NLP-based customer service dialogue method and device, computer device and storage medium

Abstract

The application relates to the technical field of financial technology, and discloses a customer service dialogue method and device based on NLP, computer equipment and a storage medium, which comprises the following steps: obtaining the question type of a question text of a user, then judging whether the difficulty level of the question type exceeds a preset difficulty level threshold value, when the difficulty level of the question type exceeds the difficulty level threshold value, sending the question text to an artificial terminal to obtain a target answer text, and when the difficulty level does not exceed the difficulty level threshold value, matching the target answer text of the question text from a question knowledge base. It can be seen that when the difficulty level of the question type of the question text of the user exceeds the preset difficulty level threshold value, the question text is sent to the artificial terminal, the intelligent switching of the artificial terminal is realized, the purpose of improving the artificial switching efficiency can be achieved, compared with the prior art, the user does not need to send a keyword to actively switch the artificial terminal, and the experience satisfaction of the user in the financial field is improved.

Description

Technical Field

[0001] This invention relates to the field of financial technology, and in particular to a customer service dialogue method, apparatus, computer equipment, and storage medium based on NLP. Background Technology

[0002] With the rapid development of science and technology, intelligent customer service technology has been widely used in the financial field, such as e-commerce customer service, life insurance customer service, and auto insurance customer service. For example, a life insurance customer sends their questions to the server through the life insurance consultation page on the client side. The server's intelligent customer service system then intelligently generates a response based on the user's question and sends it to the user's client.

[0003] However, existing intelligent customer service technology primarily relies on intelligent customer service systems to answer customer questions. Only when a customer sends keywords such as "need human assistance" will they be transferred to a human agent. In practice, in many cases, such as when the customer's question is complex or when the customer is emotionally unstable, the intelligent customer service system struggles to generate a satisfactory response. Even if the customer sends keywords to be transferred to a human agent, the customer experience suffers.

[0004] Therefore, there is an urgent need for an efficient method to transfer customers to human customer service to improve customer experience satisfaction. Summary of the Invention

[0005] This invention provides a customer service dialogue method, apparatus, computer equipment, and storage medium based on NLP to solve the technical problem that the poor efficiency of manual transfer in existing intelligent customer service systems leads to reduced customer experience satisfaction.

[0006] A customer service dialogue method based on NLP, the method comprising: Obtain the user's question text; The question text is classified by a question classification model to obtain a question type classification result; the question type classification result includes the question type of the question text and the difficulty level of the question type. Determine whether the difficulty level of the problem type exceeds a preset difficulty level threshold; If the difficulty level of the question type exceeds the difficulty level threshold, the question text is sent to a human terminal to obtain the target answer text; If the difficulty level does not exceed the difficulty level threshold, a question pair corresponding to the question text is matched from the question knowledge base based on the question type, and the question answer text in the question pair is used as the target answer text.

[0007] Optionally, the problem type classification result may also include the confidence level of the problem type in the above method. After classifying the question text using a question classification model to obtain the question type classification result, the method further includes: Determine whether the confidence level of the problem type is less than a preset confidence threshold; If the confidence level of the question type is less than the confidence threshold, the question text is marked as a new question type, and the question text is recorded in the question knowledge base; If the confidence level of the question type is greater than the confidence threshold, the first objective step is executed. The first objective step is to determine whether the difficulty level of the question type exceeds a preset level threshold.

[0008] Optionally, after classifying the question text using a question classification model to obtain the question type classification result, the above method further includes: The question text is subjected to sentiment recognition based on the corresponding sentiment classifier matched to the question type, and the sentiment recognition result of the question text is obtained. The sentiment recognition result is used to perform a first modification process on the question answer text in the question pair to obtain the target answer text.

[0009] Optionally, after performing sentiment recognition on the question text according to the question type using a corresponding sentiment classifier to obtain the sentiment recognition result of the question text, the above method further includes: Based on the emotion recognition results, determine whether the emotional complexity of the question text exceeds a preset complexity threshold; If the emotional complexity of the question text exceeds a preset complexity threshold, the second objective step is executed: the question text is sent to a human terminal to obtain the target response text; If the emotional complexity of the question text does not exceed a preset complexity threshold, the question answer text in the question pair is subjected to a first modification process based on the emotional recognition result to obtain the target answer text.

[0010] Optionally, in the above method, matching the question pair corresponding to the question text from the question knowledge base based on the question type includes: Extract the customer's actual inquiry question from the question text; Match the question pairs that correspond to the actual consultation question from the question knowledge base.

[0011] Optionally, in the above method, sending the question text to a human terminal to obtain the target answer text includes: The question text is sent to a human terminal to receive a customer service response. The customer service response text is then subjected to a second modification process to obtain the target response text.

[0012] Optionally, in the above method, sending the question text to a human terminal to obtain the target answer text includes: Based on the question text, the user's preferred consultation method can be determined. Based on the consultation method preference, a corresponding human terminal is matched to the user to obtain the target answer text; the human terminal includes at least an online customer service terminal, a voice customer service terminal, and a video customer service terminal.

[0013] A customer service dialogue device based on NLP, comprising: The question text acquisition unit is used to acquire the user's question text; The question type classification unit is used to classify the question text by type using a question classification model, and obtain the question type classification result of the question text; the question type classification result includes the question type of the question text and the difficulty level of the question type. The difficulty level determination unit is used to determine whether the difficulty level of the problem type exceeds a preset difficulty level threshold. The human terminal transfer unit is used to send the question text to a human terminal to obtain the target answer text when the difficulty level of the question type exceeds a preset difficulty level threshold.

[0014] The answer text matching unit is used to match a question pair corresponding to the question text from the question knowledge base based on the question type when the difficulty level does not exceed a preset difficulty level threshold, and to use the question answer text in the question pair as the target answer text. A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the NLP-based customer service dialogue method as described above.

[0015] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the NLP-based customer service dialogue method described above.

[0016] The aforementioned NLP-based customer service dialogue method, device, computer equipment, and storage medium determine the question type of the user's question text. Then, based on whether the difficulty level of the question type exceeds a preset difficulty level threshold, it determines whether the question text needs to be sent to a human terminal. If the difficulty level exceeds the threshold, the question text is sent to the human terminal, and the target answer text is obtained from the human terminal. If the difficulty level does not exceed the threshold, only question pairs corresponding to the question text can be matched from the question knowledge base, and the answer text in the question pair is used as the target answer text. Therefore, this invention, by classifying the user's question text into question types and sending the question text to the human terminal when the difficulty level exceeds a preset difficulty level threshold, achieves intelligent transfer to the human terminal, improving the efficiency of human transfer. Compared to existing technologies, it eliminates the need for users to actively send keywords to be transferred to a human terminal, which is beneficial for improving the user experience satisfaction in the financial sector. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the implementation of an NLP-based customer service dialogue method according to an embodiment of the present invention. Figure 2 This is another implementation flowchart of a customer service dialogue method based on NLP disclosed in an embodiment of the present invention; Figure 3 This is a flowchart illustrating another implementation of an NLP-based customer service dialogue method disclosed in an embodiment of the present invention. Figure 4 This is a flowchart illustrating another implementation of an NLP-based customer service dialogue method disclosed in an embodiment of the present invention. Figure 5 This is a partial implementation flowchart of an NLP-based customer service dialogue method disclosed in an embodiment of the present invention; Figure 6 This is a partial implementation flowchart of an NLP-based customer service dialogue method disclosed in an embodiment of the present invention; Figure 7 This is a partial implementation flowchart of an NLP-based customer service dialogue method disclosed in an embodiment of the present invention; Figure 8This is a schematic diagram of the structure of a customer service dialogue device based on NLP, as disclosed in an embodiment of the present invention. Figure 9 This is a schematic diagram of the structure of a basic computer device disclosed in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0021] It should also be understood that the term “and / or” as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0022] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0023] Furthermore, in the description of this invention and the appended claims, the terms "first," "second," "third," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.

[0024] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of the invention include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0025] This invention discloses a customer service dialogue method, apparatus, computer device, and storage medium based on NLP. It determines the question type by analyzing the user's question text, and then decides whether to send the question text to a human terminal based on whether the difficulty level of the question type exceeds a preset difficulty level threshold. If the difficulty level exceeds the threshold, the question text is sent to the human terminal, where the target answer text is obtained. If the difficulty level does not exceed the threshold, only question pairs corresponding to the question text can be matched from the question knowledge base, and the answer text in the question pair is used as the target answer text. Therefore, this invention achieves intelligent transfer of human terminals by classifying user question texts by question type and sending the question text to the human terminal when the difficulty level exceeds a preset difficulty level threshold. This improves the efficiency of human transfer. Compared to existing technologies, it eliminates the need for users to actively send keywords to be transferred to a human terminal, which is beneficial for improving the user experience satisfaction in the financial sector.

[0026] It should be understood that in this invention, the user's question text will only be transferred to a human terminal to obtain the target answer text from a human customer service representative when the difficulty level of the question type exceeds a preset difficulty level threshold. When the difficulty level of the user's question text does not exceed the preset difficulty level threshold, the intelligent customer service system will generate the target answer text based on the question text, or query the corresponding target answer text from the question knowledge base. This improves the efficiency of human terminal transfers and also achieves the goal of efficiently utilizing human resources.

[0027] like Figure 1 As shown in the figure, this embodiment discloses a flowchart of an NLP-based customer service dialogue method. This method is applicable to electronic devices with natural language processing capabilities, such as mobile phones, tablets, personal computers, and servers. The specific steps are described below: S101: Obtain the user's question text.

[0028] In this implementation, the question text in this embodiment is entered by the user on the customer service page of the client. The client can be an electronic device such as a user's mobile phone, laptop, or tablet. The user can generate the question text through methods such as Pinyin input or voice input, and send the question text to the customer service page of the client so that the question text reaches the corresponding server, such as a life insurance service terminal in the financial field. In this way, the user's question text can be obtained.

[0029] For example, taking a mobile phone as the client, a user's mobile phone has a relevant insurance APP of a financial company installed. The user opens the customer service consultation page in the APP and edits the question they want to ask into text and sends it to the consultation page. In this way, the APP's server or service terminal can obtain the user's question text.

[0030] S102: Classify the question text by type using a question classification model to obtain the question type classification result of the question text.

[0031] The problem type classification result includes the problem type in the problem text and the difficulty level of the problem type. It can be understood that the difficulty level of the problem type can be a pre-assigned difficulty level; that is, when the problem type classification result is obtained, the difficulty level of the problem type can be determined based on the problem type in the classification result. In the financial field, the problem types in this embodiment can include order inquiries, technical support, complaints and suggestions, etc., where the difficulty level of complaints and suggestions can be higher than that of order inquiries and technical support.

[0032] The question text is preprocessed to obtain its word vector representation. This word vector representation is then input into a question classification model to obtain the question type classification result. The preprocessing includes, but is not limited to, removing stop words, word segmentation, part-of-speech tagging, and named entity recognition.

[0033] In this implementation, the question classification model can be a text matching model based on the attention mechanism. The word vector representation obtained by preprocessing the question text is input into the text matching model to obtain the question type classification result of the question text.

[0034] Furthermore, in this embodiment, the difficulty level of a question type can be obtained based on a pre-defined question type difficulty level table. This table stores question types and their corresponding difficulty levels in key-value pairs. Based on the question type in the question type classification results, the question type difficulty level table is queried to obtain the difficulty level of the question type.

[0035] S103: Determine whether the difficulty level of the problem type exceeds the preset difficulty level threshold.

[0036] It should be understood that, depending on the type of question, the difficulty level corresponding to the question type may be level 1, level 2, or level 3, etc. When the difficulty level of the question type exceeds the preset difficulty level threshold, step S104 is executed, that is, the question text is sent to the human terminal to obtain the target answer text; when the difficulty level of the question type does not exceed the preset difficulty level threshold, step S105 is executed, that is, based on the question type, a question pair corresponding to the question text is matched from the question knowledge base, and the question answer text in the question pair is used as the target answer text.

[0037] It should be noted that the above difficulty levels, such as Level 1, Level 2, or Level 3, are merely illustrative examples in this embodiment. The difficulty levels may be Level 1, Level 2, Level 3, Level 4, or Level 5, and the classification of difficulty levels is not limited in this embodiment.

[0038] For example, taking a difficulty level threshold of 3 as an example, if the difficulty level of the question text is 4, then it can be determined that the difficulty level of the question type exceeds the difficulty level threshold, and the question text is sent to the human terminal. If the difficulty level of the question text is 2, then it can be determined that the difficulty level of the question type does not exceed the difficulty level threshold, and a question pair corresponding to the question text is matched from the question knowledge base based on the question type, and the question answer text in the question pair is used as the target answer text.

[0039] S104: Send the question text to the human terminal to obtain the target answer text.

[0040] It should be understood that the human customer service terminal has human customer service personnel who will write the target answer text based on the text content of the question sent to the human customer service terminal, and return the target answer text to the user's client through the human customer service terminal.

[0041] S105: Based on the question type, match the question pair corresponding to the question text from the question knowledge base, and use the question answer text in the question pair as the target answer text.

[0042] It should be understood that a question pair refers to a pair of question text and question answer text. Each question pair in the question knowledge base has a corresponding type tag. Question pairs with the same type tag as the question type are matched from the question knowledge base. Then, the target question pair that is closest to the question text is matched from these question pairs with the same type tag. The question answer text in the target question pair is selected as the target answer text, and the intelligent customer service sends the target answer text to the user's consultation page on the client.

[0043] In summary, this invention discloses an NLP-based customer service dialogue method. By judging the question type of the user's question text, the method determines whether the question text needs to be sent to a human operator based on whether the difficulty level of the question type exceeds a preset difficulty level threshold. If the difficulty level exceeds the threshold, the question text is sent to the human operator, and the target answer text is obtained from the human operator. If the difficulty level does not exceed the threshold, only question pairs corresponding to the question text can be matched from the question knowledge base, and the answer text in the question pair is used as the target answer text. Therefore, this embodiment achieves intelligent transfer of human operators by classifying the user's question text by question type and sending the question text to the human operator when the difficulty level exceeds the preset difficulty level threshold. This improves the efficiency of human operator transfer. Compared to existing technologies, it eliminates the need for users to actively send keywords to be transferred to a human operator, which is beneficial for improving the user experience satisfaction in the financial sector.

[0044] based on Figure 1 In a specific implementation, the following steps may be included after step S102, such as... Figure 2 As shown: In one embodiment, the problem type classification result may also include the confidence level of the problem type. The confidence level represents the credibility of the problem type classified by the problem classification model. The higher the confidence level, the higher the accuracy of the problem type.

[0045] S106: Determine whether the confidence level of the question type is less than the preset confidence level threshold.

[0046] If the confidence level of the question type is less than the preset confidence threshold, proceed to step S107, which marks the question text as a new question type and records it, and proceed to step S104, which sends the question text to the human terminal to obtain the target answer text; if the confidence level of the question type is greater than the preset confidence threshold, proceed to step S103, which determines whether the difficulty level of the question type exceeds the preset difficulty level threshold.

[0047] S107: Mark the question text as a new question type and record the question text in the question knowledge base.

[0048] It should be understood that when the confidence level of a question type is less than the preset confidence threshold, it is highly likely that a new question text has appeared. In this case, the intelligent customer service cannot match the target answer text from the question knowledge base and needs to send the question text to a human customer service representative for a response. However, to enable the intelligent customer service to provide an intelligent response when encountering this type of question again without transferring to a human representative, the question text can be marked as a new question type and recorded in the question knowledge base. This, combined with the human customer service representative's response (i.e., the target answer question), forms a question pair, which is also stored in the question knowledge base. Thus, when the intelligent customer service encounters these same questions again, it can find the corresponding target answer text from the question knowledge base.

[0049] In summary, the intelligent customer service system in this embodiment can adaptively learn from new question texts and the target response texts of human customer service personnel, which can greatly improve the adaptability and flexibility of the intelligent customer service system.

[0050] based on Figure 1 In a specific implementation, the following steps may be included after step S102, as follows: Figure 3 As shown: S108: Match the corresponding sentiment classifier to the question type to perform sentiment recognition on the question text, and obtain the sentiment recognition result of the question text.

[0051] Among them, the emotion recognition result is used to perform the first modification process on the question-answer text in the question pair to obtain the target answer text.

[0052] It should be understood that in the financial field, such as life insurance consultations, when users have a poor experience with the life insurance products they purchased, they inevitably experience negative emotions. In such cases, the user's question text will reflect emotions such as anger and urgency. Therefore, it is necessary to perform sentiment recognition on the question text to determine the user's current emotional category, and then adjust the tone of the response to the customer's question according to the user's current emotional category, thereby achieving the goal of soothing the user's emotions. That is, based on the sentiment recognition results, the question-answer text in the question pair undergoes a first modification process to obtain the target answer text. The first modification process refers to modifying the question-answer text emotionally according to the user's emotional category.

[0053] For example, when a user makes a suggestion, the intelligent customer service system can modify the original message, "We have received your complaint and suggestion. Our customer service staff will contact you by phone to discuss the details later," to "Dear life insurance user, thank you for taking the time to provide us with your valuable suggestion. Our customer service staff will contact you within 24 hours." By modifying the message, the system affirms the user's identity and provides a timeframe, which can better soothe the user's emotions.

[0054] In this specific implementation, a sentiment classifier can be used to classify the user's question text according to sentiment. The user's question text undergoes preprocessing such as stop word removal, word segmentation, part-of-speech tagging, and named entity recognition to obtain word vector representations. These word vector representations are then input into the sentiment classifier to obtain the sentiment classification result of the question text.

[0055] In summary, this embodiment identifies the user's current emotional category by analyzing the user's question text, and then modifies the intelligent customer service's answer text according to the emotional category, thereby improving user experience satisfaction at the emotional level.

[0056] based on Figure 3 In a specific implementation, the following steps may be included after step S108, as follows: Figure 4 As shown: S109: Determine whether the emotional complexity of the question text exceeds the preset complexity threshold based on the emotion recognition results.

[0057] If the emotional complexity of the question text exceeds a preset complexity threshold, step S104 is executed, that is, the question text is sent to the human terminal to obtain the target answer text. If the emotional complexity of the question text does not exceed the preset complexity threshold, step S105 is executed, that is, based on the question type, a question pair corresponding to the question text is matched from the question knowledge base, and the question answer text in the question pair is used as the target answer text.

[0058] Specifically, the emotion classification results in this embodiment include, but are not limited to, one or more emotion categories such as joy, anger, worry, contemplation, grief, fear, and surprise. Normally, after the emotion classifier classifies the question text, it obtains each emotion category and its corresponding confidence level. The emotion classification result is usually the emotion category with the highest confidence level. When the emotion classifier classifies the question text, if the confidence levels of multiple emotion categories exceed a preset emotion confidence threshold, the more emotion categories exceed the threshold, the higher the emotional complexity of the question text. When the emotional complexity exceeds the preset threshold, step S104 is executed, that is, the question text is sent to the human terminal to obtain the target answer text.

[0059] For example, taking a sentiment confidence threshold of 60% as an example, when the confidence levels of 3 sentiment categories exceed 60%, the sentiment complexity of the question text is 3; when the confidence levels of 4 sentiment categories exceed 60%, the sentiment complexity of the question text is 4, and so on. Setting the sentiment complexity threshold to 3 means that when the confidence levels of 4 sentiment categories exceed 60%, the sentiment complexity of the question text is determined to exceed the preset complexity threshold.

[0060] S110: Based on the question type, match the question pair corresponding to the question text from the question knowledge base, and perform the first modification processing on the question answer text in the question pair based on the sentiment recognition result to obtain the target answer text.

[0061] It should be understood that in the financial field, such as life insurance consultations, when users have a poor experience with the life insurance products they purchased, they inevitably experience negative emotions. In such cases, the user's question text will reflect emotions such as anger and urgency. Therefore, it is necessary to perform sentiment recognition on the question text to determine the user's current emotional category, and then adjust the tone of the response to the customer's question according to the user's current emotional category, thereby achieving the goal of soothing the user's emotions. That is, based on the sentiment recognition results, the question-answer text in the question pair undergoes a first modification process to obtain the target answer text. The first modification process refers to modifying the question-answer text emotionally according to the user's emotional category.

[0062] In summary, this embodiment assesses the emotional complexity of the user's question text. When the emotional complexity exceeds a complexity threshold, the question text is sent to a human operator, achieving intelligent transfer and improving the efficiency of human operator transfers. Furthermore, by judging the difficulty level of the question type and sending the question text to a human operator when the difficulty level exceeds a difficulty level threshold, the intelligent transfer methods for human operators are enriched. Compared to existing technologies, this approach eliminates the need for users to actively send keywords to be transferred to a human operator, thus improving user experience satisfaction in the financial sector.

[0063] Based on Figure 1 In its specific implementation, step S105 can be achieved in the following way, such as... Figure 5 As shown: S501: Extract the customer's actual inquiry question from the question text.

[0064] It should be understood that the question text sent by users on the client's consultation page will contain a lot of invalid text in most cases. It is necessary to remove this invalid text from the question text in order to extract the actual question that the user needs to ask.

[0065] For example, if a user's question text is "I have been looking for a long time but I can't find the life insurance renewal window. How can I renew my life insurance?", analysis shows that the user's actual question is "the steps to renew life insurance".

[0066] S502: Match the question pairs from the question knowledge base that correspond to the actual consultation question.

[0067] It should be understood that when traversing the question knowledge base to match corresponding question pairs for question text, the larger the volume of question text data, the lower the matching efficiency, thus reducing the efficiency of matching question pairs. However, the actual consultation question is a "concise version" of the question text. Matching corresponding question pairs from the question knowledge base based on the actual consultation question can greatly reduce the waste of computing resources and improve the efficiency of question pair matching.

[0068] In one embodiment, multiple actual customer inquiries may be extracted from the question text. The more actual inquiries extracted, the higher the complexity of the questions. This makes it difficult for the intelligent customer service to provide a good target answer text all at once. In this case, the intelligent customer service can answer the questions separately for each extracted actual inquiry, or directly send the question text to a human terminal for a human customer service representative to answer. This can further enrich the intelligent transfer methods of the human terminal.

[0069] based on Figure 1 In its specific implementation, step S104 can be achieved through the following steps, such as... Figure 6 As shown: S601: Send the question text to the human terminal and receive the customer service response text.

[0070] It should be understood that the human customer service terminal has human customer service personnel who will write a customer service response text based on the text content of the question sent to the human customer service terminal, and then return the customer service response text to the user's client through the human customer service terminal.

[0071] S602: Perform a second modification on the customer service response text to obtain the target response text.

[0072] Specifically, in this embodiment, the second modification process refers to modifying the customer service response text in a human customer service tone to obtain the target response text in an intelligent customer service tone. Furthermore, the second modification process may also include the first modification process, namely, performing emotional modification processing on the customer service response text.

[0073] Therefore, in this embodiment, by modifying the customer service response text in the tone of a human customer service representative to the target response text in the tone of an intelligent customer service representative, users can feel that the intelligent customer service can already solve their problems well, without the need to transfer to a human customer service representative. This avoids users sending keywords to transfer to a human customer service representative immediately when they want to ask simple questions. Based on this, not only can the frequency of intelligent customer service be increased, but also the efficient use of human resources can be ensured.

[0074] based on Figure 1 In its specific implementation, step S104 can also be achieved through the following steps, such as... Figure 7 As shown: S701: Based on the question text, obtain the user's preferred consultation method.

[0075] In this specific implementation, the user's preferred consultation method can be obtained by capturing keywords in the question text, such as intelligent customer service, voice customer service, and video customer service. When the corresponding keyword is captured, the user's preferred consultation method is obtained. When the corresponding keyword is not captured, one of the preferred consultation methods, such as intelligent customer service, can be defaulted.

[0076] S701: Match the user with the corresponding human terminal based on the consultation method preference to obtain the target answer text.

[0077] Among them, human-operated terminals include at least intelligent customer service terminals, voice customer service terminals, and video customer service terminals.

[0078] When a user prefers voice customer service, a voice customer service terminal will be matched to the user; when a user prefers video customer service, a video customer service terminal will be matched to the user, and so on.

[0079] It should be understood that the aforementioned intelligent customer service terminal, voice customer service terminal, and video customer service terminal are only some of the human terminals in this embodiment, and the specific type of human terminal is not limited in this embodiment.

[0080] In summary, this embodiment utilizes different human terminals to enrich users' consultation options, thereby improving user service satisfaction.

[0081] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0082] like Figure 8 The diagram shown is a structural schematic of a customer service dialogue device based on NLP disclosed in an embodiment of the present invention. This method is applicable to electronic devices with natural language processing capabilities, such as mobile phones, tablets, personal computers, and servers.

[0083] In a specific implementation, the apparatus in this embodiment may include the following units: The question text acquisition unit 801 is used to acquire the user's question text; The question type classification unit 802 is used to classify the question text by type using a question classification model, and obtain the question type classification result of the question text; the question type classification result includes the question type and the difficulty level of the question type in the question text. Difficulty level judgment unit 803 is used to determine whether the difficulty level of the question type exceeds the preset difficulty level threshold; The human terminal switching unit 804 is used to send the question text to the human terminal to obtain the target answer text when the difficulty level of the question type exceeds a preset difficulty level threshold.

[0084] The answer text matching unit 805 is used to match the question pair corresponding to the question text from the question knowledge base based on the question type when the difficulty level does not exceed the preset difficulty level threshold, and to take the question answer text in the question pair as the target answer text. In one implementation, the problem type classification result also includes the confidence level of the problem type; The device in this embodiment also includes the following units: The confidence level judgment unit is responsible for determining whether the confidence level of the question type is less than a preset confidence level threshold. The new question type marking unit is used to mark the question text as a new question type when the confidence level of the question type is less than the confidence level threshold, and to record the question text in the question knowledge base; If the confidence level of the question type is greater than the confidence threshold, execute the first objective step, which is to determine whether the difficulty level of the question type exceeds the preset level threshold.

[0085] In one implementation, the apparatus in this embodiment further includes the following units: The sentiment classification unit is used to match the corresponding sentiment classifier according to the question type to perform sentiment recognition on the question text and obtain the sentiment recognition result of the question text; the sentiment recognition result is used to perform the first modification processing on the question answer text in the question pair to obtain the target answer text.

[0086] In one implementation, the apparatus in this embodiment further includes the following units: The sentiment complexity judgment unit is used to determine whether the sentiment complexity of the question text exceeds a preset complexity threshold based on the sentiment recognition results. If the emotional complexity of the question text exceeds the preset complexity threshold, the manual terminal transfer unit 104 is triggered. If the emotional complexity of the question text does not exceed the preset complexity threshold, the question answer text in the question pair is subjected to the first modification process based on the emotion recognition result to obtain the target answer text.

[0087] In one implementation, the response text matching unit 105 can be used for: Extracting the customer's actual inquiry from the question text; Match question pairs from the question knowledge base that correspond to the actual consultation questions.

[0088] In one implementation, the manual terminal switching unit 104 can be used for: Send the question text to a human operator and receive a response from customer service. The customer service response text undergoes a second modification process to obtain the target response text.

[0089] In one implementation, the manual terminal switching unit 104 can also be used for: Based on the question text, we can determine the user's preferred consultation method. Based on the user's preferred consultation method, a corresponding human customer service terminal is matched to obtain the target response text; the human customer service terminal includes at least online customer service terminal, voice customer service terminal, and video customer service terminal.

[0090] For specific limitations regarding the NLP-based customer service dialogue device, please refer to the relevant limitations of the method mentioned above, which will not be repeated here. Each module in the aforementioned NLP-based customer service dialogue device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0091] This application discloses a computer device, which can be a server, and its internal structure diagram can be as follows: Figure 9 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements an NLP-based customer service dialogue method.

[0092] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Obtain the user's question text; The question text is classified by a question classification model to obtain the question type classification result; the question type classification result includes the question type and the difficulty level of the question type. Determine whether the difficulty level of the question type exceeds a preset difficulty level threshold; If the difficulty level of the question exceeds the difficulty level threshold, the question text will be sent to a human terminal to obtain the target answer text; If the difficulty level does not exceed the difficulty level threshold, the question pair corresponding to the question text is matched from the question knowledge base based on the question type, and the question answer text in the question pair is used as the target answer text.

[0093] Embodiment 4 of this application discloses a computer-readable storage medium that, when executed by a processor in a computer device, enables the computer device to perform various steps of any embodiment of an NLP-based customer service dialogue method disclosed in this invention. The computer-readable storage medium may be non-volatile or volatile.

[0094] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Obtain the user's question text; The question text is classified by a question classification model to obtain the question type classification result; the question type classification result includes the question type and the difficulty level of the question type. Determine whether the difficulty level of the question type exceeds a preset difficulty level threshold; If the difficulty level of the question exceeds the difficulty level threshold, the question text will be sent to a human terminal to obtain the target answer text; If the difficulty level does not exceed the difficulty level threshold, the question pair corresponding to the question text is matched from the question knowledge base based on the question type, and the question answer text in the question pair is used as the target answer text.

[0095] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0096] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0097] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A customer service dialogue method based on NLP, characterized in that, The method includes: Obtain the user's question text; The question text is classified by a question classification model to obtain a question type classification result; the question type classification result includes the question type of the question text and the difficulty level of the question type. Determine whether the difficulty level of the problem type exceeds a preset difficulty level threshold; If the difficulty level of the question type exceeds the difficulty level threshold, the question text is sent to a human terminal to obtain the target answer text; If the difficulty level does not exceed the difficulty level threshold, a question pair corresponding to the question text is matched from the question knowledge base based on the question type, and the question answer text in the question pair is used as the target answer text. After classifying the question text using a question classification model to obtain the question type classification result, the method further includes: The question text is subjected to sentiment recognition based on the corresponding sentiment classifier matched according to the question type, and the sentiment recognition result of the question text is obtained; the sentiment recognition result is used to perform a first modification process on the question answer text in the question pair to obtain the target answer text; After performing sentiment recognition on the question text according to the corresponding sentiment classifier matched to the question type, and obtaining the sentiment recognition result of the question text, the method further includes: Based on the emotion recognition results, it is determined whether the emotional complexity of the question text exceeds a preset complexity threshold; wherein, the emotion recognition results include the confidence scores of multiple emotion categories, and when the confidence scores of multiple emotion categories exceed the preset confidence threshold, the more emotion categories whose confidence scores exceed the confidence threshold, the higher the emotional complexity of the question text. If the emotional complexity of the question text exceeds a preset complexity threshold, the second objective step is executed: the question text is sent to a human terminal to obtain the target response text; If the emotional complexity of the question text does not exceed a preset complexity threshold, the question answer text in the question pair is subjected to a first modification process based on the emotional recognition result to obtain the target answer text.

2. The NLP-based customer service dialogue method as described in claim 1, characterized in that, The problem type classification results also include the confidence level of the problem type; After classifying the question text using a question classification model to obtain the question type classification result, the method further includes: Determine whether the confidence level of the problem type is less than a preset confidence threshold; If the confidence level of the question type is less than the confidence threshold, the question text is marked as a new question type, and the question text is recorded in the question knowledge base; If the confidence level of the question type is greater than the confidence threshold, the first objective step is executed. The first objective step is to determine whether the difficulty level of the question type exceeds a preset level threshold.

3. The NLP-based customer service dialogue method as described in claim 1, characterized in that, The step of matching question pairs corresponding to the question text from the question knowledge base based on the question type includes: Extract the customer's actual inquiry question from the question text; Match the question pairs that correspond to the actual consultation question from the question knowledge base.

4. The NLP-based customer service dialogue method as described in claim 1, characterized in that, The step of sending the question text to a human terminal to obtain the target answer text includes: The question text is sent to a human terminal to receive a customer service response. The customer service response text is then subjected to a second modification process to obtain the target response text.

5. The NLP-based customer service dialogue method as described in any one of claims 1-4, characterized in that, The step of sending the question text to a human terminal to obtain the target answer text includes: Based on the question text, the user's preferred consultation method can be determined. Based on the consultation method preference, a corresponding human terminal is matched to the user to obtain the target answer text; the human terminal includes at least an online customer service terminal, a voice customer service terminal, and a video customer service terminal.

6. A customer service dialogue device based on NLP, characterized in that, include: The question text acquisition unit is used to acquire the user's question text; The question type classification unit is used to classify the question text by type using a question classification model, and obtain the question type classification result of the question text; the question type classification result includes the question type of the question text and the difficulty level of the question type. The difficulty level determination unit is used to determine whether the difficulty level of the problem type exceeds a preset difficulty level threshold. The human terminal transfer unit is used to send the question text to a human terminal to obtain the target answer text when the difficulty level of the question type exceeds a preset difficulty level threshold. The answer text matching unit is used to match a question pair corresponding to the question text from the question knowledge base based on the question type when the difficulty level does not exceed a preset difficulty level threshold, and to use the question answer text in the question pair as the target answer text. After classifying the question text using a question classification model to obtain the question type classification result, the method further includes: The question text is subjected to sentiment recognition based on the corresponding sentiment classifier matched according to the question type, and the sentiment recognition result of the question text is obtained; the sentiment recognition result is used to perform a first modification process on the question answer text in the question pair to obtain the target answer text; After performing sentiment recognition on the question text according to the corresponding sentiment classifier matched to the question type, and obtaining the sentiment recognition result of the question text, the method further includes: Based on the emotion recognition results, it is determined whether the emotional complexity of the question text exceeds a preset complexity threshold; wherein, the emotion recognition results include the confidence scores of multiple emotion categories, and when the confidence scores of multiple emotion categories exceed the preset confidence threshold, the more emotion categories whose confidence scores exceed the confidence threshold, the higher the emotional complexity of the question text. If the emotional complexity of the question text exceeds a preset complexity threshold, the second objective step is executed: the question text is sent to a human terminal to obtain the target response text; If the emotional complexity of the question text does not exceed a preset complexity threshold, the question answer text in the question pair is subjected to a first modification process based on the emotional recognition result to obtain the target answer text.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the NLP-based customer service dialogue method as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the NLP-based customer service dialogue method as described in any one of claims 1 to 5.

Images