embodiments of the present application, after determining the target dialogue data of the service provider, all the dialogue data from the self-question data to the target dialogue data of the service provider will be spliced. It can be considered that the spliced dialogue data reflects the complete context of a question and its reply, containing more semantic information and context information.

[0047] Step S208: Perform a reply detection on the spliced dialogue data to check whether there is a reply to the question data, and determine the service quality of the service provider according to the detection result.

[0048] The target dialogue data determined in the foregoing steps may be reply data for the question data, or may be other dialogue data, and it needs to be analyzed and detected in combination with the entire spliced dialogue data to determine. If it is determined to be reply data, it can be considered that the service provider has reasonably answered the question of the service recipient, achieving a certain service quality.

[0049] It should be noted that in a feasible manner, if the question data includes multiple sub-questions, and the service provider has replied to some of the sub-questions, it can be considered that the service provider has replied to the question corresponding to the question data, so as to improve the quality inspection efficiency on the premise of meeting the basic quality inspection standards. However, it is not limited to this. It is also possible to judge whether each sub-question has been replied, and only when all are replied, it is considered that the service provider has replied to the question corresponding to the question data.

[0050] Next, the above process will be described by taking an online medical consultation scenario as an example, as Figure 2B shown.

[0051] Suppose the dialogue segment to be quality-inspected includes the following dialogue: Patient: "Doctor, my foot is swollen. Please take a look. What's going on? Is there any good solution?"; Doctor: "Did you sprain it? What caused it?"; Patient: "I fell off my bike yesterday"; Patient: "But I didn't feel anything wrong with my foot at that time"; Doctor: "Have you taken any treatment?"; Patient: "No"; Doctor: "You can apply cold compress and pay attention to rest. If it doesn't get better, it is recommended to go to the hospital"; Patient: "Which hospital is better?"; Doctor: "For your illness, it can be seen in general hospitals. Just find a nearby regular hospital"; Patient: "Okay, thank you, doctor". In this example, the service recipient is the patient and the service provider is the doctor.

[0052] Based on the above dialogue segment, first, the dialogue data <patient: "Doctor, my foot is swollen. Can you see what's going on? Is there any good way?"> is extracted; then, it is identified whether it is the patient's problem data; because the dialogue data is the patient's problem data, it is then determined whether the next dialogue data <doctor: "Is it twisted? What caused it?"> is the target dialogue data for the patient's problem data; because the dialogue data is the doctor's dialogue data, it is still a question sentence instead of a statement sentence used in a conventional reply, so the next doctor's dialogue data <doctor: "Have you done any treatment?"> is obtained, and it is still determined that it cannot be used as the target dialogue data; the next doctor's dialogue data <doctor: "You can apply cold compress and pay attention to rest. If it cannot be relieved, it is recommended to go to the hospital for a checkup"> is obtained, and it is identified as a possible reply to the problem data, and it is determined as the target dialogue data. The above process is simply illustrated by a page in the figure, where "X" indicates that the dialogue data is non-target dialogue data, and "√" indicates that the dialogue data is target dialogue data. Based on this, the dialogues from <patient: "Doctor, my foot is swollen. Can you tell me what's going on? Is there any good way?"> to <doctor: "You can apply cold compress and take a rest. If it doesn't work, I suggest you go to the hospital"> are spliced ​​to form the spliced ​​dialogue data, that is, <patient: "Doctor, my foot is swollen. Can you tell me what's going on? Is there any good way?"; Doctor: "Is it twisted? What caused it?"; Patient: "I fell off my bike yesterday"; Patient: "But I didn't feel any pain in my foot at the time"; Doctor: "Have you done anything?"; Patient: "No"; Doctor: "You can apply cold compress and take a rest. If it doesn't work, I suggest you go to the hospital">; In this example, the dialogue data is input into a neural network model for judging whether the question and answer match, and the response detection is performed to output the matching score. Assuming that the output matching score is 8 points (10 points is the full score), it can be considered that the doctor has responded to the patient's question in a targeted manner and the service quality is good.

[0053] It can be seen that in the scenario of interactive services through conversations in this embodiment, the device for service quality inspection can perform automatic quality inspection based on the conversation segments to be quality inspected. Since the interactive service scenario usually involves the questioner and the answerer, that is, the service recipient and the service provider, service quality inspection is mostly carried out on the question responses of the service provider to the service recipient. Therefore, in the solution provided in this embodiment, if the conversation data extracted from the conversation segment is the question data of the service recipient, several sentences of conversation data after the question data will be obtained and spliced, including the question data of the service recipient, the conversation data of the service provider, and other possible conversation data between the two. Based on this, a reply detection is then performed on whether the service provider has replied to the question of the service recipient, so as to achieve the service quality detection of the service provider. It can be seen that through the solution of this embodiment, on the one hand, there is no longer a need for manual service quality inspection, reducing the relatively high quality inspection cost and improving the quality inspection efficiency; on the other hand, for non-standard Q&A conversations (such as conversations that are not in the form of one question and one answer), through the method of splicing and then detecting conversation data, the problem of inaccurate detection caused by detecting through one question and one answer conversation pairs in the traditional method can be effectively avoided, and the accuracy of service quality detection is improved.

[0054] Embodiment 2

[0055] Refer to Figure 3 , which shows a flowchart of the steps of a conversation quality inspection method according to Embodiment 2 of the present application.

[0056] The conversation quality inspection method of this embodiment includes the following steps:

[0057] Step S302: Extract the conversation data to be processed from the conversation segment to be quality inspected.

[0058] Among them, the conversation segment contains multiple pieces of conversation data for service interaction between the service provider and the service recipient.

[0059] In a feasible manner, this step can be implemented as: obtaining the conversation segment to be quality inspected; extracting the conversations in the conversation segment one by one, and generating a conversation data sequence containing role information and conversations according to the time sequence of the conversations in the conversation segment and the role information corresponding to each conversation; extracting the conversation data to be processed from the conversation data sequence.

[0060] In an interactive conversation, usually each sentence of the conversation corresponds to a corresponding identifier, which can represent the role information of both parties in the conversation. As described in the previous Embodiment 1, ID-1 represents the patient, and ID-2 represents the doctor. Based on this, in this embodiment, a conversation data sequence will be generated according to the time sequence of the conversation and the role information of the conversation. For example, <ID-1: "Doctor, my foot is swollen. Please take a look. What's going on? Is there any good solution?" --- ID-2: "Did you sprain it? What caused it?" --- ID-1: "I fell off my bike yesterday"...>. It can be seen that the conversation data extracted from this conversation data sequence contains both role information and conversation content. In this way, both the efficiency of subsequent conversation data processing is improved, and the management of conversation data is facilitated.

[0061] It should be noted that in practical applications, the above form of ID identification, the form of direct role marking can also be equally applicable. For example, replace ID-1 with patient XXX, replace ID-2 with doctor YYY, etc. However, it is not limited to this, and other ways to determine role information are also equally applicable.

[0062] Step S304: Determine whether the conversation data to be processed is the conversation data of the served party or the conversation data of the server; if it is the conversation data of the server, then execute Step S306; if it is the conversation data of the served party, then execute Step S308.

[0063] In this embodiment, the conversation data of the served party and the conversation data of the server are respectively processed according to the roles corresponding to the conversation data. That is to say, subsequently, while determining whether the server has replied to the question of the served party based on the conversation data, the service attitude of the server will also be synchronously detected to more accurately and comprehensively reflect the quality of the service provided by the server.

[0064] In specific implementation, optionally, the role information of the conversation data to be processed can be obtained; determine whether the conversation data to be processed is the conversation data of the server or the conversation data of the served party according to the role information; if the conversation data to be processed is the conversation data of the server, then execute Step S306, that is, perform quality inspection on the service attitude of the conversation data to be processed and obtain the quality inspection result of the service attitude; if the conversation data to be processed is the conversation data of the served party, then execute Step S308, that is, if the conversation data to be processed is the question data of the served party, then perform the operation of determining the target conversation data of the server located after the question data from the conversation segment.

[0065] Step S306: Perform quality inspection on the service attitude of the conversation data to be processed and obtain the quality inspection result of the service attitude. Then, proceed to execute Step S314.

[0066] In a feasible manner, when the dialogue data to be processed is the dialogue data of the service provider, the dialogue data can be matched with preset service attitude keywords, and according to the matching result and the service attitude evaluation information corresponding to the keyword, the service attitude quality inspection result of the service provider can be obtained. The implementation of service attitude quality inspection in this way is relatively simple and the quality inspection speed is fast.

[0067] In another feasible manner, when the dialogue data to be processed is the dialogue data of the service provider, the dialogue data can be input into a pre-trained neural network model for service attitude evaluation, and according to the result output by the model, the service attitude quality inspection result of the service provider can be obtained. In this way, the accuracy of the obtained service attitude quality inspection result is higher.

[0068] However, it is not limited to this, and other service attitude quality inspection methods are also applicable to the solution of the embodiments of the present application.

[0069] Step S308: If the dialogue data to be processed is the problem data of the service recipient, determine the target dialogue data of the service provider located after the problem data from the dialogue segment.

[0070] In the specific implementation of this step, since it has been determined that it is the dialogue data of the service recipient, but it may be problem data or non-problem data. Therefore, after determining that the dialogue data to be processed is the dialogue data of the service recipient, further determine whether it is problem data. After determining that it is problem data, determine the target dialogue data of the service provider located after the problem data from the dialogue segment.

[0071] That is, in a feasible manner, this step can be implemented as: if it is determined that the dialogue data to be processed is the dialogue data of the service recipient according to the role information of the dialogue data to be processed, then perform identification detection on whether the dialogue data to be processed is problem data; if it is determined that the dialogue data to be processed is problem data according to the identification detection result, then determine the target dialogue data of the service provider whose time sequence is after the problem data from the dialogue segment.

[0072] Among them, the identification detection of problem data can adopt a pre-trained neural network model with the function of identifying problem data, but it is not limited to this, and other methods or algorithms that can identify problem data are also applicable, such as judging according to whether the dialogue data contains interrogative words and / or question marks, etc.

[0073] Generally speaking, interactive conversations have time information. In this embodiment, this time information is used to determine the time sequence of the conversation data, so that after determining that the conversation data of the served person is problem data, the target conversation data of the server can be found from the conversation segments according to the time sequence of the conversation data and which is after the problem data in terms of time sequence. However, it is not limited to this. If the conversation data has corresponding sequence numbers, the target conversation data whose sequence is after the problem data can also be found from the conversation segments, and so on.

[0074] In a feasible way, the determination of the target conversation data of the server can also use a pre-trained neural network model. For example, the problem data and the current conversation data can be input into the neural network model together, and the matching degree between the two is output by the neural network model. If the matching degree is higher than a preset threshold, the current conversation data can be used as the target conversation data. If the matching degree is lower than the preset threshold, the next conversation data is continued to be extracted. Among them, the specific setting of the preset threshold can be set by those skilled in the art according to actual needs, and the embodiments of the present application do not limit this.

[0075] Although the target conversation data can be selected, whether this target conversation data is a reply data for the problem data still needs further processing and more accurate judgment.

[0076] Step S310: Concatenate the problem data, the intermediate conversation data between the problem data and the target conversation data of the server, and the target conversation data of the server in the conversation segments.

[0077] In a feasible way, according to the time sequence of the conversation data in the conversation segments, the problem data, the intermediate conversation data whose time sequence is between the problem data and the target conversation data of the server, and the target conversation data of the server can be concatenated. Thus, the time information inherent in the conversation data itself is fully utilized, the efficiency of conversation data concatenation is improved, and the cost of conversation data concatenation is saved. However, as mentioned above, if the conversation data has corresponding sequence numbers, concatenation can also be performed based on the sequence numbers. The concatenated conversation data contains richer information, providing an effective basis for subsequent processing.

[0078] Step S312: Perform a reply detection on whether there is a reply to the problem data in the concatenated conversation data, and obtain the detection result of the reply detection.

[0079] Reply detection based on the spliced dialogue data can also be implemented using a pre-trained neural network model capable of detecting the matching degree of question replies. In a feasible approach, the neural network model used in this step can be reused with the neural network model for determining the target dialogue data. However, different from the input question data and the current dialogue data when determining the target dialogue data, the spliced dialogue data with richer semantic information needs to be input when performing reply detection. However, this is not limited to this. The neural network model for reply detection can also be different from the neural network for determining the target dialogue data, as long as it has the corresponding functions.

[0080] The neural network model for reply detection will output the matching degree between the question data and the target dialogue data based on the spliced dialogue data. Since this step is based on the spliced dialogue data for detection, the output matching degree is more accurate and can be used as the basis for evaluating the service quality of the service provider.

[0081] In a feasible approach, the reply detection of whether there is a reply to the question data in the spliced dialogue data can be implemented as follows: perform a first encoding on the dialogue content in the spliced dialogue data to obtain dialogue content encoding data; perform a second encoding on the role information in the spliced dialogue data to obtain role encoding data; splice the content encoding data and the role encoding data to obtain the spliced encoding data; perform reply detection on whether there is a reply to the question data based on the spliced encoding data. Since there are obvious differences in both the data size and data type between the dialogue content and the role information, the method of encoding the dialogue content and the role information separately can effectively improve the encoding efficiency and is also convenient for the unified management of the dialogue content and the role information. Among them, the first encoding and the second encoding are different encoding methods. However, in specific applications, those skilled in the art can select appropriate encoding methods according to the above characteristics of the dialogue content and the role information, including but not limited to the word vector method, the method of encoding through a neural network model, and so on.

[0082] Further optionally, when performing a first encoding on the dialogue content in the spliced dialogue data to obtain dialogue content encoding data, corresponding role information can be added to the dialogue content in the spliced dialogue data; perform a first encoding on the dialogue content after adding the role information to obtain dialogue content encoding data. The encoding data formed in this way is more easily detected and recognized by the neural network model, improving the detection and recognition efficiency of the neural network model.

[0083] Step S314: Determine the service quality of the service provider according to the service attitude quality inspection result and the detection result of the reply detection.

[0084] After obtaining the service attitude quality inspection results and the detection results of reply detection, the service quality of the service provided by the service provider can be comprehensively evaluated by combining the performance of the service provider in terms of attitude and reply, so as to obtain a relatively accurate service quality evaluation result.

[0085] Through this embodiment, in the scenario of interactive services through conversations, the device for service quality inspection can automatically perform quality inspection based on the conversation segments to be quality inspected. Since the interactive service scenario usually involves the questioner and the answerer, that is, the service recipient and the service provider, service quality inspection is mostly carried out on the service provider's reply to the service recipient's question. Therefore, in the solution provided by this embodiment, if the conversation data extracted from the conversation segment is the question data of the service recipient, several sentences of conversation data after the question data will be obtained and spliced, including the question data of the service recipient, the conversation data of the service provider, and other possible conversation data between the two. Based on this, reply detection is performed on whether the service provider has replied to the service recipient's question, so as to realize the service quality detection of the service provider. It can be seen that through the solution of this embodiment, on the one hand, there is no need for manual service quality inspection, which reduces the relatively high quality inspection cost and improves the quality inspection efficiency; on the other hand, for non-standard question-and-answer conversations (such as conversations that are not in the form of one question and one answer), through the method of splicing and detecting conversation data, the problem of inaccurate detection caused by detecting through the traditional method of question-and-answer conversation pairs can be effectively avoided, and the accuracy of service quality detection is improved.

[0086] Embodiment III

[0087] Referring to Figure 4 , a step flow chart of a conversation quality inspection method according to Embodiment III of the present application is shown.

[0088] This embodiment uses the interactive conversation in the online consultation scenario as an example to illustrate the conversation quality inspection method of the embodiments of the present application.

[0089] The conversation quality inspection method of this embodiment includes the following steps:

[0090] Step S402: Select a complete conversation that needs to be quality inspected for online consultation from the conversation library of online consultation.

[0091] In this embodiment, the conversations generated during the online consultation process are stored in the conversation library. In a feasible manner, one online consultation order can correspond to one online consultation conversation. For example, one order number corresponds to one complete conversation of online consultation.

[0092] In addition, if the dialogue contains audio, corresponding text can be obtained through conversion using technologies such as ASR. Moreover, in this embodiment, the multiple complete dialogues stored in the dialogue library are dialogues with role information. Or rather, each piece of dialogue data in each complete dialogue contains role information and dialogue content.

[0093] Step S404: Extract dialogue data one by one from the complete dialogue.

[0094] Since the dialogue data contains role information and dialogue content, the extracted dialogue data is a sequence of dialogue data containing role information [(R i , U i ), where 1 ≤ i ≤ N. Here, i represents the i-th piece of dialogue data in the dialogue data sequence, and N is the total number of all dialogue data in the complete dialogue. R i ∈{PAT, DOC} represents the dialogue role, PAT represents the patient, DOC represents the doctor, and U i represents the text content of the i-th dialogue.

[0095] Step S406: Determine the dialogue role; if the i-th piece of dialogue data is the doctor's dialogue data, then execute Step S408; if the i-th piece of dialogue data is the patient's dialogue data, then execute Step S412.

[0096] For example, the current dialogue can be determined as the doctor's dialogue data or the patient's dialogue data based on the role information in the dialogue data.

[0097] Step S408: If the i-th piece of dialogue data is the doctor's dialogue data, use the model to perform multi-dimensional service attitude recognition.

[0098] If the i-th piece of dialogue data is the doctor's dialogue data, that is, (R i , U i ), and R i = DOC, then use U i as the input and input it into a pre-trained multi-dimensional model (there may be multiple or just one) for recognition (including harassment, abuse, other service attitudes, etc.) to obtain the service attitude label output by the model. This label can reflect the doctor's service attitude, including but not limited to being polite, gentle, kind, neutral, enthusiastic, cold, etc. except for harassment and abuse. The specific label settings can be appropriately set by those skilled in the art according to requirements as long as they can reflect the doctor's service attitude.

[0099] Among them, the models used in this step include but are not limited to CNN convolutional neural network models, RNN recurrent convolutional neural network models, Transformer models, and other pre-trained models with service attitude recognition functions, etc.

[0100] Step S410: Determine whether the doctor's reply is compliant; if it is compliant, return to step S404 to continue extracting the next sentence of conversation data; if it is not compliant, execute step S420.

[0101] That is, according to the service attitude recognition result of step S408, determine whether the doctor's reply conforms to the service attitude specification. Among them, the service attitude specification can also be set by those skilled in the art according to the actual situation, and the embodiments of the present application do not limit this either.

[0102] In practical applications, if the doctor's reply is compliant, on the one hand, the compliant result can be output, and on the other hand, return to step S404 to continue execution. Of course, the result can also be output only when the doctor's reply is not compliant.

[0103] Step S412: If the i-th conversation data is the patient's conversation data, use the question model to identify the patient's conversation data.

[0104] If the i-th conversation data is the patient's conversation data, that is, (R i , U i ), R i = PAT, then use U i as the input and input it into the pre-trained question model for identification to identify whether the patient's conversation data is question data related to medical consultation. According to the output of the question model, obtain the prediction label L i ∈ {Q, O}, where Q represents that U i is question data related to medical consultation, and 0 represents other situations.

[0105] Among them, the models used in this step include but are not limited to CNN convolutional neural network models, RNN recurrent convolutional neural network models, Transformer models, and other pre-trained models with the function of identifying question data, etc.

[0106] Step S414: Determine whether the patient's conversation data is question data according to the recognition result; if it is not question data, return to step S404; if it is question data, execute step S416.

[0107] According to the prediction label L i ∈ {Q, O} output by the question model, if L i = 0, return to step S404 to continue extracting the next conversation data; if L i = Q, enter the next step.

[0108] Step S416: When the current conversation data is the patient's question data, extract the patient's question data, the intermediate conversation data between the question data and the doctor's target reply data, and the doctor's target reply data from the complete conversation and splice them together.

[0109] For example, if it is determined that the current conversation data is the patient's question data, i.e., (R i , U i , L i ), R i = PAT, L i = Q, select the doctor's conversation data after the patient's conversation data from the complete conversation in sequence, and determine the doctor's target reply data from it. Combine the conversation data of this patient (R i , U i ), the doctor's target reply data (R j , U j ), and the intermediate conversation data between the two conversation data {(R k , U k )} to obtain [(R i , U i ), {(R k , U k )}, (R j , U j ), 1 ≤ i < k < j ≤ N.

[0110] Step S418: Identify using a question-answer matching model.

[0111] In specific implementation, the sentence encoding module can be used to perform the first encoding on the conversation content part in the combined conversation data obtained in step S416, i.e., [U i , {U k}, U j to obtain [H i , {H k}, H j , and then superimpose the second encoding corresponding to the role information to obtain [H i + RH i , {H k + RH k}, H j + RH j , where RH i is the role encoding corresponding to the role information of the i-th conversation data.

[0112] Among them, the above first encoding and second encoding methods include but are not limited to using word vectors, using pre-trained models, etc.

[0113] In addition, when performing encoding, in a feasible method, the first encoding and the second encoding can be directly added. For example: U: In Hefei; R: [Patient's role information], then the first encoding can be performed on the part of "U: In Hefei", the second encoding can be performed on the part of "R: [Patient's role information]", and then the first encoding and the second encoding are added.

[0114] In another feasible way, character information prompts can be added to the original content. For example, the original dialogue data is: U: From Hefei; R:

Patient's role information

Patient's role information

Patient's role information

[0115] Use the added encoding as the input of the question-and-answer matching model to identify whether there is reply data in the dialogue data of one or more doctors that reasonably answers the patient's question data. That is, to identify whether the doctor effectively replied to the patient's question. If so, on the one hand, the identification result can be output, and on the other hand, step S404 can be returned to continue extracting the next sentence of dialogue data; if not, it indicates that the doctor did not answer the patient's question, and step S420 is executed. Of course, it is also possible to only output the result when the doctor did not answer the patient's question and execute step S420.

[0116] Among them, the models used in this step include but are not limited to CNN convolutional neural network models, RNN recurrent convolutional neural network models, Transformer models, and other pre-trained models with question-and-answer matching and recognition functions.

[0117] Step S420: Evaluate the service quality of the doctor based on the doctor's service attitude and the reply to the patient's question.

[0118] For example, the service of the doctor can be scored according to pre-set rules, etc.

[0119] Through this embodiment, on the one hand, various models are used for corresponding identification and detection. The models replace manual labor, improve the quality inspection efficiency, and reduce the quality inspection cost; in addition, compared with various complex rules for identification and detection, the various models in this embodiment have better generalization and recall capabilities, and at the same time, identifying based on the context of the spliced dialogue data can also reduce false recalls; on the other hand, after disassembling the complete dialogue and then inputting it into the model for processing, the difficulty of the model to identify whether the doctor's reply is effective is reduced.

[0120] Embodiment 4

[0121] Refer to Figure 5 , which shows the step flowchart of a dialogue quality inspection method according to Embodiment 4 of the present application.

[0122] In this embodiment, taking the example that the dialogue quality inspection method described in the foregoing embodiment is executed by a server (such as a background server or the cloud, etc.), the dialogue quality inspection method of the present application will be described from the perspective of the client. However, those skilled in the art should understand that if the client has sufficient software and hardware performance, the dialogue quality inspection methods in the above-mentioned multiple embodiments can also be implemented on the client, or part of them can be implemented on the client and part on the server.

[0123] The dialogue quality inspection method of this embodiment includes the following steps:

[0124] Step S502: Send request information for requesting dialogue quality inspection of a dialogue segment of a service interaction.

[0125] Among them, the request information contains information about the dialogue segment to be quality inspected.

[0126] For example, the client receives the input information for quality inspecting the dialogue segment to be quality inspected through a web page or an application program interface, generates corresponding request information through corresponding trigger options (such as a submit button, voice operation, gesture operation, etc.), and sends it to the server. The information of the dialogue segment includes but is not limited to: information of the dialogue order, information of the dialogue identifier (such as name, address, ID number), etc. The specific form of the information of the dialogue segment in the embodiments of the present application is not limited.

[0127] Step S502: Receive feedback information for the request information.

[0128] Among them, the feedback information contains quality inspection result information about the service quality provided by the service provider to the service recipient determined according to the dialogue data in the dialogue segment.

[0129] Based on the information of the dialogue segment, the server can obtain the dialogue segment to be quality inspected, and then can use the dialogue quality inspection method described in any one of the foregoing Embodiments 1 to 3 to process the dialogue segment to be quality inspected to obtain quality inspection result information about the service quality provided by the service provider to the service recipient, such as the evaluation score mentioned above.

[0130] Step S504: Obtain service adjustment information that matches the quality inspection result information, and according to the service adjustment information, instruct to replace the service provider or send service behavior adjustment instruction information to the service provider.

[0131] Among them, the service adjustment information can be set by those skilled in the art according to specific service scenarios. If the service quality is very poor, in addition to requiring the service provider to adjust their service behavior, the service provider can also be directly replaced, such as not dispatching service orders to them for a certain period or allocating service receivers, etc. If there are certain defects in the service quality but it has not reached the bottom line of adjustment that requires replacement, corresponding service behavior adjustment instruction information can be sent to the service provider to require the service provider to adjust their service behavior according to this instruction information and provide better service to the service receiver. For example, adjusting the words and tone of the conversation, etc.

[0132] Through this embodiment, according to actual needs, the server can be requested to perform quality inspection on the service quality of the service provider based on the conversation segments in the interactive service. It can be widely applied to various interactive service scenarios, has good applicability and compatibility, and improves the efficiency of service quality inspection.

[0133] Embodiment Five

[0134] Refer to Figure 6A , which shows the step flowchart of a dialogue quality inspection method according to Embodiment Five of the present application.

[0135] Taking online medical consultation as an example in this embodiment, it is still assumed that the dialogue quality inspection method described in the foregoing embodiment is executed by the server (such as a background server or the cloud, etc.), and the dialogue quality inspection method of the present application is described from the perspective of the client.

[0136] Step S602: Send request information for requesting quality inspection of the service quality of the doctor for the dialogue segment of the online medical consultation.

[0137] Among them, the request information contains information about the dialogue segment.

[0138] For example, the client can receive the information input for quality inspection of the dialogue segment to be quality inspected through the web page of the online medical consultation or the interface of the online medical consultation application program, and generate corresponding request information to send to the server. The information of the dialogue segment includes but is not limited to: information of the online medical consultation order, information of the online medical consultation identifier (such as name, address, ID number), etc. The specific form of the information of the dialogue segment in the embodiments of the present application is not limited.

[0139] Step S604: Receive feedback information for the request information.

[0140] Among them, the feedback information contains the quality inspection result information of the service quality provided by the doctor to the patient determined according to the dialogue data in the dialogue segment.

[0141] Based on the information of the conversation segment in the online consultation, the server can obtain the conversation segment to be quality-checked, and then can process the conversation segment to be quality-checked by using the conversation quality-checking method described in any one of the foregoing Embodiments 1 to 3, so as to obtain the quality-check result information of the service quality of the online consultation service provided by the doctor to the patient, such as the evaluation score described above.

[0142] Step S606: Obtain service adjustment information that matches the quality-check result information, and according to the service adjustment information, instruct to replace the doctor serving the patient or send service behavior adjustment instruction information to the doctor.

[0143] Among them, the service adjustment information can be set by those skilled in the art according to the service scenario of the online consultation, and the embodiments of the present application do not limit this.

[0144] Hereinafter, an exemplary description of the above process will be given with a scenario example, as Figure 6B shown.

[0145] Figure 6B In this example, a user with the authority to quality-check the online consultation service quality inputs the information of the conversation segment to be quality-checked through the interface of the online consultation application program on the client side. In this example, the information of the conversation segment to be quality-checked is indicated by the online consultation order number. Suppose the online consultation order number is "12345". After the user clicks the "Quality Check" button, the client generates a request message according to the online consultation order number and sends it to the server. After receiving the request message, the server obtains the online consultation order number from it and queries the corresponding conversation segment from the database. Furthermore, the service quality of the conversation segment is quality-checked by using the conversation quality-checking method described in any one of the foregoing Embodiments 1 to 3. In this example, suppose after the service quality check, if it is determined that the service quality of the doctor corresponding to the online consultation order number is 8 points (out of 10 points), then this score is fed back to the client. After receiving this score, the client gives a prompt message of "This doctor is great. I hope to continue to work hard~". In another case, suppose it is determined that the service quality of the doctor corresponding to the online consultation order number is 6 points, then the server still feeds back this score to the client. After receiving this score, the client gives an adjustment instruction message of "The service quality needs to be improved. Please adjust your service behavior and avoid using negative words." If the person requesting the quality check is a doctor, then the doctor can adjust his subsequent service behavior according to this adjustment instruction message.

[0146] Through this embodiment, according to the actual needs in the online consultation service, the server can be requested to quality-check the service quality of the doctor based on the conversation segment in the online consultation interaction service, which can effectively improve the efficiency of the online consultation service quality check.

[0147] Embodiment 6

[0148] Refer toFigure 7A , showing a step flowchart of a dialogue quality inspection method according to Embodiment 6 of the present application.

[0149] In this embodiment, taking customer service as an example, it is still assumed that the dialogue quality inspection method described in the foregoing embodiment is executed by a server (such as a background server or a cloud, etc.), and the dialogue quality inspection method of the present application is described from the perspective of the client.

[0150] Step S702: Send a request message for requesting service quality inspection of the customer service.

[0151] Among them, the request message contains identification information of the customer service.

[0152] Generally speaking, the customer service can be provided in the form of online text interaction. For example, in e-commerce, customer service staff interact with users before or after sales through an online customer service interface, etc., but it is not limited to the form of text interaction. This interaction can also be in the form of voice or a mixture of voice and text. When the interaction content includes voice, the voice can be converted into text and then the dialogue quality inspection of this embodiment can be carried out. And in some scenarios, such as telephone customer service, it is completely in the form of voice dialogue. In this scenario, the voice dialogue can also be converted into a text dialogue to carry out the dialogue quality inspection of this embodiment.

[0153] A complete service usually corresponds to an identification, that is, the identification information of the customer service. Based on this identification information, the customer service to be subjected to service quality inspection and its corresponding dialogue segment can be determined.

[0154] Step S704: Receive a feedback message for the request message.

[0155] Among them, the feedback message contains quality inspection result information of the service quality of the customer service determined according to the identification information.

[0156] In an optional specific implementation, the feedback message contains a dialogue segment corresponding to the customer service determined according to the identification information, and quality inspection result information of the service quality of the customer service determined according to the dialogue data in the dialogue segment.

[0157] In this embodiment, after the server obtains the identification information of the customer service, it can search for the corresponding dialogue segment of the customer service based on this identification information. For example, the dialogue segment of the complete dialogue of a certain customer service identified by a certain ID. Furthermore, based on this dialogue segment, the server can use the dialogue quality inspection method described in any one of Embodiments 1 to 3 of the foregoing embodiment to perform quality inspection processing on this dialogue segment to obtain quality inspection result information of the service quality of this customer service, such as the evaluation score described above, and feedback it to the client.

[0158] Step S706: Obtain service adjustment information that matches the quality inspection result information, and according to the service adjustment information, instruct to replace the customer service staff or send service behavior adjustment instruction information to the customer service staff.

[0159] Among them, the service adjustment information can be set by those skilled in the art according to the service scenario of customer service, and the embodiments of the present application do not limit this.

[0160] The following uses a scenario example to exemplarily illustrate the above process, as Figure 7B shown.

[0161] Figure 7B In [the figure], an input box and a quality inspection trigger button for requesting quality inspection of a certain customer service are displayed on the interface. The user can enter the identification information of the customer service for which quality inspection is requested, such as S1001, in this input box. After the user clicks the "Quality Inspection" button, a request message for requesting service quality inspection of the customer service is triggered and sent to the server. The identification information of the customer service to be quality inspected is carried in this request message, such as "S1001". After receiving this request message, the server obtains the identification information of the customer service from it. Furthermore, according to this identification information, it searches in the database storing multiple customer services and their corresponding conversation segments for the conversation segment corresponding to "S1001". After obtaining the conversation segment corresponding to "S1001", service quality inspection is performed on this conversation segment based on any one of the conversation quality inspection methods described in the foregoing Embodiments 1 to 3. In this example, it is assumed that after service quality inspection, if it is determined that the service quality of this customer service is 7 points (out of 10 points), then this score is fed back to the client. After receiving this score, the client gives a prompt message of "Your service needs improvement, please pay attention~".

[0162] It can be seen that through this embodiment, the quality inspection of the service quality of various customer services, including online customer service and telephone customer service, etc., can be effectively realized, and the efficiency of customer service quality inspection is effectively improved.

[0163] Embodiment 7

[0164] Referring to Figure 8 , a schematic structural diagram of an electronic device according to Embodiment 6 of the present application is shown. The specific implementation of the electronic device is not limited in the specific embodiments of the present application.

[0165] As Figure 8 shown, the electronic device may include: a processor 802, a communications interface 804, a memory 806, and a communication bus 808.

[0166] Among them:

[0167] The processor 802, communication interface 804, and memory 806 communicate with each other via a communication bus 808.

[0168] The communication interface 804 is used to communicate with other electronic devices or servers.

[0169] The processor 802 is used to execute the program 810, and specifically can execute the relevant steps in the above-mentioned embodiments of the dialogue quality inspection method.

[0170] Specifically, the program 810 may include program code, and the program code includes computer operation instructions.

[0171] The processor 802 may be a CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application. One or more processors included in the intelligent device may be of the same type of processor, such as one or more CPUs; or may be of different types of processors, such as one or more CPUs and one or more ASICs.

[0172] The memory 806 is used to store the program 810. The memory 806 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk memory.

[0173] The program 810 is specifically used to cause the processor 802 to execute the operations corresponding to the dialogue quality inspection method described in any one of the foregoing Embodiments 1 to 6.

[0174] For the specific implementation of each step in the program 810, reference may be made to the corresponding steps and descriptions in the corresponding units in the above-mentioned embodiments of the dialogue quality inspection method, and there are corresponding beneficial effects, which will not be elaborated here. Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described devices and modules can refer to the corresponding process descriptions in the foregoing method embodiments, and will not be elaborated here.

[0175] The embodiments of the present application also provide a computer program product, including computer instructions, and the computer instructions instruct a computing device to execute the operations corresponding to any one of the above-mentioned multiple method embodiments of the dialogue quality inspection method.

[0176] It should be noted that according to the needs of implementation, each component / step described in the embodiments of the present application can be split into more components / steps, or two or more components / steps or partial operations of components / steps can be combined into new components / steps to achieve the purpose of the embodiments of the present application.

[0177] The method according to the embodiments of the present application can be implemented in hardware, firmware, or be implemented as software or computer code that can be stored in a recording medium (such as a CD ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or be implemented as computer code that is originally stored in a remote recording medium or a non-transitory machine-readable medium and downloaded through a network and will be stored in a local recording medium. Thus, the method described herein can be stored as such software processing on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component (such as RAM, ROM, flash memory, etc.) that can store or receive software or computer code. When the software or computer code is accessed and executed by the computer, the processor, or the hardware, the dialogue quality inspection method described herein is implemented. In addition, when a general-purpose computer accesses the code for implementing the dialogue quality inspection method shown herein, the execution of the code converts the general-purpose computer into a dedicated computer for executing the dialogue quality inspection method shown herein.

[0178] Those of ordinary skill in the art can realize that the units and method steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the embodiments of the present application.

[0179] The above embodiments are only used to illustrate the embodiments of the present application, rather than to limit the embodiments of the present application. Those of ordinary skill in the relevant technical field can make various changes and modifications without departing from the spirit and scope of the embodiments of the present application. Therefore, all equivalent technical solutions also belong to the scope of the embodiments of the present application. The patent protection scope of the embodiments of the present application shall be defined by the claims.

Claims

1. A method for dialogue quality inspection, including: extracting dialogue data to be processed from a dialogue segment to be quality inspected, where the dialogue segment contains multiple pieces of dialogue data for service interaction between a service provider and a service recipient; if the dialogue data to be processed is the problem data of the service recipient, then determining the target dialogue data of the service provider located after the problem data from the dialogue segment, and the target dialogue data is the reply data of the problem data; concatenating the problem data, the intermediate dialogue data between the problem data and the target dialogue data of the service provider, and the target dialogue data of the service provider in the dialogue segment; performing a reply detection on whether there is a reply to the problem data for the concatenated dialogue data, and determining the service quality of the service provider according to the detection result; wherein, the extracting dialogue data to be processed from the dialogue segment to be quality inspected includes: obtaining the dialogue segment to be quality inspected; extracting each piece of dialogue in the dialogue segment one by one, and generating a dialogue data sequence including the role information and the dialogue according to the time sequence of the dialogue in the dialogue segment and the role information corresponding to each piece of dialogue; extracting the dialogue data to be processed from the dialogue data sequence.

2. The method according to claim 1, wherein, the method further includes: obtaining the role information of the dialogue data to be processed; judging whether the dialogue data to be processed is the dialogue data of the service provider or the dialogue data of the service recipient according to the role information; if the dialogue data to be processed is the dialogue data of the service provider, then performing a service attitude quality inspection on the dialogue data to be processed and obtaining a service attitude quality inspection result; if the dialogue data to be processed is the dialogue data of the service recipient, then performing the operation of determining the target dialogue data of the service provider located after the problem data from the dialogue segment if the dialogue data to be processed is the problem data of the service recipient.

3. The method according to any one of claims 1-2, wherein, the determining the target dialogue data of the service provider located after the problem data from the dialogue segment if the dialogue data to be processed is the problem data of the service recipient includes: if it is determined that the dialogue data to be processed is the dialogue data of the service recipient according to the role information of the dialogue data to be processed, then performing an identification detection on whether the dialogue data to be processed is problem data; if it is determined that the dialogue data to be processed is problem data according to the identification detection result, then determining the target dialogue data of the service provider whose time sequence is located after the problem data from the dialogue segment.

4. The method according to claim 3, wherein, the concatenating the problem data, the intermediate dialogue data between the problem data and the target dialogue data of the service provider, and the target dialogue data of the service provider in the dialogue segment includes: According to the time sequence of the conversation data in the conversation segment, splice the problem data, the intermediate conversation data whose time sequence is between the problem data and the target conversation data of the service provider, and the target conversation data of the service provider.

5. The method according to claim 3, wherein, the reply detection for whether there is a reply to the problem data in the spliced conversation data includes: performing a first encoding on the conversation content in the spliced conversation data to obtain conversation content encoding data; performing a second encoding on the role information in the spliced conversation data to obtain role encoding data; splicing the conversation content encoding data and the role encoding data to obtain spliced encoding data; performing reply detection on whether there is a reply to the problem data based on the spliced encoding data.

6. The method according to claim 5, wherein, the performing a first encoding on the conversation content in the spliced conversation data to obtain conversation content encoding data includes: adding corresponding role information to the conversation content in the spliced conversation data; performing a first encoding on the conversation content after adding the role information to obtain conversation content encoding data.

7. A conversation quality inspection method, including: sending request information for requesting conversation quality inspection of a conversation segment of a service interaction, wherein the request information contains information about the conversation segment to be quality inspected; receiving feedback information for the request information, wherein the feedback information contains quality inspection result information of the service quality of the service provided by the service provider to the service recipient determined according to the conversation data in the conversation segment, and the service quality is determined by the method according to any one of claims 1-6; obtaining service adjustment information matching the quality inspection result information, and instructing to replace the service provider or sending service behavior adjustment instruction information to the service provider according to the service adjustment information.

8. A conversation quality inspection method, including: sending request information for requesting quality inspection of the service quality of a doctor for an online consultation conversation segment, wherein the request information contains information about the conversation segment; receiving feedback information for the request information, wherein the feedback information contains quality inspection result information of the service quality of the service provided by the doctor to the patient determined according to the conversation data in the conversation segment, and the service quality is determined by the method according to any one of claims 1-6; obtaining service adjustment information matching the quality inspection result information, and instructing to replace the doctor serving the patient or sending service behavior adjustment instruction information to the doctor according to the service adjustment information.

9. A conversation quality inspection method, including: sending request information for requesting service quality inspection of a customer service, wherein the request information contains identification information of the customer service; Receive feedback information for the request information, where the feedback information includes quality inspection result information of the service quality of the customer service determined according to the identification information, and the service quality is determined by the method described in any one of claims 1-6; Obtain service adjustment information matching the quality inspection result information, and according to the service adjustment information, instruct to replace the customer service personnel or send service behavior adjustment instruction information to the customer service personnel.

10. The method according to claim 9, wherein, the feedback information includes a dialogue segment corresponding to the customer service determined according to the identification information, and quality inspection result information of the service quality of the customer service determined according to the dialogue data in the dialogue segment.

11. An electronic device, comprising: a processor, a memory, a communication interface and a communication bus, and the processor, the memory and the communication interface complete mutual communication through the communication bus; the memory is used for storing at least one executable instruction, and the executable instruction causes the processor to execute the operations corresponding to the dialogue quality inspection method described in any one of claims 1-10.

12. A computer storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the dialogue quality inspection method described in any one of claims 1-10.

13. A computer program product, including computer instructions, and the computer instructions instruct a computing device to execute the operations corresponding to the dialogue quality inspection method described in any one of claims 1-10.

Images