Customer service dialogue question extraction method and device, electronic equipment and storage medium

Abstract

The invention provides a customer service dialogue question extraction method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a customer service dialogue text, the customer service dialogue text comprises a customer identifier and a statement corresponding to the customer identifier; semantic recognition is conducted on statements corresponding to the customer identifiers to obtain statement topics, statement division is conducted on the statements corresponding to the customer identifiers based on the statement topics to obtain customer statements and statement types corresponding to the customer statements, and if the statement types of the customer statements are professional inquiry types, the customer statements serve as candidate statements, and a customer question list is obtained; and comparing the statement theme of the candidate statement with the statement theme of the historical customer question in the customer question list, if the statement theme of the candidate statement is inconsistent with the statement theme of the historical customer question in the customer question list, performing question extraction on the candidate statement to obtain a newly added customer question, and adding the newly added customer question into the customer question list, through the method, the technical problems of resource waste and low answer accuracy caused by too frequent customer service dialogue question extraction are solved.

Description

Technical Field

[0001] The present application relates to the technical field of customer service dialogue processing, and specifically to a method, device, electronic device, and storage medium for extracting questions from customer service dialogues. Background Art

[0002] In hotline customer service systems, automatic speech recognition (ASR) technology is widely used to convert customer voice messages into text for subsequent semantic analysis and processing. However, due to limitations in ASR technology when processing speech signals, factors such as pauses, changes in speech rate, and unclear pronunciation can lead to illogical segmentation in the parsed text. Parsing a sentence into multiple segments can lead to numerous problems when using the original customer utterance directly for question processing and answer generation. Due to the large number of text fragments and their incomplete semantics, large models may frequently parse and respond to these incomplete texts, resulting in an excessively high frequency of answer generation. This not only wastes significant computing resources and increases the system's processing burden, but also makes it difficult for customer service personnel to quickly filter out valuable information when faced with a large number of incoherent and inaccurate responses, hindering timely and effective responses to customers, severely impacting customer service efficiency and quality.

[0003] Currently, most related technologies focus on direct recognition of customer voice and simple semantic analysis. When processing customers' cumbersome voice descriptions and incomplete text parsed by ASR, it is difficult to ensure the accuracy and efficiency of generated answers, and cannot meet the actual needs of hotline customer service scenarios for fast and accurate responses to customer questions. Summary of the Invention

[0004] In view of the shortcomings of the above-mentioned related technologies, the present application provides a method, device, electronic device and storage medium for extracting questions from customer service conversations to solve the technical problems in the related technologies of too frequent extraction of customer service conversation questions resulting in waste of resources and low answer accuracy.

[0005] The present application provides a method for extracting questions from a customer service conversation, the method comprising: obtaining a customer service conversation text, and performing sentence segmentation on the customer service conversation text to obtain customer sentences; performing semantic recognition on the customer sentences to obtain a sentence type and a sentence subject of the customer sentences; if the sentence type of the customer sentence is a professional inquiry type, using the customer sentence as a candidate sentence; obtaining a customer question list, comparing the sentence subject of the candidate sentence with the sentence subject of historical customer questions in the customer question list, and if the comparison is inconsistent, performing question extraction on the candidate sentence to obtain a new customer question, and adding the new customer question to the customer question list.

[0006] In one embodiment of the present application, a historical conversation list is obtained, and the customer statement is added to the historical conversation list as a new customer statement, and the historical conversation list includes multiple historical customer statements; if the statement type of the new customer statement is a professional inquiry type, the statement subject of the new customer statement is compared with the statement subject of the historical customer statements in the historical conversation list. If the comparison is inconsistent, the new customer statement is subjected to question extraction to obtain a new customer question.

[0007] In one embodiment of the present application, the customer service conversation text is preliminarily divided according to punctuation marks in the customer service conversation text to obtain basic division sentences; semantic recognition is performed on the basic division sentences to obtain the sentence subjects of the basic division sentences; if the basic division sentence has the same sentence subject as another basic division sentence, the basic division sentence and the other basic division sentence are sentence merged to obtain the customer sentence; if there are multiple sentence subjects in the basic division sentence, the basic division sentence is split based on the multiple sentence subjects to obtain the customer sentence.

[0008] In one embodiment of the present application, if the customer statement includes preset question words, the sentence type of the customer statement is determined to be an inquiry type; keyword extraction is performed on the customer statement of the inquiry type to obtain inquiry keywords, and the inquiry keywords are compared with preset professional keywords. If the comparison is consistent, the sentence type of the customer statement is determined to be a professional inquiry type.

[0009] In one embodiment of the present application, a training data set is obtained, wherein the training data set includes training sentences and sentence topics corresponding to the training sentences; the training data set is input into a preset language model for training to obtain a sentence topic detection model, wherein the input of the sentence topic detection model is the customer sentence, and the output of the sentence topic detection model is the sentence topic.

[0010] In one embodiment of the present application, the newly added customer question is input into a retrieval enhancement generation model, which includes a preset professional inquiry knowledge base; the retrieval enhancement generation model obtains the most relevant document corresponding to the customer question through semantic vector retrieval, and obtains the supplementary document corresponding to the customer question through keyword retrieval; the most relevant document and the supplementary document are deduplicated to obtain the retained document after deduplication, and the answer to the question corresponding to the customer question is generated based on the retained document.

[0011] In one embodiment of the present application, a voice conversation between the customer and the customer service representative is obtained, and the voice conversation is converted into a customer service conversation text based on speech recognition.

[0012] An embodiment of the present application also provides a device for extracting questions from customer service conversations, the device comprising: a text input module for acquiring customer service conversation text and performing sentence division on the customer service conversation text to obtain customer sentences; a semantic recognition module for performing semantic recognition on the customer sentences to obtain the sentence type and sentence subject of the customer sentences; a sentence classification module for using the customer sentence as a candidate sentence if the sentence type of the customer sentence is a professional inquiry type; a question extraction module for acquiring a customer question list, comparing the sentence subject of the candidate sentence with the sentence subject of historical customer questions in the customer question list; if the comparison is inconsistent, performing question extraction on the candidate sentence to obtain a new customer question, and adding the new customer question to the customer question list.

[0013] An embodiment of the present application also provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the electronic device implements the question extraction method for customer service conversation as described in any of the above embodiments.

[0014] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a computer processor, the computer executes the method for extracting questions from a customer service conversation as described in any of the above embodiments.

[0015] The beneficial effects of the present application include: a method, device, electronic device and storage medium for extracting questions from a customer service conversation, wherein the method obtains a customer service conversation text, the customer service conversation text includes a customer identifier and a sentence corresponding to the customer identifier; semantically identifies the sentence corresponding to the customer identifier to obtain a sentence topic, divides the sentence corresponding to the customer identifier based on the sentence topic, obtains the customer sentence and the sentence type corresponding to the customer sentence, if the sentence type of the customer sentence is a professional inquiry type, uses the customer sentence as a candidate sentence, obtains a customer question list, compares the sentence topic of the candidate sentence with the sentence topic of historical customer questions in the customer question list, if the comparison is inconsistent, extracts questions from the candidate sentence to obtain a new customer question, and adds the new customer question to the customer question list, controls the frequency of question extraction by comparing the sentence topic of the customer sentence with the sentence topic of historical customer questions, thereby achieving answers to fewer questions, reducing the frequency of calling the knowledge base, reducing the answers to repeated questions and resource consumption, and improving the accuracy of customer service responses to customer questions.

[0016] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a schematic diagram of an implementation environment of a method for extracting questions from a customer service conversation, as shown in an exemplary embodiment of the present application; Figure 2 This is a flowchart of a method for extracting questions from customer service conversations, as shown in an exemplary embodiment of the present application; Figure 3 This is a schematic diagram of a question extraction process shown in an exemplary embodiment of the present application; Figure 4 This is a block diagram of a device for extracting questions from customer service conversations, as shown in an exemplary embodiment of the present application; Figure 5 It is a structural diagram of an electronic device shown in an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0018] The following describes the embodiments of the present application through specific examples. Those skilled in the art can easily understand the other advantages and effects of the present application from the content disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other unless they conflict.

[0019] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present application. Therefore, the illustrations only show components related to the present application and are not drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component can be changed at will, and the component layout type may also be more complicated.

[0020] It should be noted that in this application, "first," "second," and the like are merely used to distinguish similar objects, and do not limit the order or precedence of similar objects. The variations of "including," "having," and the like indicate that the scope of the subject of the term is not exclusive, in addition to the examples shown in the term.

[0021] It should be understood that the various numbers, step numbers, and other reference numerals in this application are provided for ease of description and are not intended to limit the scope of this application. The order of reference numerals in this application does not necessarily imply a specific order of execution; the order of execution of each process is determined by its function and inherent logic.

[0022] In the following description, a large number of details are discussed to provide a more thorough explanation of the embodiments of the present application. However, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present application difficult to understand.

[0023] The embodiments of the present application respectively propose a method for extracting questions from customer service conversations, a device for extracting questions from customer service conversations, an electronic device, a computer-readable storage medium, and a computer program product. These embodiments will be described in detail below.

[0024] See also Figure 1 , Figure 1 This is a schematic diagram of an implementation environment of a method for extracting questions from a customer service conversation, shown as an exemplary embodiment of the present application.

[0025] like Figure 1 As shown, the implementation environment may include a customer service terminal 101 and a computer device 102, wherein the computer device 102 may be at least one of a microcomputer, an embedded computer, a neural network computer, etc., and the customer service terminal 101 may be a customer service computer, a customer service mobile phone, or a customer service telephone, etc. The computer device 102 obtains the customer service conversation text of the customer service terminal 101, and the customer service conversation text includes a customer identifier and a sentence corresponding to the customer identifier; semantic recognition is performed on the sentence corresponding to the customer identifier to obtain a sentence topic, and sentence division is performed on the sentence corresponding to the customer identifier based on the sentence topic to obtain a customer sentence and a sentence type corresponding to the customer sentence; if the sentence type of the customer sentence is a professional inquiry type, the customer sentence is used as a candidate sentence, and the candidate sentence is compared with the sentence topic of the historical customer question in the customer question list; if the comparison is inconsistent, question extraction is performed based on the candidate sentence to obtain a new customer question, and the new customer question is added to the customer question list, so as to obtain the answer to the question based on the customer question list query, and the answer to the question is returned to the customer service terminal 101 so that the customer service can answer the customer based on the answer to the question.

[0026] See also Figure 2 , Figure 2 This is a flowchart of a method for extracting questions from customer service conversations, as shown in an exemplary embodiment of the present application. This method can be applied to Figure 1 The implementation environment shown is as follows. The method may also be applicable to other exemplary implementation environments and be specifically executed by devices in other implementation environments. This embodiment does not limit the implementation environment to which the method is applicable.

[0027] like Figure 2As shown, in an exemplary embodiment, the question extraction method for customer service dialogue includes at least steps S210 to S240, which are described in detail as follows: Step S210: Acquire a customer service conversation text, where the customer service conversation text includes a customer ID and a sentence corresponding to the customer ID.

[0028] In one embodiment of the present application, obtaining the customer service conversation text includes: obtaining a voice conversation between the customer and the customer service, and converting the voice conversation into the customer service conversation text based on voice recognition.

[0029] For example, the voice conversation between the customer and the customer service is obtained through the real-time recording of the customer service call, and the voice conversation is converted into customer service conversation text based on voice recognition. The customer service conversation text includes the customer service identifier and the corresponding sentence and the customer identifier and the corresponding sentence. For example, the customer service conversation text includes "Customer: Where is the seat adjustment button of this car?", where "Customer:" is the customer identifier and "Where is the seat adjustment button of this car?" is the customer sentence.

[0030] Step S220 , performing semantic recognition on the sentence corresponding to the customer identifier to obtain a sentence subject, and performing sentence division on the sentence corresponding to the customer identifier based on the sentence subject to obtain a customer sentence and a sentence type corresponding to the customer sentence.

[0031] In one embodiment of the present application, sentence division of sentences corresponding to customer identifiers based on sentence topics includes: performing preliminary division of the customer service conversation text according to punctuation marks in the customer service conversation text to obtain basic division sentences; performing semantic recognition on the basic division sentences to obtain sentence topics of the basic division sentences, and if the basic division sentence has the same sentence topic as another basic division sentence, merging the basic division sentence with the other basic division sentence to obtain a customer sentence; if there are multiple sentence topics in the basic division sentence, splitting the basic division sentence based on the multiple sentence topics to obtain a customer sentence.

[0032] In one embodiment of the present application, the process of summarizing the sentences corresponding to the customer identifier in combination with the context and extracting the complete customer question can be: for example, a customer question may require multiple sentences to fully describe. In this case, the sentences corresponding to the customer identifier can be merged or divided based on the sentence subject, so as to obtain a customer sentence with complete meaning to extract the customer question. For another example, after obtaining a customer sentence with complete meaning, it is also possible to determine whether the customer sentence is a professional inquiry type based on preset question words and preset professional keywords. If the customer sentence is a professional inquiry type, the customer sentence is used as a candidate sentence, and the sentence subject of the candidate sentence is compared with the sentence subject of the historical customer question in the customer question list. If the comparison is different, a new customer question is extracted for the candidate sentence and added to the customer question list. If the comparison is the same, no question is extracted for the candidate sentence.

[0033] For example, based on the customer identifier in the customer service conversation text, corresponding sentences are identified and preliminarily segmented according to punctuation to obtain basic segmented sentences. For example, based on the customer identifier in the customer service conversation text "Customer: I can't find the seat adjustment button. Where is the seat adjustment button for this car?", the corresponding sentence is obtained: "I can't find the seat adjustment button. Where is the seat adjustment button for this car?". First, the sentence is preliminarily segmented according to punctuation to obtain the basic segmented sentences "I can't find the seat adjustment button." and "Where is the seat adjustment button for this car?". Then, semantic recognition is performed on these two basic segmented sentences to determine that the topic of both sentences is about seat adjustment. The two basic segmented sentences are then merged to obtain the customer sentence "I can't find the seat adjustment button. Where is the seat adjustment button for this car?"

[0034] For example, based on the customer identifier in the customer service conversation text "Customer: I would like to ask where the seat adjustment button is, and I also don't know how to connect the music player to Bluetooth.", the corresponding sentence is obtained: "I would like to ask where the seat adjustment button is, and I also don't know how to connect the music player to Bluetooth." First, the sentence is preliminarily divided by punctuation marks to obtain the basic division sentence "I would like to ask where the seat adjustment button is, and I also want to know how to connect the music player to Bluetooth." Then, semantic recognition is performed on this basic division sentence to obtain that the sentence topic of one sentence is about seat adjustment, and the sentence topic of the other sentence is about Bluetooth connection. Then, this basic division sentence is split into two customer sentences: "I would like to ask where the seat adjustment button is?" and "I also want to know how to connect the music player to Bluetooth?".

[0035] In one embodiment of the present application, obtaining the sentence type corresponding to the customer sentence includes: if the customer sentence includes preset question words, determining the sentence type of the customer sentence as an inquiry type; performing keyword extraction on the customer sentence of the inquiry type to obtain inquiry keywords, comparing the inquiry keywords with preset professional keywords, and if the comparison is consistent, determining the sentence type of the customer sentence as a professional inquiry type.

[0036] Exemplarily, preset question words may include: "how," "where," "how," "please ask," and "what." If a customer statement includes preset question words, the statement type of the customer statement is determined to be an inquiry type. For example, "Where are the seat adjustment buttons for this car?" and "What is the address of the latest store?" If both customer statements include preset question words, the statement types of both sentences are determined to be an inquiry type. Preset professional keywords can be set to vehicle nouns such as "seat," "tire," "air conditioning," and "window," or other professional keywords. From "Where are the seat adjustment buttons for this car?", the query keyword extracted is "seat," so the statement type of "Where are the seat adjustment buttons for this car?" can be determined to be a professional inquiry type. From "Where is the address of the latest store?", the query keyword extracted is "store address," which is inconsistent with the preset professional keywords. Therefore, the statement type of "Where is the address of the latest store?" is a non-professional inquiry type.

[0037] In one embodiment of the present application, before performing semantic recognition on the sentence corresponding to the customer identifier to obtain the sentence topic, it includes: obtaining a training data set, the training data set includes training sentences and sentence topics corresponding to the training sentences; inputting the training data set into a preset language model for training to obtain a sentence topic detection model, the input of the sentence topic detection model is the customer sentence, and the output of the sentence topic detection model is the sentence topic.

[0038] In step S230 , if the statement type of the customer statement is a professional inquiry type, the customer statement is used as a candidate statement.

[0039] For example, if the statement type of the customer's statement is a professional inquiry type, it is also necessary to determine whether the statement subject of the customer's statement is consistent with the statement subject of the previously extracted question. If they are consistent, this question will no longer be extracted. If they are inconsistent, the question will be extracted based on the customer's statement. Therefore, the customer's statement is determined as a candidate statement, and the candidate statement needs to be further judged whether a question needs to be extracted.

[0040] Step S240, obtain a customer question list, compare the sentence subject of the candidate sentence with the sentence subject of the historical customer question in the customer question list, if the comparison is inconsistent, extract the question from the candidate sentence to obtain a new customer question, and add the new customer question to the customer question list.

[0041] Exemplarily, question extraction is performed on candidate sentences that are inconsistent with the sentence topics of historical customer questions to obtain new customer questions, and the customer questions are added to the customer question list, wherein question extraction on the candidate sentences includes: performing word segmentation and part-of-speech recognition on the customer sentence to obtain multiple sentence participles and parts of speech corresponding to the sentence participles, extracting the question words and the verbs and nouns near the question words to obtain new customer questions, thereby realizing question extraction.

[0042] In one embodiment of the present application, after obtaining the customer statement and the statement type corresponding to the customer statement, it also includes: obtaining a historical conversation list, adding the customer statement as a new customer statement to the historical conversation list, the historical conversation list includes multiple historical customer statements; if the statement type of the new customer statement is a professional inquiry type, then the statement subject of the new customer statement is compared with the statement subject of the historical customer statements in the historical conversation list. If the comparison is inconsistent, the new customer statement is subjected to question extraction to obtain a new customer question.

[0043] In one embodiment of the present application, after question extraction is performed on candidate sentences to obtain new customer questions, it also includes: inputting the new customer questions into a retrieval enhancement generation model, which includes a preset professional inquiry knowledge base; the retrieval enhancement generation model obtains the most relevant documents corresponding to the customer questions through semantic vector retrieval, and obtains supplementary documents corresponding to the customer questions through keyword retrieval; deduplicating the most relevant documents and supplementary documents to obtain the retained documents after deduplication, and generating answers to the customer questions based on the retained documents.

[0044] Exemplarily, after the user question is extracted, the question will be sent to the RAG (retrieval-augmented generation model) system for answering using a large model. This RAG system contains a knowledge base of customer service car questions (pre-set professional inquiry knowledge base). In the system, the most relevant documents will first be retrieved based on the semantic vector of the customer question, and keyword retrieval will be used as a supplement to obtain supplementary documents. The relevant documents and supplementary documents obtained by the two retrieval methods are deduplicated to obtain retained documents, and the retained documents and user questions are input into a pre-trained re-ranking model to obtain a similarity score for each retained document. The retained documents with a similarity score lower than the preset score threshold are filtered out, and then the customer question and the filtered retained documents are input into the large language model together. The large language model generates the answer to the customer question with reference to the filtered retained documents. Among them, the input of the re-ranking model can be user questions and retained documents, and the output can be the similarity score between the retained documents and the user questions. The similarity score represents the relevance of the retained documents to the user questions.

[0045] See also Figure 3 , Figure 3This is a schematic diagram of a question extraction process shown in an exemplary embodiment of the present application. Figure 3 As shown in the figure, ASR (Automatic Speech Recognition) speech recognition technology aims to convert the lexical content of human speech into computer-readable input, such as keystrokes, binary codes, or character sequences. First, ASR parses the conversation between the customer service representative and the customer to extract a complete sentence. It then determines whether the sentence is a customer statement. The sentence is added to the historical conversation list to determine whether it raises a complete question. If so, it determines whether the sentence is a professional inquiry. If so, a list of customer questions is obtained, and the subject of the sentence is compared with the subject of historical customer questions in the customer question list. If the comparison is inconsistent, the sentence is different from the historical customer questions. Question extraction is performed on the sentence to obtain a new customer question, which is then added to the customer question list.

[0046] See also Figure 4 , Figure 4 This is a block diagram of a customer service conversation question extraction device shown in an exemplary embodiment of the present application. The device can be applied to Figure 1 The implementation environment shown is as follows. The apparatus may also be applicable to other exemplary implementation environments and specifically configured in other devices. This embodiment does not limit the implementation environment to which the apparatus is applicable.

[0047] like Figure 4 As shown, the exemplary customer service dialogue question extraction device includes: The text input module 401 is used to obtain the customer service conversation text, which includes the customer ID and the sentence corresponding to the customer ID; Semantic recognition module 402, configured to perform semantic recognition on the sentence corresponding to the customer identifier to obtain a sentence subject, and to classify the sentence corresponding to the customer identifier based on the sentence subject to obtain a customer sentence and a sentence type corresponding to the customer sentence; The statement classification module 403 is configured to use the customer's statement as a candidate statement if the statement type of the customer's statement is a professional inquiry type; The question extraction module 404 is used to obtain a customer question list, compare the sentence subject of the candidate sentence with the sentence subject of the historical customer questions in the customer question list, and if the comparison is inconsistent, extract the question from the candidate sentence to obtain a new customer question, and add the new customer question to the customer question list.

[0048] The text input module 401 obtains the customer service conversation text, including: obtaining the voice conversation between the customer and the customer service, and converting the voice conversation into the customer service conversation text based on voice recognition.

[0049] The text input module 401 divides the sentences corresponding to the customer identifier based on the sentence topics, including: performing preliminary division of the customer service conversation text according to punctuation marks in the customer service conversation text to obtain basic division sentences; performing semantic recognition on the basic division sentences to obtain the sentence topics of the basic division sentences, if the basic division sentence has the same sentence topic as another basic division sentence, then merging the basic division sentence with the other basic division sentence to obtain the customer sentence; if there are multiple sentence topics in the basic division sentence, then splitting the basic division sentence based on the multiple sentence topics to obtain the customer sentence.

[0050] The semantic recognition module 402 obtains the sentence type corresponding to the customer sentence, including: if the customer sentence includes preset question words, the sentence type of the customer sentence is determined to be an inquiry type; keyword extraction is performed on the customer sentence of inquiry type to obtain inquiry keywords, and the inquiry keywords are compared with preset professional keywords. If the comparison is consistent, the sentence type of the customer sentence is determined to be a professional inquiry type.

[0051] Before the semantic recognition module 402 performs semantic recognition on the sentence corresponding to the customer identifier to obtain the sentence topic, it includes: obtaining a training data set, the training data set includes training sentences and sentence topics corresponding to the training sentences; inputting the training data set into a preset language model for training to obtain a sentence topic detection model, the input of the sentence topic detection model is the customer sentence, and the output of the sentence topic detection model is the sentence topic.

[0052] After the question extraction module 404 obtains the customer statement and the statement type corresponding to the customer statement, it also includes: obtaining a historical conversation list, adding the customer statement as a new customer statement to the historical conversation list, and the historical conversation list includes multiple historical customer statements; if the statement type of the new customer statement is a professional inquiry type, then the statement subject of the new customer statement is compared with the statement subject of the historical customer statements in the historical conversation list. If the comparison is inconsistent, the new customer statement is extracted to obtain a new customer question.

[0053] After the question extraction module 404 extracts questions from the candidate sentences to obtain new customer questions, it also includes: inputting the new customer questions into the retrieval enhancement generation model, which includes a preset professional inquiry knowledge base; the retrieval enhancement generation model obtains the most relevant documents corresponding to the customer questions through semantic vector retrieval, and obtains supplementary documents corresponding to the customer questions through keyword retrieval; deduplicating the most relevant documents and supplementary documents to obtain the retained documents after deduplication, and generating the answers to the customer questions based on the retained documents.

[0054] Through the above device, the frequency of extracting questions is controlled by comparing customer statements with the sentence topics of historical customer questions, thereby answering fewer questions, reducing the frequency of calling the knowledge base, reducing the answers to repeated questions and resource consumption, and improving the accuracy of customer service's responses to customer questions.

[0055] It can be understood that the problem extraction device for customer service conversations provided in the above embodiment and the problem extraction method for customer service conversations provided in the above embodiment belong to the same concept, wherein the specific manner in which the problem extraction method for customer service conversations performs operations has been described in detail in the above embodiment and will not be repeated here. In actual applications, the problem extraction device for customer service conversations provided in the above embodiment can allocate the above functions to different functional modules as needed, that is, divide the internal structure of the problem extraction device for customer service conversations into different functional modules, and then implement all or part of the functions of the corresponding functional modules through the problem extraction method for customer service conversations described in the above embodiment, and no specific restrictions are imposed here. For example, the text input module 401 includes a module for executing step S210 and related steps, the semantic recognition module 402 includes a module for executing step S220 and related steps, the sentence classification module 403 includes a module for executing step S230 and related steps, and the problem extraction module 404 includes a module for executing step S240 and related steps.

[0056] Figure 5 The following is a schematic diagram showing the structure of a computer system suitable for implementing an electronic device according to an embodiment of the present application. Figure 5 The computer system 500 of the electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0057] like Figure 5 As shown, computer system 500 includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes, such as the methods described in the above embodiments, based on programs stored in read-only memory (ROM) 502 or programs loaded from storage 508 into random access memory (RAM) 503. RAM 503 also stores various programs and data required for system operation. CPU 501, ROM 502, and RAM 503 are interconnected via bus 504. An input / output (I / O) interface 505 is also connected to bus 504.

[0058] The following components are connected to the I / O interface 505: an input section 506 including a keyboard, mouse, and the like; an output section 507 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and speakers; a storage section 508 including devices such as a hard disk; and a communication section 509 including a network interface card such as a LAN (Local Area Network) card or a modem. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the I / O interface 505 as needed. Removable media 511, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 510 as needed, so that computer programs read from the media can be installed in the storage section 508 as needed.

[0059] In particular, according to embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for executing the methods illustrated in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via the communication section 509 and / or installed from removable media 511. When executed by the central processing unit (CPU) 501, the computer program performs the various functions defined in the system of the present application.

[0060] It should be noted that the computer-readable medium described in the embodiments of this application may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may, for example, be an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or component, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. This propagated data signal may take a variety of forms, including, but not limited to, an electromagnetic signal, an optical signal, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. A computer program embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, or any suitable combination thereof.

[0061] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. Among them, each box in the flowchart or block diagram can represent a module, program segment, or part of the code, and the above-mentioned module, program segment, or part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0062] The units involved in the embodiments described in this application may be implemented by software or hardware, and the units described may also be set in a processor. In some cases, the names of these units do not constitute limitations on the units themselves.

[0063] Another aspect of the present application provides a computer-readable storage medium storing a computer program that, when executed by a computer processor, causes the computer to perform the aforementioned method for extracting customer service conversation questions. The computer-readable storage medium may be included in the electronic device described in the above embodiments, or may exist independently and not be incorporated into the electronic device.

[0064] Another aspect of the present application provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method for extracting questions from customer service conversations provided in each of the above embodiments.

[0065] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Anyone skilled in the art may modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, any equivalent modifications or alterations accomplished by a person of ordinary skill in the art without departing from the spirit and technical concepts disclosed in this application shall be covered by the claims of this application.

Claims

1. A method for extracting questions from customer service conversations, characterized in that: The method comprises: Obtaining a customer service conversation text, the customer service conversation text including a customer identifier and a sentence corresponding to the customer identifier; Performing semantic recognition on the sentence corresponding to the customer identifier to obtain a sentence subject, and performing sentence segmentation on the sentence corresponding to the customer identifier based on the sentence subject to obtain a customer sentence and a sentence type corresponding to the customer sentence; If the statement type of the customer statement is a professional inquiry type, the customer statement is used as a candidate statement; Obtain a customer question list, compare the sentence subject of the candidate sentence with the sentence subject of the historical customer questions in the customer question list, and if the comparison is inconsistent, extract the question from the candidate sentence to obtain a new customer question, and add the new customer question to the customer question list.

2. The method for extracting questions from customer service dialogues according to claim 1, characterized in that: After obtaining the customer statement and the statement type corresponding to the customer statement, the method further includes: Obtain a historical conversation list, and add the customer statement as a new customer statement to the historical conversation list, wherein the historical conversation list includes multiple historical customer statements; If the statement type of the newly added customer statement is a professional inquiry type, the statement subject of the newly added customer statement is compared with the statement subject of the historical customer statements in the historical conversation list. If the comparison is inconsistent, the question of the newly added customer statement is extracted to obtain the new customer question.

3. The method for extracting questions from customer service dialogues according to claim 1, characterized in that: Sentence division of the sentences corresponding to the customer identifier based on the sentence subject includes: Preliminarily dividing the customer service conversation text according to punctuation marks in the customer service conversation text to obtain basic division sentences; Performing semantic recognition on the basic partitioning statement to obtain a statement subject of the basic partitioning statement, and if the statement subject of the basic partitioning statement is the same as that of another basic partitioning statement, performing statement merging on the basic partitioning statement and the other basic partitioning statement to obtain the customer statement; If there are multiple statement subjects in the basic partition statement, the basic partition statement is split based on the multiple statement subjects to obtain the customer statement.

4. The method for extracting questions from customer service dialogues according to any one of claims 1 to 3, characterized in that: The statement types corresponding to the customer statement include: If the customer's statement includes a preset question word, determining the statement type of the customer's statement as an inquiry type; Keywords are extracted from the customer's statement of the inquiry type to obtain inquiry keywords, and the inquiry keywords are compared with preset professional keywords. If the comparison is consistent, the statement type of the customer's statement is determined to be a professional inquiry type.

5. The method for extracting questions from customer service dialogues according to any one of claims 1 to 3, characterized in that: Before performing semantic recognition on the sentence corresponding to the customer identifier to obtain the sentence subject, the method further includes: Obtaining a training data set, the training data set including training sentences and sentence topics corresponding to the training sentences; The training data set is input into a preset language model for training to obtain a sentence topic detection model, wherein the input of the sentence topic detection model is the customer sentence, and the output of the sentence topic detection model is the sentence topic.

6. The method for extracting questions from customer service dialogues according to any one of claims 1 to 3, characterized in that: After extracting questions from the candidate sentences to obtain new customer questions, the following steps are also included: Inputting the newly added customer question into a search enhancement generation model, wherein the search enhancement generation model includes a preset professional inquiry knowledge base; The retrieval enhancement generation model obtains the most relevant document corresponding to the customer question through semantic vector retrieval, and obtains supplementary documents corresponding to the customer question through keyword retrieval; Deduplication is performed on the most relevant document and the supplementary document to obtain a retained document after deduplication, and an answer corresponding to the customer question is generated based on the retained document.

7. The method for extracting questions from customer service dialogues according to any one of claims 1 to 3, characterized in that: Obtaining customer service conversation text includes: The voice conversation between the customer and the customer service representative is obtained, and the voice conversation is converted into customer service conversation text based on speech recognition.

8. A device for extracting questions from customer service conversations, characterized in that: The device comprises: A text input module, configured to obtain customer service conversation text, wherein the customer service conversation text includes a customer identifier and a sentence corresponding to the customer identifier; a semantic recognition module configured to perform semantic recognition on the sentence corresponding to the customer identifier to obtain a sentence subject, and to perform sentence classification on the sentence corresponding to the customer identifier based on the sentence subject to obtain a customer sentence and a sentence type corresponding to the customer sentence; a statement classification module, configured to use the customer statement as a candidate statement if the statement type of the customer statement is a professional inquiry type; The question extraction module is used to obtain a customer question list, compare the sentence subject of the candidate sentence with the sentence subject of the historical customer questions in the customer question list, and if the comparison is inconsistent, extract the question from the candidate sentence to obtain a new customer question, and add the new customer question to the customer question list.

9. An electronic device, characterized in that: The electronic device comprises: one or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, enables the electronic device to implement the question extraction method for customer service dialogue according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor of a computer, the computer is caused to execute the method for extracting questions from a customer service dialogue according to any one of claims 1 to 7.

Images