Customer service voice intelligent detection system and method based on machine learning
By using a machine learning-based intelligent voice detection system for customer service, voice data from customer inquiries and responses from intelligent customer service robots are extracted and analyzed to determine the validity of responses and transfer customers to human customer service. This solves the problem that intelligent customer service robots cannot accurately understand customer questions, enabling timely resolution of customer issues and improved customer satisfaction.
Patent Information
- Application Number
- CN202411156243.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-22
- Publication Date
- 2024-11-15
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Intelligent customer service robots often fail to accurately understand customer needs when dealing with complex issues, resulting in irrelevant answers or the provision of unrelated information, thus failing to effectively resolve customer problems.
A machine learning-based intelligent voice detection system for customer service is adopted. Through speech recognition and semantic feature encoding, voice data of customer inquiries and responses from intelligent customer service robots are extracted, and semantic feature encoding is performed to determine whether the response effectively solves the customer's problem, and the connection is transferred to human customer service when necessary.
This improved the optimization of customer service resource allocation, enabled timely resolution of customer issues, and enhanced customer satisfaction.
Smart Images

Figure CN118968973A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of intelligent detection, and more specifically, to a customer service voice intelligent detection system and method based on machine learning. Background Art
[0002] For enterprises, compared with traditional manual customer service, intelligent customer service robots can provide 24-hour service, identify customer voice consultations, complaints and other information, and automatically reply to some simple and repetitive questions, which not only improves response speed and service efficiency, but also reduces labor costs. However, irrelevant answers, self-talk, and stereotyped answers are also problems that exist in current customer service robots. In order to solve the problem that customers cannot solve the problem after repeated communication with intelligent customer service, we are looking forward to a customer service voice intelligent detection solution based on machine learning. Summary of the invention
[0003] In response to the shortcomings of the prior art, the present application provides a customer service voice intelligent detection system and method based on machine learning.
[0004] According to one aspect of the present application, a customer service voice intelligent detection system based on machine learning is provided, which includes: A voice call data collection module is used to obtain the voice call data to be analyzed; A voice call data segmentation module is used to extract customer consultation voice data and intelligent customer service robot reply voice data from the voice call data to be analyzed; A semantic association feature encoding module, used to perform association semantic feature encoding on the customer consultation voice data and the intelligent customer service robot reply voice data to obtain a consultation-reply semantic association feature vector; The voice transfer intelligent detection result generation module is used to obtain a detection result of whether the voice call needs to be transferred to manual customer service based on the consultation-reply semantic association feature vector.
[0005] According to another aspect of the present application, a customer service voice intelligent detection method based on machine learning is provided, which includes: Obtain the voice call data to be analyzed; Extracting customer consultation voice data and intelligent customer service robot reply voice data from the voice call data to be analyzed respectively; Performing correlation semantic feature encoding on the customer consultation voice data and the intelligent customer service robot reply voice data to obtain a consultation-reply semantic correlation feature vector; Based on the consultation-reply semantic association feature vector, a detection result of whether the voice call needs to be transferred to manual customer service is obtained.
[0006] This application has significant technical effects due to the adoption of the above technical solutions: The customer service voice intelligent detection system and method based on machine learning provided in this application first extracts customer consultation voice data and intelligent customer service robot reply voice data from the voice call data to be detected, and then performs semantic analysis on the customer consultation voice data and the intelligent customer service robot reply voice data to obtain customer consultation semantic features and customer service reply semantic features, and then associates the customer service semantic features with the customer consultation semantic features to determine whether the reply content of the intelligent customer service robot effectively solves the customer problem. When it is detected that the problem has not been solved, the voice call is transferred to the manual customer service in a timely manner. In this way, it can ensure that while optimizing the allocation of customer service resources, customer problems are solved in a timely manner, thereby improving customer satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS
[0007] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.
[0008] Figure 1 The present invention is a flowchart of a method for intelligent customer service voice detection based on machine learning according to an embodiment of the present application.
[0009] Figure 2 The present invention is a flowchart of the customer service voice intelligent detection method based on machine learning according to an embodiment of the present application, in which the customer consultation voice data and the intelligent customer service robot reply voice data are encoded with associated semantic features to obtain a consultation-reply semantic associated feature vector.
[0010] Figure 3 This is a flowchart of a customer service voice intelligent detection method based on machine learning according to an embodiment of the present application, which obtains a detection result of whether a voice call needs to be transferred to a manual customer service based on the consultation-reply semantic association feature vector.
[0011] Figure 4 This is a system block diagram of a customer service voice intelligent detection system based on machine learning according to an embodiment of the present application.
[0012] Figure 5 is a block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION
[0013] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.
[0014] In today's rapidly developing digital age, companies are facing growing customer service needs. As an innovative technology, intelligent customer service robots provide companies with an efficient and economical solution. Compared with traditional manual customer service, intelligent customer service robots can provide 24-hour uninterrupted service. They can identify and process customer inquiries, complaints and other requests through voice, and automatically reply to some common and repetitive questions. This automated service not only significantly improves response speed and service efficiency, but also greatly reduces the company's labor costs, enabling companies to reallocate resources to other key areas. However, there are also some challenges in the actual application of intelligent customer service robots. For example, they may not accurately understand the customer's complex questions or specific contexts, resulting in irrelevant answers or irrelevant information. Therefore, a customer service voice intelligent detection solution based on machine learning is expected. + In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also demonstrated a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. The development of deep learning and neural networks has provided new solutions and solutions for customer service voice intelligent detection.
[0015] Figure 1 Flowchart of the customer service voice intelligent detection method based on machine learning according to an embodiment of the present application. Figure 1 As shown, according to the machine learning-based customer service voice intelligent detection method according to the embodiment of the present application, the method includes: S110, obtaining voice call data to be analyzed; S120, respectively extracting customer consultation voice data and intelligent customer service robot reply voice data from the voice call data to be analyzed; S130, performing associated semantic feature encoding on the customer consultation voice data and the intelligent customer service robot reply voice data to obtain a consultation-reply semantic associated feature vector; S140, based on the consultation-reply semantic associated feature vector, obtaining a detection result of whether the voice call needs to be transferred to manual customer service.
[0016] Specifically, in the technical solution of the present application, it is first necessary to obtain the voice call data to be analyzed. It should be understood that the voice call data to be analyzed includes customer consultation voice data, and the specific content includes specific questions and needs raised by customers when interacting with intelligent robot customer service. These problems may involve various types such as products and services, and will also include background information of the problems mentioned by customers during the consultation process. The background context information helps the intelligent customer service to better understand the problem. The voice call data to be analyzed will also include specific reply information of the intelligent customer service robot to customer questions, and these reply information may contain steps and suggestions for solving customer problems. By analyzing the customer consultation content and the reply content of the intelligent customer service robot, it can be judged whether the answer of the customer service robot is directly related to the customer's problem. If the reply deviates from the customer's needs, it means that the robot has failed to effectively solve the problem. At this time, it is necessary to transfer the call to the manual customer service in time, so as to effectively and timely solve the customer problem and improve customer satisfaction. In particular, in an implementable embodiment of the present application, the customer service system is generally equipped with a recording function, which can be directly obtained through the customer service system. It is worth noting that the recording and the method of obtaining call data here need to be legal and compliant.
[0017] In order to more clearly identify and understand the expressions and intentions of different roles, it is necessary to extract customer consultation voice data and intelligent customer service robot reply voice data from the voice call data to be analyzed. Specifically, customer consultation voice represents the customer's needs and problems and is the starting point of the analysis; while the intelligent customer service reply voice is the customer service's response to the customer's consultation and is the core object of the analysis. By separating the two, they can be processed and analyzed independently, thereby more accurately evaluating the performance of intelligent customer service.
[0018] Figure 2 The present invention is a flowchart of the method for intelligent customer service voice detection based on machine learning according to an embodiment of the present application, in which the customer consultation voice data and the intelligent customer service robot reply voice data are encoded with associated semantic features to obtain a consultation-reply semantic associated feature vector. Figure 2 As shown, the associating semantic feature encoding of the customer consultation voice data and the intelligent customer service robot reply voice data to obtain a consultation-reply semantic association feature vector includes: S131, performing voice recognition on the customer consultation voice data and the intelligent customer service robot reply voice data respectively to obtain a customer consultation text and a customer service robot reply text; S132, performing semantic feature extraction on the customer consultation text to obtain a customer consultation semantic feature vector; S133, performing semantic feature extraction on the customer service robot reply text to obtain a customer service reply robot semantic feature vector; S134, associating the customer consultation semantic feature vector with the customer service reply robot semantic feature vector to obtain the consultation-reply semantic association feature vector.
[0019] Considering that the original voice data is a digital audio stream and cannot be directly used for subsequent semantic analysis, in order to convert the audio data into a method that can be understood and processed by the computer, it is necessary to perform voice recognition on the customer consultation voice data and the intelligent customer service robot reply voice data respectively to obtain the customer consultation text and the customer service robot reply text. Through voice recognition technology, audio data can be converted into text data, and text data can be more conveniently segmented and processed for semantic analysis, so that it can be used to evaluate the relevance and accuracy of the intelligent customer service robot's reply. Specifically, voice recognition technology can accurately convert the information in the voice into text representation, avoiding semantic understanding deviations caused by errors. The accuracy of this information expression is conducive to improving the quality of customer service analysis, because any slight semantic deviation will affect the judgment of the effectiveness of the customer service reply.
[0020] The customer consultation text after speech recognition is a long sequence of text paragraphs. In order for the computer to more accurately understand the customer's intentions and the content of the expression, it is necessary to segment the customer consultation text and pass it through the customer consultation text semantic feature encoder to obtain the customer consultation semantic feature vector. The structure of the customer consultation text semantic feature encoder in this application is a first text semantic encoder based on a converter including an embedding layer. It should be understood that the word segmentation process can divide the long sequence of text into independent vocabulary units. After word segmentation, each word represents a specific semantics, and then the text after word segmentation is processed based on a converter including an embedding layer to perform semantic analysis of the customer consultation text. Those of ordinary skill in the art should know that the embedding layer is a common structure for processing text data in a deep learning model. It maps each word to a vector in a high-dimensional space, which can capture the semantic information of the word. The dimension of word embedding is usually much smaller than the size of the vocabulary, so it can reduce the number of parameters and computational complexity of the model while retaining the semantic relationship between words. In the customer consultation text semantic feature encoder, the embedding layer converts the word segmented words into corresponding word embedding vectors, providing a basis for subsequent processing. The transformer is a deep learning model based on the self-attention mechanism. It is composed of multiple encoders and decoders stacked together, each of which contains multiple self-attention layers and feedforward neural network layers. In the customer consultation text semantic feature encoder, the transformer-based encoder captures the dependencies between words through the self-attention mechanism, no matter how far apart these words are in the text. This mechanism enables the transformer model to better understand the contextual information of the customer consultation text, thereby extracting more accurate semantic features.
[0021] In order to accurately reflect the overall meaning of the intelligent customer service robot's reply text, and thus provide strong support for subsequent quality assessments, it is necessary to segment the customer service robot's reply text and pass it through the customer service robot's reply text semantic feature encoder to obtain a customer service reply robot semantic feature vector. In this application, the customer service robot reply text semantic feature encoder is a converter-based second text semantic encoder that includes an embedding layer. It should be understood that through the processing of the semantic feature encoder, the text content can be converted into a feature vector containing key reply text semantic features. Compared with traditional text processing methods, the converter-based encoder can more accurately extract the semantic features of the reply text. It uses a self-attention mechanism, which can take into account the contextual information of the entire reply text key word sequence when processing each reply text key word, thereby avoiding the information loss or distortion problems that may exist in traditional methods.
[0022] Considering that customer consultation and customer service response are the two core parts of a complete conversation, in order to understand the contextual relationship of the conversation in the voice call data, and thus understand whether the intelligent customer service robot response can meet customer needs, the technical solution of this application needs to associate the customer consultation semantic feature vector and the customer service response robot semantic feature vector to obtain the consultation-response semantic association feature vector. When it is found that the semantic association between the customer service response and the customer consultation is low, it means that the intelligent customer service robot response has deviations, misunderstandings, or irrelevant answers. This association analysis provides a data basis for the automated evaluation of the quality of customer service responses.
[0023] Figure 3 The present invention is a flowchart of a method for intelligently detecting customer service voice based on machine learning according to an embodiment of the present application, which detects whether a voice call needs to be transferred to a human customer service based on the inquiry-reply semantic association feature vector. Figure 3 As shown, the detection result of whether the voice call needs to be transferred to manual customer service is obtained based on the consultation-reply semantic association feature vector, including: S141, performing back-propagation expression compensation on the consultation-reply semantic association feature vector based on relevant information to obtain an optimized consultation-reply semantic association feature vector; S142, passing the optimized consultation-reply semantic association feature vector through a reply quality classifier to obtain a classification result, and the classification result is used to indicate whether the reply of the intelligent customer service robot effectively solves the customer's consultation problem; S143, in response to the reply of the intelligent customer service robot not effectively solving the customer's consultation problem, promptly transferring the voice call to manual customer service.
[0024] In particular, in the technical solution of the present application, considering that the weights of the reply quality classifier are learned during the model training process, these weights are crucial to how the model interprets the input semantic association feature vector. If the weights do not match the input feature vector, the classifier may not be able to accurately judge the quality of the reply. During the classification process, if the weights of the classifier are inconsistent with the distribution of the input feature vector, it may cause the features of certain categories to be incorrectly emphasized or ignored, resulting in category coherence interference. This interference may affect the judgment accuracy of the classifier, resulting in the inability to correctly identify whether the reply of the intelligent customer service robot is valid. In order to avoid category coherence interference, in the technical solution of the present application, the consultation-reply semantic association feature vector is compensated by back-propagation expression based on relevant information to obtain an optimized consultation-reply semantic association feature vector.
[0025] Specifically, in an embodiment of the present application, the consultation-reply semantic association feature vector is subjected to back-propagation expression compensation based on relevant information to obtain an optimized consultation-reply semantic association feature vector, including: calculating the product between the consultation-reply semantic association feature vector and its transposed vector to obtain a consultation-reply semantic association feature autocorrelation expression matrix; calculating the coherent interference phase between the category decoding weight matrix of the reply quality classifier and the consultation-reply semantic association feature autocorrelation expression matrix to obtain a consultation-reply semantic association feature-category weight fine-grained coupling representation matrix; performing probabilistic processing on the consultation-reply semantic association feature-category weight fine-grained coupling representation matrix based on the Sigmoid activation function to obtain a consultation-reply semantic association feature-category weight fine-grained coupling probabilistic representation matrix; performing back-propagation expression compensation on the consultation-reply semantic association feature vector based on the consultation-reply semantic association feature-category weight fine-grained coupling probabilistic representation matrix to obtain a category-related compensation representation vector; calculating the position-weighted sum between the category-related compensation representation vector and the consultation-reply semantic association feature vector to obtain the optimized consultation-reply semantic association feature vector.
[0026] More specifically, in an embodiment of the present application, a possible implementation method for calculating the coherent interference phase between the category decoding weight matrix of the reply quality classifier and the consultation-reply semantic association feature autocorrelation expression matrix to obtain the consultation-reply semantic association feature-category weight fine-grained coupling representation matrix may be: the coherent interference phase between the category decoding weight matrix of the reply quality classifier and the consultation-reply semantic association feature autocorrelation expression matrix is processed by the following formula to obtain the consultation-reply semantic association feature-category weight fine-grained coupling representation matrix; wherein the formula is:
[0027] in, represents the class decoding weight matrix of the reply quality classifier row vector, The first row vector, represents the class decoding weight matrix of the reply quality classifier The row vector and the autocorrelation expression matrix of the classification feature The covariance matrix of the data set to which the row vector belongs, represents the inverse of the covariance matrix, The first part of the probabilistic representation matrix of the fine-grained coupling of classification features and class weights The eigenvalue of the location.
[0028] More specifically, in an embodiment of the present application, a possible implementation method of performing back-propagation expression compensation on the consultation-reply semantic association feature vector based on the consultation-reply semantic association feature-category weight fine-grained coupling probabilistic representation matrix to obtain a category-related compensation representation vector may be: performing back-propagation expression compensation processing on the consultation-reply semantic association feature vector according to the following formula to obtain a category-related compensation representation vector; wherein the formula is:
[0029] in, represents the category-dependent compensation representation vector, represents the semantic association feature-category weight fine-grained coupling probabilistic representation matrix of the consultation reply, represents the inquiry-response semantic association feature vector, Represents matrix multiplication.
[0030] In the technical solution of the present application, when the consultation-reply semantic association feature vector is classified by the reply quality classifier, since the weight of the reply quality classifier also needs to adapt to the consultation-reply semantic association feature vector, category coherent interference with the consultation-reply semantic association feature vector may occur. Based on this, in the technical solution of the present application, the consultation-reply semantic association feature vector is compensated by back-propagation expression based on relevant information, which first represents the autocorrelation interference feature spectrum with the autocorrelation matrix of the consultation-reply semantic association feature vector, and then simulates the fine-grained feature-category label coherent interference phase between the consultation-reply semantic association feature and the weight matrix of the reply quality classifier by calculating the correlation measure between the autocorrelation matrix of the consultation-reply semantic association feature vector and the category decoding weight matrix of the reply quality classifier, and then performs back-propagation expression compensation on the consultation-reply semantic association feature vector with the consultation-reply semantic association feature-category weight fine-grained coupling representation matrix, so as to restore the equivalent probability intensity representation of the consultation-reply semantic association feature vector in the absence of interference by reverse compensation, and obtain the optimized consultation-reply semantic association feature vector to improve the accuracy of the classification result.
[0031] Next, the optimized consultation-reply semantic association feature vector is passed through the reply quality classifier to obtain a classification result, and the classification result is used to indicate whether the reply of the intelligent customer service robot effectively solves the problem consulted by the customer. It should be understood that the optimized consultation-reply semantic association feature vector contains whether the reply of the intelligent customer service robot can effectively match the customer's inquiry information, that is, whether the intelligent customer service robot effectively solves the customer's problem. By inputting the semantic association features of the consultation-reply into the classifier, the quality of the reply can be automatically and objectively quantified and evaluated, reducing human errors. The reply quality classifier is a model based on machine learning, which receives the optimized consultation-reply semantic association feature vector as input, and divides the reply into different quality categories (i.e., "solved" and "unsolved") by learning and analyzing these features. When the classification result indicates that the reply of the intelligent customer service robot does not effectively solve the problem consulted by the customer, the voice call is promptly transferred to the manual customer service, and the manual customer service provides more accurate and professional solutions, thereby ensuring that the customer's problem is properly solved.
[0032] In summary, the customer service voice intelligent detection method based on machine learning based on the embodiment of the present application is explained, which first extracts customer consultation voice data and intelligent customer service robot reply voice data from the voice call data to be detected, and then performs semantic analysis on the customer consultation voice data and the intelligent customer service robot reply voice data to obtain customer consultation semantic features and customer service reply semantic features, and then associates the customer service semantic features with the customer consultation semantic features to determine whether the reply content of the intelligent customer service robot effectively solves the customer problem. When it is detected that the problem has not been solved, the voice call is transferred to the manual customer service in a timely manner. In this way, it can ensure that while optimizing the allocation of customer service resources, customer problems are solved in a timely manner, thereby improving customer satisfaction.
[0033] Figure 4 : is a system block diagram of a customer service voice intelligent detection system based on machine learning according to an embodiment of the present application. Figure 4 As shown, according to the machine learning-based customer service voice intelligent detection system 100 according to the embodiment of the present application, it includes: a voice call data collection module 110, used to obtain the voice call data to be analyzed; a voice call data segmentation module 120, used to extract the customer consultation voice data and the intelligent customer service robot reply voice data from the voice call data to be analyzed respectively; a semantic association feature encoding module 130, used to perform association semantic feature encoding on the customer consultation voice data and the intelligent customer service robot reply voice data to obtain a consultation-reply semantic association feature vector; a voice transfer intelligent detection result generation module 140, used to obtain a detection result of whether the voice call needs to be transferred to manual customer service based on the consultation-reply semantic association feature vector.
[0034] Here, those skilled in the art can understand that the specific functions and operations of each unit and module in the above-mentioned customer service voice intelligent detection system 100 based on machine learning have been referred to above. Figures 1 to 3 The description of the customer service voice intelligent detection method based on machine learning has been introduced in detail, and therefore, its repeated description will be omitted.
[0035] In summary, the customer service voice intelligent detection system 100 based on machine learning according to the embodiment of the present application is explained, which first extracts customer consultation voice data and intelligent customer service robot reply voice data from the voice call data to be detected, and then performs semantic analysis on the customer consultation voice data and the intelligent customer service robot reply voice data to obtain customer consultation semantic features and customer service reply semantic features, and then associates the customer service semantic features with the customer consultation semantic features to determine whether the reply content of the intelligent customer service robot effectively solves the customer problem. When it is detected that the problem has not been solved, the voice call is transferred to the manual customer service in a timely manner. In this way, it can ensure that while optimizing the allocation of customer service resources, customer problems can be solved in a timely manner, thereby improving customer satisfaction.
[0036] As described above, the customer service voice intelligent detection system 100 based on machine learning according to the embodiment of the present application can be implemented in various wireless terminals, such as a server for customer service voice intelligent detection based on machine learning, etc. In one example, the customer service voice intelligent detection system 100 based on machine learning according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the customer service voice intelligent detection system 100 based on machine learning can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the customer service voice intelligent detection system 100 based on machine learning can also be one of the many hardware modules of the wireless terminal.
[0037] Alternatively, in another example, the customer service voice intelligent detection system 100 based on machine learning and the wireless terminal may also be separate devices, and the customer service voice intelligent detection system 100 based on machine learning may be connected to the wireless terminal via a wired and / or wireless network and transmit interaction information in accordance with an agreed data format.
[0038] Below, reference Figure 5 To describe an electronic device according to an embodiment of the present application. Figure 5 1 is a block diagram of an electronic device according to an embodiment of the present application. For example, the electronic device 10 can be provided as a server or a terminal device. Figure 5 The electronic device 10 includes a processing component 11, which further includes one or more processors, and a memory resource represented by a memory 12 for storing instructions executable by the processing component 11, such as an application. The application stored in the memory 12 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 11 is configured to execute instructions to perform the above method.
[0039] The electronic device 10 may further include a power supply component 13 configured to perform power management of the electronic device 10, a wired or wireless network interface 14 configured to connect the electronic device 10 to a network, and an input / output (I / O) interface 15. The electronic device 10 may operate based on an operating system stored in the memory 12, such as Windows Server™, Mac OSX™, Unix™, Linux™, FreeBSD™ or the like.
[0040] In an embodiment of the present application, a computer-readable storage medium is also provided, such as a memory 12 including computer program instructions. The computer program instructions can be executed by a processing component 11 of an electronic device 10 to complete the above method.
[0041] The present application may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer program instructions for causing a processor to implement various aspects of the present application.
[0042] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.
[0043] The computer program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer program instructions from the network and forwards the computer program instructions for storage in a computer-readable storage medium in each computing / processing device.
[0044] The computer program instructions for performing the operation of the present application can be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The computer program instructions can be executed completely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer, partially on the remote computer, or completely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet). In some embodiments, by using the state information of the computer program instructions to personalize an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA) or a programmable logic array (PLA), the electronic circuit can execute the computer program instructions, thereby realizing various aspects of the present application.
[0045] Various aspects of the present application are described herein with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each box in the flowchart and / or block diagram and the combination of boxes in the flowchart and / or block diagram can be implemented by computer program instructions.
[0046] These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.
[0047] Computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.
Claims
1. A customer service voice intelligent detection method based on machine learning, characterized in that: include: Obtain the voice call data to be analyzed; Extracting customer consultation voice data and intelligent customer service robot reply voice data from the voice call data to be analyzed respectively; Performing correlation semantic feature encoding on the customer consultation voice data and the intelligent customer service robot reply voice data to obtain a consultation-reply semantic correlation feature vector; Based on the consultation-reply semantic association feature vector, a detection result of whether the voice call needs to be transferred to manual customer service is obtained.
2. The customer service voice intelligent detection method based on machine learning according to claim 1 is characterized in that: The customer consultation voice data and the intelligent customer service robot reply voice data are encoded with associated semantic features to obtain a consultation-reply semantic associated feature vector, including: Performing voice recognition on the customer consultation voice data and the intelligent customer service robot reply voice data respectively to obtain a customer consultation text and a customer service robot reply text; Extracting semantic features from the customer consultation text to obtain a customer consultation semantic feature vector; Performing semantic feature extraction on the customer service robot reply text to obtain a customer service reply robot semantic feature vector; The customer consultation semantic feature vector and the customer service reply robot semantic feature vector are associated to obtain the consultation-reply semantic association feature vector.
3. The customer service voice intelligent detection method based on machine learning according to claim 2 is characterized in that: The method further comprises extracting semantic features from the customer consultation text to obtain a customer consultation semantic feature vector, comprising: segmenting the customer consultation text and passing the result through a customer consultation text semantic feature encoder to obtain the customer consultation semantic feature vector.
4. The customer service voice intelligent detection method based on machine learning according to claim 3 is characterized in that: The semantic feature extraction is performed on the customer service robot reply text to obtain a customer service robot reply semantic feature vector, including: after the customer service robot reply text is segmented, a customer service robot reply text semantic feature encoder is used to obtain the customer service robot reply semantic feature vector.
5. The customer service voice intelligent detection method based on machine learning according to claim 4 is characterized in that: The customer consultation text semantic feature encoder is a first text semantic encoder based on a converter and includes an embedding layer, and the customer service robot reply text semantic feature encoder is a second semantic encoder based on a converter and includes an embedding layer.
6. The customer service voice intelligent detection method based on machine learning according to claim 5 is characterized in that: Based on the inquiry-reply semantic association feature vector, a detection result of whether the voice call needs to be transferred to a manual customer service is obtained, including: Performing back-propagation expression compensation on the inquiry-reply semantic association feature vector based on relevant information to obtain an optimized inquiry-reply semantic association feature vector; The optimized inquiry-reply semantic association feature vector is passed through a reply quality classifier to obtain a classification result, wherein the classification result is used to indicate whether the reply of the intelligent customer service robot effectively solves the customer's inquiry problem; In response to the fact that the reply of the intelligent customer service robot does not effectively solve the problem consulted by the customer, the voice call is promptly transferred to the manual customer service.
7. The customer service voice intelligent detection method based on machine learning according to claim 6 is characterized in that: The consultation-reply semantic association feature vector is subjected to back-propagation expression compensation based on relevant information to obtain an optimized consultation-reply semantic association feature vector, including: Calculating the product between the consultation-response semantic association feature vector and its transposed vector to obtain a consultation-response semantic association feature autocorrelation expression matrix; Calculating the coherent interference phase between the category decoding weight matrix of the reply quality classifier and the consultation-reply semantic association feature autocorrelation expression matrix to obtain the consultation-reply semantic association feature-category weight fine-grained coupling representation matrix; The consultation reply semantic association feature-category weight fine-grained coupling representation matrix is subjected to probabilistic processing based on a Sigmoid activation function to obtain a consultation reply semantic association feature-category weight fine-grained coupling probabilistic representation matrix; Based on the consultation reply semantic association feature-category weight fine-grained coupling probabilistic representation matrix, back-propagation expression compensation is performed on the consultation-reply semantic association feature vector to obtain a category-related compensation representation vector; A position-weighted sum between the category-related compensation representation vector and the consultation-reply semantic association feature vector is calculated to obtain the optimized consultation-reply semantic association feature vector.
8. A customer service voice intelligent detection system based on machine learning, characterized in that: include: A voice call data collection module is used to obtain the voice call data to be analyzed; A voice call data segmentation module is used to extract customer consultation voice data and intelligent customer service robot reply voice data from the voice call data to be analyzed; A semantic association feature encoding module, used to perform association semantic feature encoding on the customer consultation voice data and the intelligent customer service robot reply voice data to obtain a consultation-reply semantic association feature vector; The voice transfer intelligent detection result generation module is used to obtain a detection result of whether the voice call needs to be transferred to manual customer service based on the consultation-reply semantic association feature vector.
9. The customer service voice intelligent detection system based on machine learning according to claim 8 is characterized in that: The semantic association feature encoding module includes: A voice data speech recognition unit, used to perform speech recognition on the customer consultation voice data and the intelligent customer service robot reply voice data respectively to obtain a customer consultation text and a customer service robot reply text; A customer inquiry semantic feature extraction unit, used for extracting semantic features from the customer inquiry text to obtain a customer inquiry semantic feature vector; A customer service robot reply semantic feature extraction unit, used to extract semantic features from the customer service robot reply text to obtain a customer service robot reply semantic feature vector; The customer consultation semantic feature vector and the customer service reply robot semantic feature vector are associated to obtain the consultation-reply semantic association feature vector.
10. The customer service voice intelligent detection system based on machine learning according to claim 9 is characterized in that: The voice transfer intelligent detection result generation module includes: A feature optimization unit, used for performing back-propagation expression compensation on the consultation-reply semantic association feature vector based on relevant information to obtain an optimized consultation-reply semantic association feature vector; A feature vector classification unit, used to pass the optimized consultation-reply semantic association feature vector through a reply quality classifier to obtain a classification result, wherein the classification result is used to indicate whether the reply of the intelligent customer service robot effectively solves the customer's consultation problem; The semantic transfer unit is used to transfer the voice call to the manual customer service in a timely manner in response to the reply of the intelligent customer service robot not effectively solving the problem consulted by the customer.
Citation Information
Cited By
Unmanned cleaning robot cleaning control system based on deep learning
CN119165861A
