Voice quality inspection method and device based on artificial intelligence, computer equipment and medium

Abstract

The invention belongs to the field of artificial intelligence and the field of financial science and technology, and relates to a voice quality inspection method based on artificial intelligence, which comprises the following steps: receiving voice data to be subjected to quality inspection, and preprocessing the voice data to obtain target data; calling a concept recognition model to perform recognition processing on the target data to obtain a corresponding target concept; obtaining an explanation corresponding to the target concept; generating a loyalty quantized value and a readability quantized value corresponding to the interpretation based on an evaluation rule; screening a target explanation from all explanations based on the loyalty quantized value and the readability quantized value; screening target question keywords from the target explanations based on a keyword library; and generating a quality inspection report corresponding to the voice data based on the target question keyword. The invention further provides a voice quality inspection device based on artificial intelligence, computer equipment and a storage medium. In addition, the invention also relates to a block chain technology, and the quality inspection report can be stored in a block chain. According to the invention, the efficiency and accuracy of voice quality inspection are effectively improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence development technology and the field of financial technology, and in particular to artificial intelligence-based speech quality inspection methods, devices, computer equipment and storage media. Background Art

[0002] In the call center operations of financial enterprises, voice quality inspection is a key link in evaluating the quality of customer service. Voice quality inspection refers to the quality management and inspection of voice content, to check whether customer service or sales personnel use standard terms and whether the service is in place. Generally, the audio to be inspected is obtained by recording customer service calls, or by double-recording the sales process. By conducting voice quality inspection on the audio to be inspected, it can promote corporate marketing and market development, improve customer satisfaction, and improve customer service.

[0003] The current voice quality inspection method mainly uses manual evaluation and makes human judgments based on business rules. However, human judgments are inevitably prone to omissions and have problems such as low efficiency, long time consumption and high cost, making it impossible to effectively inspect voice quality. Summary of the invention

[0004] The purpose of the embodiments of the present application is to propose a speech quality inspection method, device, computer equipment and storage medium based on artificial intelligence to solve the technical problem that the existing speech quality inspection method mainly adopts manual listening and evaluation, and makes human judgment according to business rules. However, human judgment is inevitably prone to omissions, and has problems such as low efficiency, long time consumption and high cost, and cannot effectively perform speech quality inspection.

[0005] In order to solve the above technical problems, the present application embodiment provides a voice quality inspection method based on artificial intelligence, which adopts the following technical solutions:

[0006] Receiving voice data to be quality-checked, and preprocessing the voice data to obtain corresponding target data;

[0007] Calling a pre-built concept recognition model to perform recognition processing on the target data to obtain a target concept corresponding to the target data;

[0008] Obtaining an explanation corresponding to the target concept;

[0009] Generating a fidelity quantification value and a readability quantification value corresponding to the interpretation based on a preset evaluation rule;

[0010] Based on the fidelity quantification value and the readability quantification value, selecting a target interpretation from all the interpretations;

[0011] Filtering target question keywords from the target explanation based on a preset keyword library;

[0012] A quality inspection report corresponding to the voice data is generated based on the target question keywords.

[0013] Furthermore, the step of preprocessing the voice data to obtain corresponding target data specifically includes:

[0014] Performing text conversion processing on the voice data to obtain corresponding first text data;

[0015] Performing word segmentation processing on the first text data to obtain corresponding second text data;

[0016] Performing vectorization processing on the second text data to obtain corresponding vector data;

[0017] Performing standardization processing on the vector data to obtain corresponding processed data;

[0018] The processed data is used as the target data.

[0019] Furthermore, the step of generating the fidelity quantization value and the readability quantization value corresponding to the interpretation based on the preset evaluation rules specifically includes:

[0020] Get the preset fidelity algorithm and readability algorithm;

[0021] Calculating the interpretation based on the fidelity algorithm to obtain a fidelity quantization value corresponding to the interpretation;

[0022] The interpretation is calculated based on the readability algorithm to obtain a readability quantification value corresponding to the interpretation.

[0023] Furthermore, the step of selecting a target interpretation from all the interpretations based on the fidelity quantification value and the readability quantification value specifically includes:

[0024] Obtaining a loyalty threshold corresponding to the loyalty;

[0025] Get the readability threshold corresponding to the readability;

[0026] Filtering out, from all the interpretations, a designated interpretation whose fidelity quantization value is greater than the fidelity threshold and whose readability quantization value is greater than the readability threshold;

[0027] The designated interpretation is taken as the target interpretation.

[0028] Furthermore, the step of generating a quality inspection report corresponding to the voice data based on the target question keywords specifically includes:

[0029] Based on the preset information type, obtain the associated information corresponding to the target question keyword;

[0030] Generate corresponding potential question data based on the target question keyword and the associated information;

[0031] Call the preset quality inspection report template;

[0032] The quality inspection report corresponding to the voice data is generated based on the potential problem data and the quality inspection report template.

[0033] Furthermore, before the step of calling the pre-built concept recognition model to perform recognition processing on the target data to obtain the target concept corresponding to the target data, the step further includes:

[0034] Get pre-built speech sample data;

[0035] Call the preset initial recognition model;

[0036] Performing pre-training processing on the initial recognition model based on the speech sample data to obtain a corresponding first recognition model;

[0037] Acquire pre-built specified speech sample data corresponding to the specified concept;

[0038] Fine-tune the first recognition model based on the specified voice sample data to obtain a corresponding second recognition model;

[0039] The second recognition model is used as the concept recognition model.

[0040] Further, after the step of generating a quality inspection report corresponding to the voice data based on the target question keywords, the method further includes:

[0041] Determine the target team corresponding to the voice quality inspection process;

[0042] Call the preset data sending interface;

[0043] The quality inspection report is sent to the target team based on the data sending interface.

[0044] In order to solve the above technical problems, the embodiment of the present application also provides a speech quality inspection device based on artificial intelligence, which adopts the following technical solution:

[0045] A processing module, used for receiving voice data to be quality inspected, and preprocessing the voice data to obtain corresponding target data;

[0046] A recognition module is used to call a pre-built concept recognition model to perform recognition processing on the target data to obtain a target concept corresponding to the target data;

[0047] A first acquisition module, used to acquire an explanation corresponding to the target concept;

[0048] A first generating module, configured to generate a fidelity quantization value and a readability quantization value corresponding to the interpretation based on a preset evaluation rule;

[0049] A first screening module, configured to screen a target interpretation from all the interpretations based on the fidelity quantification value and the readability quantification value;

[0050] A second screening module, used to screen out target question keywords from the target explanation based on a preset keyword library;

[0051] The second generating module is used to generate a quality inspection report corresponding to the voice data based on the target question keywords.

[0052] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0053] Receiving voice data to be quality-checked, and preprocessing the voice data to obtain corresponding target data;

[0054] Calling a pre-built concept recognition model to perform recognition processing on the target data to obtain a target concept corresponding to the target data;

[0055] Obtaining an explanation corresponding to the target concept;

[0056] Generating a fidelity quantification value and a readability quantification value corresponding to the interpretation based on a preset evaluation rule;

[0057] Based on the fidelity quantification value and the readability quantification value, selecting a target interpretation from all the interpretations;

[0058] Filtering target question keywords from the target explanation based on a preset keyword library;

[0059] A quality inspection report corresponding to the voice data is generated based on the target question keywords.

[0060] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0061] Receiving voice data to be quality-checked, and preprocessing the voice data to obtain corresponding target data;

[0062] Calling a pre-built concept recognition model to perform recognition processing on the target data to obtain a target concept corresponding to the target data;

[0063] Obtaining an explanation corresponding to the target concept;

[0064] Generating a fidelity quantification value and a readability quantification value corresponding to the interpretation based on a preset evaluation rule;

[0065] Based on the fidelity quantification value and the readability quantification value, selecting a target interpretation from all the interpretations;

[0066] Filtering target question keywords from the target explanation based on a preset keyword library;

[0067] A quality inspection report corresponding to the voice data is generated based on the target question keywords.

[0068] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0069] The present application first receives speech data to be quality inspected, and preprocesses the speech data to obtain corresponding target data; then calls a pre-built concept recognition model to perform recognition processing on the target data to obtain a target concept corresponding to the target data; then obtains an explanation corresponding to the target concept; and generates a fidelity quantization value and a readability quantization value corresponding to the explanation based on preset evaluation rules; subsequently, based on the fidelity quantization value and the readability quantization value, a target explanation is screened out from all the explanations; further based on a preset keyword library, target problem keywords are screened out from the target explanation; finally, a quality inspection report corresponding to the speech data is generated based on the target problem keywords. After preprocessing the speech data to be quality inspected to obtain target data, the present application uses a concept recognition model to identify and process the target data to obtain a target concept, and obtains an explanation corresponding to the target concept, and then evaluates the fidelity and readability of the explanation based on the evaluation rules and screens out the target explanation, and then screens out target problem keywords from the target explanation based on a preset keyword library, and finally automatically and accurately generates a quality inspection report corresponding to the speech data based on the target problem keywords, thereby reducing reliance on manual evaluation, eliminating the need for excessive manual participation, and effectively improving the efficiency and accuracy of speech quality inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0071] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0072] Figure 2 A flowchart of an embodiment of a speech quality inspection method based on artificial intelligence according to the present application;

[0073] Figure 3 It is a structural schematic diagram of an embodiment of a speech quality inspection device based on artificial intelligence according to the present application;

[0074] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0075] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0076] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0077] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0078] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0079] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0080] The terminal devices 101, 102, and 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, and the like.

[0081] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0082] It should be noted that the artificial intelligence-based speech quality inspection method provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the artificial intelligence-based speech quality inspection device is generally arranged in the server / terminal device.

[0083] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0084] Continue to refer Figure 2 , showing a flow chart of an embodiment of a voice quality inspection method based on artificial intelligence according to the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted. The voice quality inspection method based on artificial intelligence provided in the embodiment of the present application can be applied to any scenario that requires voice quality inspection, and the voice quality inspection method based on artificial intelligence can be applied to products in these scenarios, for example, voice quality inspection in the financial insurance field. The voice quality inspection method based on artificial intelligence includes the following steps:

[0085] Step S201, receiving voice data to be quality-checked, and preprocessing the voice data to obtain corresponding target data.

[0086] In this embodiment, the electronic device (e.g. Figure 1 The server / terminal device shown in the figure) can obtain the voice data to be inspected through a wired connection or a wireless connection. It should be pointed out that the above-mentioned wireless connection method may include but is not limited to 3G / 4G / 5G connection, Wi-Fi connection, Bluetooth connection, Wi MAX connection, Zigbee connection, UWB (ultrawi deband) connection, and other wireless connection methods currently known or to be developed in the future. The execution subject of the present application may specifically be an intelligent voice quality inspection system. It can be applied to the voice quality inspection business scenario of customer service quality assessment in the call center of a financial enterprise (such as an insurance company or a bank). Among them, the above-mentioned voice data to be inspected may come from a call recording between a customer and a customer service representative, an online chat voice message, etc. In addition. The above-mentioned specific implementation process of preprocessing the voice data to obtain the corresponding target data will be further described in detail in the present application in subsequent specific embodiments, and will not be elaborated on here.

[0087] Step S202: calling a pre-built concept recognition model to perform recognition processing on the target data to obtain a target concept corresponding to the target data.

[0088] In this embodiment, the above-mentioned target data is input into a pre-built concept recognition model, and the concept recognition model parses the target data in the form of text vectors to identify the key concepts in the speech to be inspected, namely the above-mentioned target concepts. Key concepts are words, phrases or semantic patterns related to potential indications of customer service issues. Concept recognition identifies key concepts by comparing the input text with the learned concept representations. Among them, the number of the above-mentioned target concepts may include 1 or more. In addition, the construction process of the above-mentioned concept recognition model will be further described in detail in the subsequent specific embodiments of this application, and will not be elaborated on here.

[0089] Step S203: obtaining an explanation corresponding to the target concept.

[0090] In this embodiment, after the target concept corresponding to the target data is identified using the concept recognition model, explanations corresponding to each target concept are automatically generated for each target concept. For example, for "positive service attitude", an explanation may be generated: "The customer service staff always remains polite and patient during the call, actively asks about customer needs and provides solutions." Among them, all generated explanations are stored in the data center or cloud storage service of the financial enterprise, and are associated with corresponding call records, key concepts and other information. The data management system within the financial enterprise is responsible for organizing these explanations, including classification, tagging and indexing, for subsequent retrieval and analysis.

[0091] Step S204: Generate a fidelity quantization value and a readability quantization value corresponding to the interpretation based on a preset evaluation rule.

[0092] In this embodiment, the specific implementation process of generating the fidelity quantization value and the readability quantization value corresponding to the interpretation based on the preset evaluation rules will be further described in detail in subsequent specific embodiments of the present application, and will not be elaborated on here.

[0093] Step S205 , based on the fidelity quantification value and the readability quantification value, select a target interpretation from all the interpretations.

[0094] In this embodiment, the specific implementation process of selecting the target interpretation from all the interpretations based on the fidelity quantization value and the readability quantization value will be described in further detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0095] Step S206: Filter out target question keywords from the target explanation based on a preset keyword library.

[0096] In this embodiment, a keyword library related to potential problems in evaluating the service quality of customer service is pre-built according to actual business needs. It is pre-defined what types of problems need to be marked, such as improper service, customer dissatisfaction, inaccurate information, etc. and used as the above potential problems. Correspondingly, the keyword library related to potential problems at least includes keywords such as improper service, customer dissatisfaction, inaccurate information, etc. The above target interpretation can be matched with all the keywords contained in the preset keyword library, and the specified keywords that are successfully matched are used as the above target problem keywords.

[0097] Step S207: Generate a quality inspection report corresponding to the voice data based on the target question keywords.

[0098] In this embodiment, the specific implementation process of generating the quality inspection report corresponding to the voice data based on the target question keywords will be described in further detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0099] The present application first receives speech data to be quality inspected, and preprocesses the speech data to obtain corresponding target data; then calls a pre-built concept recognition model to perform recognition processing on the target data to obtain a target concept corresponding to the target data; then obtains an explanation corresponding to the target concept; and generates a fidelity quantization value and a readability quantization value corresponding to the explanation based on preset evaluation rules; subsequently, based on the fidelity quantization value and the readability quantization value, a target explanation is screened out from all the explanations; further based on a preset keyword library, target problem keywords are screened out from the target explanation; finally, a quality inspection report corresponding to the speech data is generated based on the target problem keywords. After preprocessing the speech data to be quality inspected to obtain target data, the present application uses a concept recognition model to identify and process the target data to obtain a target concept, and obtains an explanation corresponding to the target concept, and then evaluates the fidelity and readability of the explanation based on the evaluation rules and screens out the target explanation, and then screens out target problem keywords from the target explanation based on a preset keyword library, and finally automatically and accurately generates a quality inspection report corresponding to the speech data based on the target problem keywords, thereby reducing reliance on manual evaluation, eliminating the need for excessive manual participation, and effectively improving the efficiency and accuracy of speech quality inspection.

[0100] In some optional implementations, preprocessing the speech data to obtain corresponding target data in step S201 includes the following steps:

[0101] The voice data is converted into text to obtain corresponding first text data.

[0102] In this embodiment, the voice data may be converted into text form by using voice recognition technology, thereby obtaining the first text data.

[0103] The first text data is processed by word segmentation to obtain corresponding second text data.

[0104] In this embodiment, the above-mentioned word segmentation processing refers to the process of dividing a continuous text into individual words or phrases, which is necessary for subsequent model processing.

[0105] Vectorization is performed on the second text data to obtain corresponding vector data.

[0106] In this embodiment, the vectorization processing refers to converting the second text data after word segmentation into a numerical vector. The vectorization processing can be implemented by using word embeddings technology, such as Word2Vec, GlobE or BERT, which can map vocabulary to vectors in a high-dimensional space.

[0107] The vector data is standardized to obtain corresponding processed data.

[0108] In this embodiment, the above-mentioned standardization processing may at least include processing such as removing noise, adjusting volume, unifying sampling rate, etc. By performing standardization processing on the vector data, it is ensured that the model can process the input data more stably.

[0109] The processed data is used as the target data.

[0110] The present application performs text conversion processing on the speech data to obtain corresponding first text data; then performs word segmentation processing on the first text data to obtain corresponding second text data; then performs vectorization processing on the second text data to obtain corresponding vector data; subsequently, the vector data is standardized to obtain corresponding processed data, and the processed data is used as the target data. The present application performs text conversion processing, word segmentation processing, vectorization processing, and standardization processing on the speech data, so as to quickly and intelligently complete the preprocessing of the speech data, improve the processing efficiency of the preprocessing of the speech data, and ensure the data accuracy of the generated target data.

[0111] In some optional implementations of this embodiment, step S204 includes the following steps:

[0112] Get the preset fidelity algorithm and readability algorithm.

[0113] In this embodiment, the fidelity quantification is the output stability of the model under concept perturbation. The fidelity algorithm specifically includes: γ(a,ξ,δ)=|x|1∑ht∈f(x)δ(y,y′), where γ(a,ξ,δ): represents the fidelity quantification value. a: represents the perturbation parameter, which controls the degree of perturbation to the concept. ξ: represents the perturbation function, which perturbs the concept according to the perturbation parameter a. δ: represents the function that measures the output change, usually a distance metric, such as Euclidean distance or cosine similarity. x: represents the input text. f(x): represents the feature extraction function of the model for the input text x, and outputs a series of hidden states ht. y=g(ht): represents the probability distribution of the model output according to the hidden state ht when the concept is not perturbed. y′=g(ξ(ht,a)): represents the probability distribution of the model output according to the perturbed hidden state after the concept is perturbed. |x|1: represents the normalization of the input text length to ensure that texts of different lengths are comparable in fidelity quantification.

[0114] In addition, readability quantifies the consistency between the interpretation of the model output and human understanding. The readability algorithm specifically includes: μUC I(xi,xj) = logP(xi,xj) + c, where μUC I(xi,xj): represents the readability quantification value between vocabulary xi and xj. (xi,xj): represents the probability that vocabulary xi and xj appear together in the corpus. This probability reflects the degree of semantic association between the two words. c: is a constant term used to adjust the scale of readability quantification. In practical applications, an appropriate value can be selected as needed. log: represents the logarithmic operation, which helps to map the probability value to a more intuitive scale and amplify the difference between smaller probability values.

[0115] Specifically, the fidelity algorithm quantifies the stability of the model by measuring the output changes of the model under conceptual perturbations, while the readability algorithm quantifies the consistency between the explanation of the model output and human understanding by evaluating the co-occurrence probability between words. These two quantitative indicators together form the core of the conceptual explanation evaluation mechanism, which is used to optimize the speech quality inspection process.

[0116] The interpretation is calculated based on the fidelity algorithm to obtain a fidelity quantization value corresponding to the interpretation.

[0117] In this embodiment, the first calculation parameter required by the above-mentioned interpretation corresponding to the above-mentioned fidelity algorithm is obtained, and the first calculation parameter is substituted into the fidelity algorithm for calculation, so as to obtain a fidelity quantization value corresponding to the interpretation.

[0118] The interpretation is calculated based on the readability algorithm to obtain a readability quantification value corresponding to the interpretation.

[0119] In this embodiment, the second calculation parameter required by the readability algorithm corresponding to the above explanation is obtained, and the second calculation parameter is substituted into the readability algorithm for calculation to obtain the readability quantization value corresponding to the explanation.

[0120] The present application obtains a preset fidelity algorithm and a readability algorithm; then, based on the fidelity algorithm, the interpretation is calculated and processed to obtain a fidelity quantification value corresponding to the interpretation; and based on the readability algorithm, the interpretation is calculated and processed to obtain a readability quantification value corresponding to the interpretation. The present application calculates and processes the above interpretation based on the use of a fidelity algorithm and a readability algorithm, thereby achieving an accurate evaluation of the fidelity and readability of the interpretation, and generates a fidelity quantification value and a readability quantification value corresponding to the interpretation, thereby ensuring the data accuracy of the generated fidelity quantification value and readability quantification value.

[0121] In some optional implementations, step S205 includes the following steps:

[0122] Gets the loyalty threshold corresponding to the loyalty level.

[0123] In this embodiment, there is no specific limitation on the value of the loyalty threshold. The first minimum threshold of loyalty can be set according to actual business needs and previous related analysis to serve as the loyalty threshold.

[0124] Gets the readability threshold corresponding to the readability.

[0125] In this embodiment, there is no specific limitation on the value of the readability threshold. The second lowest threshold of readability can be set according to actual business needs and previous related analysis to serve as the loyalty threshold.

[0126] A designated interpretation whose fidelity quantization value is greater than the fidelity threshold and whose readability quantization value is greater than the readability threshold is screened out from all the interpretations.

[0127] In this embodiment, the above-mentioned loyalty threshold and readability threshold can be used to perform numerical comparisons on the loyalty quantization values ​​and readability quantization values ​​corresponding to all the interpretations, so as to screen out the specified interpretations that meet the conditions that the loyalty quantization value is greater than the loyalty threshold and the readability quantization value is greater than the readability threshold.

[0128] The designated interpretation is taken as the target interpretation.

[0129] In this embodiment, the target explanation obtained after screening is a high-quality explanation among all the explanations.

[0130] The present application obtains a loyalty threshold corresponding to the fidelity; and obtains a readability threshold corresponding to the readability; and then selects from all the interpretations a designated interpretation whose fidelity quantization value is greater than the loyalty threshold, and whose readability quantization value is greater than the readability threshold; and subsequently uses the designated interpretation as the target interpretation. The present application numerically analyzes the fidelity quantization values ​​and readability quantization values ​​corresponding to all the interpretations based on the obtained loyalty threshold corresponding to the fidelity, and the readability threshold corresponding to the readability, so as to quickly and accurately select from all the interpretations a high-quality interpretation whose fidelity quantization value is greater than the loyalty threshold, and whose readability quantization value is greater than the readability threshold, i.e., the above-mentioned target interpretation, thereby ensuring the data accuracy of the obtained target interpretation.

[0131] In some optional implementations, step S207 includes the following steps:

[0132] Based on the preset information type, the associated information corresponding to the target question keyword is obtained.

[0133] In this embodiment, the above information types may include at least explanation type, occurrence frequency type, and number of customers affected. Correspondingly, the above associated information corresponding to the target question keyword may include explanation information, namely the above target explanation, occurrence frequency, and number of customers affected.

[0134] Corresponding potential question data is generated based on the target question keyword and the associated information.

[0135] In this embodiment, the target question keywords correspond to potential problems in the voice data, which can also be understood as descriptions of potential problems. All information related to the potential problem, i.e., the potential problem data, can be obtained by sorting the target question keywords and the associated information.

[0136] Call the preset quality inspection report template.

[0137] In this embodiment, the above-mentioned quality inspection report template can be a report template file pre-constructed according to actual quality inspection analysis requirements. There is no specific limitation on the report format of the quality inspection report template. For example, it can be in the form of Word document, PDF or Excel table.

[0138] The quality inspection report corresponding to the voice data is generated based on the potential problem data and the quality inspection report template.

[0139] In this embodiment, the potential problem data can be filled into the corresponding position in the quality inspection report template to generate a quality inspection report corresponding to the voice data. In particular, charts or visualization tools can be used to enhance the readability and comprehensibility of the generated quality inspection report.

[0140] The present application obtains the associated information corresponding to the target question keyword based on a preset information type; then generates corresponding potential problem data based on the target question keyword and the associated information; then calls a preset quality inspection report template; and subsequently generates the quality inspection report corresponding to the voice data based on the potential problem data and the quality inspection report template. The present application obtains the associated information corresponding to the target question keyword based on the information type, generates corresponding potential problem data based on the target question keyword and the associated information, and then processes the potential problem data based on the use of the quality inspection report template, thereby realizing the rapid and intelligent generation of the quality inspection report corresponding to the voice data, and improving the efficiency of generating quality inspection reports.

[0141] In some optional implementations of this embodiment, before step S202, the electronic device may further perform the following steps:

[0142] Get pre-built speech sample data.

[0143] In this embodiment, a large amount of diverse initial speech data is collected in advance, and these data should cover different languages, accents, sound quality, background noise, etc. to ensure the generalization ability of the model. The source of the initial speech data can be a public speech data set, self-recorded speech data, or a speech sample generated by speech synthesis technology. Then the collected initial speech data is accurately annotated with text, that is, a corresponding text transcription is provided for each speech sample. And the annotation process needs to ensure accuracy and consistency to reduce the impact of annotation errors on model performance. After that, the annotated initial speech data is preprocessed, including denoising, normalization, framing, windowing and other operations to improve the quality of the speech signal, so as to construct the above-mentioned speech sample data. Among them, the preprocessed speech data will be easier for the model to process, thereby improving the recognition accuracy of the model.

[0144] Call the preset initial recognition model.

[0145] In this embodiment, there is no specific limitation on the model type selection of the above-mentioned initial recognition model, which can be determined according to the actual task requirements. For example, any one of deep neural network (DNN), convolutional neural network (CNN), recurrent neural network (RNN) or their variants (such as LSTM, GRU) can be selected.

[0146] The initial recognition model is pre-trained based on the speech sample data to obtain a corresponding first recognition model.

[0147] In this embodiment, the above-mentioned speech sample data is first converted into a feature vector that can be processed by the model by adopting a feature extraction method, such as Mel Frequency Cepstral Coefficient (MFCC) or Filter Bank technology. Then, the above-mentioned initial recognition model is pre-trained using the labeled speech sample data to learn the mapping relationship between the speech signal and the text. During the pre-training process, the initial recognition model will try to recognize various speech patterns, and construct corresponding acoustic models and language models, and finally obtain a pre-trained first recognition model.

[0148] Get pre-built specified speech sample data corresponding to the specified concept.

[0149] In this embodiment, designated speech sample data related to the above-mentioned designated concepts are collected in advance, and these designated speech sample data are representative and accurately labeled. In addition, the data volume of the designated speech sample data may be relatively small, but it should be sufficient to cover the main speech patterns and characteristics of the designated field. The above-mentioned designated field is specifically a vocabulary, phrase or semantic pattern related to potential indications of customer service problems.

[0150] The first recognition model is fine-tuned based on the specified voice sample data to obtain a corresponding second recognition model.

[0151] In this embodiment, the pre-trained first recognition model is first loaded into the training environment, and then the first recognition model is fine-tuned using the specified speech sample data of the specified field to optimize the performance of the model in the specified field. During the fine-tuning process, the model parameters, learning rate and other hyperparameters are adjusted to adapt to the data characteristics of the specified field. Regularization, data enhancement and other techniques are used to prevent overfitting and improve the generalization ability of the model. The fine-tuned model is further evaluated using a validation set to check the recognition accuracy and performance of the model in a specific field. The model is iteratively optimized based on the evaluation results, including adjusting the model structure, increasing training data, etc., to obtain the corresponding second recognition model. Among them, some speech data can be randomly selected from the above-mentioned specified speech sample data as the above-mentioned validation set.

[0152] The second recognition model is used as the concept recognition model.

[0153] In this embodiment, after the concept recognition model is obtained, the concept recognition model can be deployed in actual applications for users to use, and the performance of the concept recognition model is continuously monitored, and necessary adjustments and optimizations are made to the monitored concept recognition model based on user feedback.

[0154] The present application obtains pre-constructed speech sample data; then calls a preset initial recognition model; then pre-trains the initial recognition model based on the speech sample data to obtain a corresponding first recognition model; subsequently obtains pre-constructed designated speech sample data corresponding to a designated concept; further fine-tunes the first recognition model based on the designated speech sample data to obtain a corresponding second recognition model; and finally uses the second recognition model as the concept recognition model. The present application obtains a first recognition model by pre-training the initial recognition model using pre-constructed speech sample data, and then fine-tunes the first recognition model using pre-constructed designated speech sample data corresponding to a designated concept, thereby realizing fast and accurate training of a concept recognition model with good generalization ability and recognition accuracy in a designated field, effectively improving the construction intelligence of the concept recognition model and ensuring the recognition effect of the concept recognition model.

[0155] In some optional implementations of this embodiment, after step S207, the electronic device may further perform the following steps:

[0156] Identify the target team for voice quality control processing.

[0157] In this embodiment, the target team may specifically be a team responsible for handling matters related to voice quality inspection, such as a customer service team, a product team, etc.

[0158] Call the preset data sending interface.

[0159] In this embodiment, the data sending interface is a pre-built interface with a data sending function.

[0160] The quality inspection report is sent to the target team based on the data sending interface.

[0161] In this embodiment, the communication address of the target team can be obtained, and then the quality inspection report can be sent to the target team using the data sending interface according to the communication address, wherein the communication address includes an email address or a mobile phone number.

[0162] This application determines the target team corresponding to the voice quality inspection process; then calls the preset data sending interface; and subsequently sends the quality inspection report to the target team based on the data sending interface. After generating a quality inspection report corresponding to the voice data based on the target question keywords, this application will also intelligently call the data sending interface to send the quality inspection report to the target team corresponding to the voice quality inspection process, so that the target team can respond quickly and take measures to improve customer satisfaction, which is conducive to improving the work experience and work efficiency of the target team.

[0163] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0164] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned quality inspection report, the above-mentioned quality inspection report can also be stored in a blockchain node.

[0165] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0166] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0167] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0168] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0169] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0170] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a speech quality inspection device based on artificial intelligence, and the device embodiment is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0171] like Figure 3As shown, the speech quality inspection device 300 based on artificial intelligence described in this embodiment includes: a processing module 301, a recognition module 302, a first acquisition module 303, a first generation module 304, a first screening module 305, a second screening module 306 and a second generation module 307. Among them:

[0172] The processing module 301 is used to receive the voice data to be quality inspected and pre-process the voice data to obtain the corresponding target data;

[0173] The recognition module 302 is used to call a pre-built concept recognition model to perform recognition processing on the target data to obtain a target concept corresponding to the target data;

[0174] A first acquisition module 303, used to acquire an explanation corresponding to the target concept;

[0175] A first generating module 304, configured to generate a fidelity quantization value and a readability quantization value corresponding to the interpretation based on a preset evaluation rule;

[0176] A first screening module 305, configured to screen a target interpretation from all the interpretations based on the fidelity quantization value and the readability quantization value;

[0177] A second screening module 306 is used to screen target question keywords from the target explanation based on a preset keyword library;

[0178] The second generating module 307 is used to generate a quality inspection report corresponding to the speech data based on the target question keywords.

[0179] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based speech quality inspection method in the aforementioned implementation mode, and will not be repeated here.

[0180] In some optional implementations of this embodiment, the processing module 301 includes:

[0181] A first processing submodule, configured to perform text conversion processing on the voice data to obtain corresponding first text data;

[0182] A second processing submodule, configured to perform word segmentation processing on the first text data to obtain corresponding second text data;

[0183] A third processing submodule, used for performing vectorization processing on the second text data to obtain corresponding vector data;

[0184] A fourth processing submodule is used to perform standardization processing on the vector data to obtain corresponding processed data;

[0185] The first determining submodule is used to use the processed data as the target data.

[0186] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based speech quality inspection method in the aforementioned implementation mode, and will not be repeated here.

[0187] In some optional implementations of this embodiment, the first generating module 304 includes:

[0188] A first acquisition submodule is used to acquire a preset fidelity algorithm and a readability algorithm;

[0189] A first calculation submodule, configured to perform calculation processing on the interpretation based on the fidelity algorithm to obtain a fidelity quantization value corresponding to the interpretation;

[0190] The second calculation submodule is used to perform calculation processing on the interpretation based on the readability algorithm to obtain a readability quantization value corresponding to the interpretation.

[0191] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based speech quality inspection method in the aforementioned implementation mode, and will not be repeated here.

[0192] In some optional implementations of this embodiment, the first screening module 305 includes:

[0193] The second acquisition submodule is used to acquire a loyalty threshold corresponding to the loyalty;

[0194] A third acquisition submodule is used to acquire a readability threshold corresponding to the readability;

[0195] A screening submodule, for screening out, from all the interpretations, a designated interpretation whose fidelity quantization value is greater than the fidelity threshold and whose readability quantization value is greater than the readability threshold;

[0196] The second determination submodule is used to take the designated interpretation as the target interpretation.

[0197] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based speech quality inspection method in the aforementioned implementation mode, and will not be repeated here.

[0198] In some optional implementations of this embodiment, the second generating module 307 includes:

[0199] A fourth acquisition submodule is used to acquire associated information corresponding to the target question keyword based on a preset information type;

[0200] A first generating submodule, configured to generate corresponding potential question data based on the target question keyword and the associated information;

[0201] Call submodule, used to call the preset quality inspection report template;

[0202] The second generating submodule is used to generate the quality inspection report corresponding to the voice data based on the potential problem data and the quality inspection report template.

[0203] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based speech quality inspection method in the aforementioned implementation mode, and will not be repeated here.

[0204] In some optional implementations of this embodiment, the speech quality inspection device based on artificial intelligence further includes:

[0205] A second acquisition module is used to acquire pre-constructed voice sample data;

[0206] A first calling module is used to call a preset initial recognition model;

[0207] A pre-training module, used for pre-training the initial recognition model based on the speech sample data to obtain a corresponding first recognition model;

[0208] A third acquisition module is used to acquire pre-built specified speech sample data corresponding to the specified concept;

[0209] A fine-tuning module, configured to fine-tune the first recognition model based on the specified speech sample data to obtain a corresponding second recognition model;

[0210] The first determining module is used to use the second recognition model as the concept recognition model.

[0211] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based speech quality inspection method in the aforementioned implementation mode, and will not be repeated here.

[0212] In some optional implementations of this embodiment, the speech quality inspection device based on artificial intelligence further includes:

[0213] The second determination module is used to determine the target team corresponding to the voice quality inspection process;

[0214] The second calling module is used to call a preset data sending interface;

[0215] A sending module is used to send the quality inspection report to the target team based on the data sending interface.

[0216] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based speech quality inspection method in the aforementioned implementation mode, and will not be repeated here.

[0217] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0218] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital processors (DSPs), embedded devices, etc.

[0219] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0220] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart memory card (Smart Med ia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of a voice quality inspection method based on artificial intelligence, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0221] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run the computer-readable instructions stored in the memory 41 or process data, such as computer-readable instructions for running the artificial intelligence-based voice quality inspection method.

[0222] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0223] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0224] In an embodiment of the present application, voice data to be quality inspected is first received, and the voice data is preprocessed to obtain corresponding target data; then a pre-constructed concept recognition model is called to perform recognition processing on the target data to obtain a target concept corresponding to the target data; then an explanation corresponding to the target concept is obtained; and a fidelity quantization value and a readability quantization value corresponding to the explanation are generated based on preset evaluation rules; subsequently, a target explanation is screened out from all the explanations based on the fidelity quantization value and the readability quantization value; further, target problem keywords are screened out from the target explanation based on a preset keyword library; finally, a quality inspection report corresponding to the voice data is generated based on the target problem keywords. After preprocessing the speech data to be quality inspected to obtain target data, the present application uses a concept recognition model to identify and process the target data to obtain a target concept, and obtains an explanation corresponding to the target concept, and then evaluates the fidelity and readability of the explanation based on the evaluation rules and screens out the target explanation, and subsequently screens out target problem keywords from the target explanation based on a preset keyword library, and finally the target problem keywords automatically and accurately generate a quality inspection report corresponding to the speech data, reducing dependence on manual evaluation, eliminating the need for excessive manual participation, and effectively improving the efficiency and accuracy of speech quality inspection.

[0225] The present application also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the above-mentioned artificial intelligence-based speech quality inspection method.

[0226] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0227] In an embodiment of the present application, voice data to be quality inspected is first received, and the voice data is preprocessed to obtain corresponding target data; then a pre-constructed concept recognition model is called to perform recognition processing on the target data to obtain a target concept corresponding to the target data; then an explanation corresponding to the target concept is obtained; and a fidelity quantization value and a readability quantization value corresponding to the explanation are generated based on preset evaluation rules; subsequently, a target explanation is screened out from all the explanations based on the fidelity quantization value and the readability quantization value; further, target problem keywords are screened out from the target explanation based on a preset keyword library; finally, a quality inspection report corresponding to the voice data is generated based on the target problem keywords. After preprocessing the speech data to be quality inspected to obtain target data, the present application uses a concept recognition model to identify and process the target data to obtain a target concept, and obtains an explanation corresponding to the target concept, and then evaluates the fidelity and readability of the explanation based on the evaluation rules and screens out the target explanation, and subsequently screens out target problem keywords from the target explanation based on a preset keyword library, and finally the target problem keywords automatically and accurately generate a quality inspection report corresponding to the speech data, reducing dependence on manual evaluation, eliminating the need for excessive manual participation, and effectively improving the efficiency and accuracy of speech quality inspection.

[0228] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0229] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A speech quality inspection method based on artificial intelligence, characterized in that: The steps include: Receiving voice data to be quality-checked, and preprocessing the voice data to obtain corresponding target data; Calling a pre-built concept recognition model to perform recognition processing on the target data to obtain a target concept corresponding to the target data; Obtaining an explanation corresponding to the target concept; Generating a fidelity quantification value and a readability quantification value corresponding to the interpretation based on a preset evaluation rule; Based on the fidelity quantification value and the readability quantification value, selecting a target interpretation from all the interpretations; Filtering target question keywords from the target explanation based on a preset keyword library; A quality inspection report corresponding to the voice data is generated based on the target question keywords.

2. The method for speech quality inspection based on artificial intelligence according to claim 1, characterized in that: The step of preprocessing the voice data to obtain corresponding target data specifically includes: Performing text conversion processing on the voice data to obtain corresponding first text data; Performing word segmentation processing on the first text data to obtain corresponding second text data; Performing vectorization processing on the second text data to obtain corresponding vector data; Performing standardization processing on the vector data to obtain corresponding processed data; The processed data is used as the target data.

3. The speech quality inspection method based on artificial intelligence according to claim 1, characterized in that: The step of generating a fidelity quantization value and a readability quantization value corresponding to the interpretation based on a preset evaluation rule specifically includes: Get the preset fidelity algorithm and readability algorithm; Calculating the interpretation based on the fidelity algorithm to obtain a fidelity quantization value corresponding to the interpretation; The interpretation is calculated based on the readability algorithm to obtain a readability quantification value corresponding to the interpretation.

4. The method for speech quality inspection based on artificial intelligence according to claim 1, characterized in that: The step of selecting a target interpretation from all the interpretations based on the fidelity quantification value and the readability quantification value specifically includes: Obtaining a loyalty threshold corresponding to the loyalty; Get the readability threshold corresponding to the readability; Filtering out, from all the interpretations, a designated interpretation whose fidelity quantization value is greater than the fidelity threshold and whose readability quantization value is greater than the readability threshold; The designated interpretation is taken as the target interpretation.

5. The method for speech quality inspection based on artificial intelligence according to claim 1, characterized in that: The step of generating a quality inspection report corresponding to the voice data based on the target question keywords specifically includes: Based on the preset information type, obtain the associated information corresponding to the target question keyword; Generate corresponding potential question data based on the target question keyword and the associated information; Call the preset quality inspection report template; The quality inspection report corresponding to the voice data is generated based on the potential problem data and the quality inspection report template.

6. The method for speech quality inspection based on artificial intelligence according to claim 1, characterized in that: Before the step of calling the pre-built concept recognition model to perform recognition processing on the target data to obtain the target concept corresponding to the target data, the method further includes: Get pre-built speech sample data; Call the preset initial recognition model; Performing pre-training processing on the initial recognition model based on the speech sample data to obtain a corresponding first recognition model; Acquire pre-built specified speech sample data corresponding to the specified concept; Fine-tune the first recognition model based on the specified voice sample data to obtain a corresponding second recognition model; The second recognition model is used as the concept recognition model.

7. The method for speech quality inspection based on artificial intelligence according to claim 1, characterized in that: After the step of generating a quality inspection report corresponding to the voice data based on the target question keywords, the method further includes: Determine the target team corresponding to the voice quality inspection process; Call the preset data sending interface; The quality inspection report is sent to the target team based on the data sending interface.

8. A speech quality inspection device based on artificial intelligence, characterized in that: include: A processing module, used for receiving voice data to be quality inspected, and preprocessing the voice data to obtain corresponding target data; A recognition module is used to call a pre-built concept recognition model to perform recognition processing on the target data to obtain a target concept corresponding to the target data; A first acquisition module, used to acquire an explanation corresponding to the target concept; A first generating module, configured to generate a fidelity quantization value and a readability quantization value corresponding to the interpretation based on a preset evaluation rule; A first screening module, configured to screen a target interpretation from all the interpretations based on the fidelity quantification value and the readability quantification value; A second screening module, used to screen out target question keywords from the target explanation based on a preset keyword library; The second generating module is used to generate a quality inspection report corresponding to the voice data based on the target question keywords.

9. A computer device, characterized in that: It includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the speech quality inspection method based on artificial intelligence as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the artificial intelligence-based speech quality inspection method according to any one of claims 1 to 7.

Images