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

Abstract

The invention relates to the technical field of artificial intelligence, and discloses a voice call quality inspection method, device and equipment based on artificial intelligence, and a medium, and the method comprises the steps: obtaining a to-be-inspected audio according to a quality inspection request, carrying out the track division processing of the to-be-inspected audio, and obtaining a customer service audio; obtaining a first key text corresponding to the audio to be subjected to quality inspection and a second key text corresponding to the customer service audio; judging whether preset prohibited words exist in the second key text or not; if the preset prohibited words do not exist in the second key text, inputting the customer service audio and the first key text into a trained emotion recognition model, and obtaining an emotion quality inspection result; performing business process quality inspection on the second key text to obtain a business quality inspection result; performing verbal skill key point quality inspection on the second key text to obtain a verbal skill quality inspection result; and determining a final quality inspection result of the audio to be subjected to quality inspection according to the above results. The technical problem that an existing voice call quality inspection method is low in efficiency and accuracy is solved.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence technology and natural language processing technology, and in particular to a voice call quality inspection method, device, equipment and medium based on artificial intelligence. Background Art

[0002] In the insurance industry, the quality of customer service is directly related to the company's brand image and customer satisfaction. As an important means of evaluating and improving service quality, voice quality inspection is of great importance. However, the traditional voice quality inspection method relies on manual listening to recordings, which is not only inefficient, but also highly subjective and difficult to standardize. With the continuous increase in insurance business volume and the increasing diversification of customer needs, traditional quality inspection methods can no longer meet the needs of insurance companies.

[0003] In the current voice call quality inspection scheme, the entire audio clip generated between the customer and the customer service is usually analyzed directly according to the preset rules to determine whether the customer service responds to the customer's request according to the preset rules. Since the customer's request has a certain flexibility, this quality inspection method has low accuracy. In addition, since it is difficult to extract keyword information from the audio, it is necessary to analyze the entire audio clip, resulting in low efficiency of voice quality inspection. Summary of the invention

[0004] The present invention provides a voice call quality inspection method, device, equipment and medium based on artificial intelligence to solve the technical problems of low efficiency and accuracy of existing voice call quality inspection methods.

[0005] In a first aspect, a voice call quality inspection method based on artificial intelligence is provided, comprising:

[0006] Acquire the audio to be inspected according to the quality inspection request, and perform track splitting on the audio to be inspected to obtain customer service audio;

[0007] Obtain the first key text corresponding to the audio to be inspected and the second key text corresponding to the customer service audio;

[0008] Determine whether there is a preset prohibited word in the second key text, and if the preset prohibited word exists in the second key text, determine the final quality inspection result of the audio to be inspected;

[0009] If the preset prohibited words do not exist in the second key text, input the customer service audio and the first key text into a trained emotion recognition model to obtain an emotion quality inspection result, wherein the emotion recognition model is trained by samples and labels;

[0010] According to preset business rules, a business process quality inspection is performed on the second key text to obtain a business quality inspection result;

[0011] According to the preset speech rules, the second key text is quality checked on the key points of the speech to obtain the speech quality check result;

[0012] The final quality inspection result of the audio to be quality inspected is determined based on the emotion quality inspection result, the business quality inspection result and the speech quality inspection result.

[0013] In a second aspect, a voice call quality inspection device based on artificial intelligence is provided, comprising:

[0014] An audio extraction module is used to obtain the audio to be inspected according to the quality inspection request, and perform track processing on the audio to be inspected to obtain customer service audio;

[0015] A key text module, used to obtain a first key text corresponding to the audio to be quality inspected and a second key text corresponding to the customer service audio;

[0016] a prohibited word determination module, configured to determine whether there is a preset prohibited word in the second key text, and if the preset prohibited word exists in the second key text, determine a final quality inspection result of the audio to be inspected;

[0017] An emotion quality inspection module, configured to input the customer service audio and the first key text into a trained emotion recognition model to obtain an emotion quality inspection result if the preset prohibited words do not exist in the second key text, wherein the emotion recognition model is trained by samples and labels;

[0018] A business quality inspection module, used to perform business process quality inspection on the second key text according to preset business rules and obtain business quality inspection results;

[0019] A speech quality inspection module, used to perform speech key point quality inspection on the second key text according to preset speech rules, and obtain speech quality inspection results;

[0020] A result determination module is used to determine the final quality inspection result of the audio to be quality inspected based on the emotion quality inspection result, the business quality inspection result and the speech quality inspection result.

[0021] In a third aspect, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned artificial intelligence-based voice call quality inspection method when executing the computer program.

[0022] In a fourth aspect, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the above-mentioned steps of artificial intelligence-based voice call quality inspection are implemented.

[0023] In the scheme implemented by the above-mentioned artificial intelligence-based voice call quality inspection method, device, equipment and medium, after receiving the quality inspection request, the audio to be inspected is extracted, and the audio to be inspected is processed by splitting the tracks to obtain the customer service audio; then the first key text of the audio to be inspected and the second key text of the customer service audio are extracted; it is determined whether the second key text contains preset banned words. If there are preset banned words, it means that the customer service quality is poor, and the final quality inspection result of the customer service can be directly determined. In the embodiment of the present invention, by judging the preset banned words, the final quality inspection results of a part of voice calls with obviously poor service quality can be determined first, without the need for subsequent emotional quality inspection, business process quality inspection and speech key points quality inspection, thereby improving the efficiency of voice call quality inspection and enabling batch voice call quality inspection.

[0024] If there are no preset banned words, the customer service quality needs to be further evaluated. First, the customer service audio and the first key text are combined with the emotion recognition model to obtain the emotion quality inspection result; then the business quality inspection result and the speech quality inspection result are combined to determine the final quality inspection result. In the embodiment of the present invention, multimodal fusion of voice and text is performed through customer service audio and the first key text, so that the emotion recognition result is more accurate and the accuracy of the voice call quality inspection result is improved; in addition, the embodiment of the present invention combines emotion quality inspection, business process quality inspection and speech key point quality inspection to judge the quality inspection results of voice calls from multiple aspects, which can further improve the accuracy of the voice call quality inspection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0026] Figure 1 It is a schematic diagram of an application environment of a voice call quality inspection method based on artificial intelligence in one embodiment of the present invention;

[0027] Figure 2 A flowchart of a voice call quality inspection method based on artificial intelligence provided by an embodiment of the present invention;

[0028] Figure 3 for Figure 2 A schematic diagram of a process of step S20;

[0029] Figure 4 for Figure 2 A schematic diagram of a process of step S40;

[0030] Figure 5 for Figure 2 A schematic diagram of a process of step S50;

[0031] Figure 6 for Figure 2 A schematic diagram of a process of step S60;

[0032] Figure 7 A structural diagram of a voice call quality inspection device based on artificial intelligence provided by an embodiment of the present invention;

[0033] Figure 8 is a schematic diagram of a structure of a computer device in one embodiment of the present invention;

[0034] Fig. 9 It is another structural schematic diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] The voice call quality inspection method based on artificial intelligence provided by the embodiment of the present invention can be applied in the following aspects: Figure 1 In an application environment, the client communicates with the server through a network. The server can receive a quality inspection request through the client, and obtain the audio to be inspected according to the quality inspection request, perform track processing on the audio to be inspected, and obtain the customer service audio; obtain the first key text corresponding to the audio to be inspected and the second key text corresponding to the customer service audio; determine whether there are preset prohibited words in the second key text, and if there are preset prohibited words in the second key text, determine the final quality inspection result of the audio to be inspected; if there are no preset prohibited words in the second key text, input the customer service audio and the first key text into the trained emotion recognition model to obtain the emotion quality inspection result, wherein the emotion recognition model is trained by samples and labels; according to the preset business rules, perform business process quality inspection on the second key text to obtain the business quality inspection result; according to the preset speech rules, perform speech key points quality inspection on the second key text to obtain the speech quality inspection result; according to the emotion quality inspection result, the business quality inspection result and the speech quality inspection result, determine the final quality inspection result of the audio to be inspected.

[0037] In the embodiment of the present invention, by pre-setting the judgment of banned words, the final quality inspection results of a part of voice calls with obviously poor service quality can be determined first, and there is no need to conduct subsequent emotion quality inspection, business process quality inspection and speech key points quality inspection, thereby improving the efficiency of voice call quality inspection, and batch voice calls can be quality inspected; and multi-modal fusion of voice text is performed through customer service audio and the first key text, so that the result of emotion recognition is more accurate, and the accuracy of the voice call quality inspection result is improved; in addition, the embodiment of the present invention combines emotion quality inspection, business process quality inspection and speech key points quality inspection, and judges the quality inspection results of voice calls from multiple aspects, which can further improve the accuracy of the voice call quality inspection results.

[0038] The client may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and portable wearable devices. The server may be implemented by an independent server or a server cluster consisting of multiple servers. The present invention is described in detail below through specific embodiments.

[0039] A flow chart of a voice call quality inspection method based on artificial intelligence provided by an embodiment of the present invention is shown in FIG. Figure 2 As shown, the method comprises the following steps:

[0040] S10: Acquire the audio to be inspected according to the quality inspection request, and perform track separation on the audio to be inspected to obtain customer service audio.

[0041] The artificial intelligence-based voice call quality inspection method provided by the embodiment of the present invention can be applied to manual telephone customer service, intelligent customer service or intelligent assistant and other manual service systems and intelligent question-answering engines in various application scenarios. Intelligent customer service or intelligent assistant is usually implemented through a server, which can receive quality inspection requests in real time. For example, in the field of insurance applications, customers often consult insurance-related knowledge by phone, and often need manual customer service or the help of intelligent question-answering engines to answer some of the customers' insurance consultation questions. By quality inspecting the voice call audio between manual customer service and customers or the voice call audio between intelligent customer service and customers, it is possible to understand the problems that arise in the customer service process, and train manual customer service or improve intelligent customer service according to the problems that arise, thereby improving insurance promotion efficiency and user experience.

[0042] Intelligent customer service or intelligent assistant is one of the most extensive and important ways to implement natural language processing (NLP) technology in actual scenarios. Applying intelligent customer service assistants in the consulting business of the insurance industry can improve the efficiency of user consultation processing and reduce the number of manual customer service, thereby reducing labor costs. In the embodiment of the present invention, the quality inspection of the voice call quality of the intelligent customer service system is used as an example to illustrate. It is easy to understand that intelligent customer service requires a large number of samples for deep learning training, and the trained intelligent customer service needs to be tested on the test samples to observe whether the trained intelligent customer service meets the requirements. The voice call quality inspection method provided in the embodiment of the present invention can inspect the quality of sample voice calls and screen out voice calls with poor quality inspection results. Through the screened voice calls, the neural network structure or parameters of the intelligent customer service are readjusted to optimize the service quality of the intelligent customer service.

[0043] When quality inspection is required, a quality inspection request is sent to the server. After receiving the quality inspection request, the server finds the corresponding voice call audio according to the voice ID contained in the quality inspection request, and uses the voice call audio as the audio to be inspected. The voice ID represents the ID number of each voice call audio, and each voice call audio can be distinguished by the voice ID.

[0044] Then, the audio to be inspected is split into tracks to obtain customer service audio and client audio. It is easy to understand that the audio to be inspected contains the audio of both the customer and the customer service. Through audio track splitting, a mixed audio file containing multiple audio elements (such as human voices, musical instruments, etc.) can be separated into multiple independent audio tracks. In the specific implementation of the audio track splitting process, the frequency separation method can be used, such as using the filter in the audio editing software to separate according to the frequency range of different audio elements; a high-pass filter can be set to remove the low-frequency part to separate the high-frequency instrument sound, and a low-pass filter can be set to remove the high-frequency part to separate the low-frequency bass sound, etc., and the filter parameters can be continuously adjusted until a more satisfactory separation effect is obtained.

[0045] S20: Obtain a first key text corresponding to the audio to be quality inspected and a second key text corresponding to the customer service audio.

[0046] Then, the audio to be inspected is converted into the first text, and invalid information in the first text, such as pause words, repeated words, etc., is filtered out, and key information in the first text is extracted to obtain the first key text. For example, if the first text is "Hello, hello, um, hello, do I meet the insurance requirements for Ping An Fu No. 20?", then the meaningless words such as "hello, hello", "um" and so on in the first text are removed, and the first key text is "Hello, hello, do I meet the insurance requirements for Ping An Fu No. 20?".

[0047] As an example, a flow chart of step S20 in an embodiment of the present invention is as follows: Figure 3 As shown, the step of extracting the first key text corresponding to the audio to be inspected and the second key text corresponding to the customer service audio includes:

[0048] S21: using automatic speech recognition technology to recognize the texts corresponding to the audio to be inspected and the customer service audio respectively, to obtain the text corresponding to the audio to be inspected and the text corresponding to the customer service audio;

[0049] S22: Perform keyword extraction on the text corresponding to the audio to be quality inspected and the text corresponding to the customer service audio respectively to obtain the first key text and the second key text.

[0050] When extracting the first key text of the audio to be inspected in the embodiment of the present invention, the audio to be inspected is firstly subjected to text recognition using automatic speech recognition technology. Speech recognition software can be used to convert the audio to be inspected into text (i.e., the first text mentioned above). The speech recognition software can be Google Voice Assistant, Microsoft's speech recognition software, Baidu speech recognition, etc. The audio to be inspected is imported into the speech recognition software. The speech recognition software will automatically analyze the audio file and convert it into text. Then, keywords are extracted from the text to remove invalid words in the text to obtain the first key text.

[0051] Similarly, when extracting the second key text of the customer service audio, first use automatic speech recognition technology to perform text recognition on the customer service audio. You can use speech recognition software to convert the customer service audio into text (i.e., the second text mentioned above). The speech recognition software can be Google Voice Assistant, Microsoft's speech recognition software, Baidu speech recognition, etc. Import the customer service audio into the speech recognition software. The speech recognition software will automatically analyze the audio file and convert it into text; then perform keyword extraction on the text, remove invalid words in the text, and obtain the second key text.

[0052] S30: Determine whether there are preset prohibited words in the second key text. If the preset prohibited words exist in the second key text, determine the final quality inspection result of the audio to be quality inspected.

[0053] Then determine whether there are preset banned words in the second key text. If there are preset banned words in the second key text, it means that the intelligent assistant is seriously unqualified in serving customers. The final quality inspection result of the voice call quality can be directly determined to be the lowest. There is no need for subsequent emotional quality inspection, business process quality inspection, and key points quality inspection of speech, etc., which can improve the quality inspection efficiency of voice call quality.

[0054] When determining whether there are preset prohibited words in the second key text, a method of matching words with text can be used, such as string matching, regular expression matching, stem extraction and word form restoration matching, semantic matching and fuzzy matching, etc. In the embodiment of the present invention, string matching is taken as an example, and the second key text content is traversed to check whether there is a string that is exactly the same as the preset prohibited words character by character. If found, it is considered that the match is successful. This method is simple and direct, and has high matching efficiency.

[0055] Among them, the preset prohibited words include curse words, dirty words, etc. There are generally multiple preset prohibited words, which can be determined according to actual conditions, and the embodiment of the present invention does not make specific limitations on this.

[0056] It should be noted that the quality inspection results in the embodiment of the present invention are divided into four levels: unqualified, good, qualified and excellent, unqualified being the worst, good being the second, qualified indicating that the voice call quality meets the requirements, and excellent indicating that the voice call quality is very excellent. When it is determined that the second key text contains the preset prohibited words, the final quality inspection result of the audio to be inspected is directly determined to be unqualified.

[0057] In the embodiment of the present invention, by pre-setting the judgment of banned words, the final quality inspection results of a part of voice calls with obviously poor service quality can be determined first, without the need for subsequent emotion quality inspection, business process quality inspection and speech key points quality inspection, thereby improving the efficiency of voice call quality inspection and enabling batch voice call quality inspection.

[0058] S40: If the preset banned words do not exist in the second key text, the customer service audio and the first key text are input into a trained emotion recognition model to obtain an emotion quality inspection result, wherein the emotion recognition model is trained by samples and labels.

[0059] If there are no preset banned words in the second key text, the audio to be inspected will continue to be inspected for emotions, business processes, and key points of speech. In the embodiment of the present invention, the customer service audio and the first key text are input into a trained emotion recognition model to obtain an emotion inspection result. The customer service audio is used as a voice feature, and the first key text is used as a text feature. After multimodal fusion of the voice features and the text features, they are input into the emotion recognition model, so that the obtained emotion inspection result is more accurate.

[0060] As an implementation manner, a flow chart of step S40 in an embodiment of the present invention is as follows: Figure 4 As shown, the steps of inputting the semantic information and the first key text into a trained emotion recognition model to obtain an emotion quality inspection result include:

[0061] S41: Extracting the frequency spectrum features of the customer service audio and the word vector features of the first key text;

[0062] S42: concatenating the spectral feature with the word vector to obtain a concatenated feature;

[0063] S43: Input the splicing features into the trained emotion recognition model to obtain the emotion quality inspection result.

[0064] First, extract the spectral features of the customer service audio and the word vector features of the first key text, wherein the spectral features of the customer service audio can be Mel frequency spectrum (MFCC) features. MFCC features are an audio feature extraction method based on the auditory characteristics of the human ear. It simulates the sensitivity of the human ear to different frequencies and converts the audio signal into a set of feature parameters related to human auditory perception.

[0065] The steps of extracting Mel spectrum features are as follows: 1. Pre-emphasis: pre-emphasize the audio signal through a high-pass filter to enhance the energy of the high-frequency part and compensate for the decreased sensitivity of the human ear to high-frequency signals; 2. Framing: divide the audio signal into several short frames, usually with a length of 20-40 milliseconds per frame;

[0066] 3. Windowing: perform windowing on each frame. Commonly used window functions include Hamming window and Hanning window. Windowing can reduce discontinuity at the frame boundary and improve the accuracy of spectrum analysis. 4. Fast Fourier transform: perform FFT calculation on each frame to obtain the spectrum. 5. Mel filter bank: pass the spectrum through a group of Mel filters to obtain the Mel spectrum. The design of the Mel filter bank is based on the sensitivity of the human ear to different frequencies. The frequency range is usually divided into several Mel frequency bands. 6. Logarithmic operation: perform logarithmic operation on the Mel spectrum to simulate the logarithmic perception characteristics of the human ear to sound intensity. 7. Discrete cosine transform: perform discrete cosine transform on the logarithmic Mel spectrum to obtain the MFCC coefficient. The MFCC coefficient usually takes the first 12-20 coefficients as the characteristic parameters of the audio.

[0067] The word vector of the first key text may be extracted by using Word2Vec. Specifically, the word vector of the first key text may be obtained by inputting the first key text into Word2Vec.

[0068] As an example, the emotion recognition model is an LSTM model.

[0069] In the embodiment of the present invention, the emotion recognition model is an LSTM (Long Short-Term Memory) model. LSTM is a special recurrent neural network (RNN) that can effectively process long-term dependencies in sequence data.

[0070] The LSTM model initialization parameters are as follows:

[0071] Input layer: The concatenated features of the spectrum features and word vector features are used as the input of the LSTM model.

[0072] LSTM layer: Set the number of LSTM units, initially 128.

[0073] Activation function: Sigmoid is used as the activation function inside the LSTM unit.

[0074] Fully connected layer: converts the output of the LSTM layer into the final prediction category, which is divided into four categories: unqualified, good, qualified, and excellent.

[0075] Output layer: Use the Softmax activation function to output the probability of each category and finally get the sentiment quality inspection result.

[0076] After the LSTM model is initialized, the sample voice calls and labels are used to train the LSTM model to obtain the trained LSTM model. The customer service audio and the second key text are input into the trained LSTM model to obtain the sentiment quality inspection results.

[0077] S50: Performing business process quality inspection on the second key text according to preset business rules to obtain business quality inspection results.

[0078] After obtaining the sentiment quality inspection results, the second key text is subjected to business process quality inspection according to the preset business rules. For large-scale text data and repetitive quality inspection tasks, automated quality inspection methods can be used. Specifically, natural language processing technology and machine learning algorithms are used to develop automated quality inspection tools and systems. Automated quality inspection can improve quality inspection efficiency and accuracy, but it requires certain technical investment and data training.

[0079] As an implementation manner, a flowchart of step S50 provided in an embodiment of the present invention is as follows: Figure 5 As shown, the steps of performing business process quality inspection on the second key text according to the preset business rules and obtaining the business quality inspection result include:

[0080] S51: Segment processing is performed on the second key text;

[0081] S52: Performing business process quality inspection on the second key text after sentence segmentation using the preset business rules to obtain the business quality inspection result.

[0082] In the embodiment of the present invention, the second key text is firstly segmented into multiple separate sentences, wherein the preset business rules include key words of each business process node, and then semantic understanding is performed on each separate sentence and key words. If the semantics of the two are the same, the business process is considered to be correct and the business quality inspection result is passed; otherwise, the business process is considered to be wrong and the business quality inspection result is failed.

[0083] S60: According to the preset speech rules, the second key text is quality-checked on the key points of the speech to obtain the speech quality-check result.

[0084] According to the preset business rules, the second key text is quality-checked for the key points of the speech. For large-scale text data and repetitive quality inspection tasks, automated quality inspection methods can be used. Specifically, natural language processing technology and machine learning algorithms are used to develop automated quality inspection tools and systems. Automated quality inspection can improve quality inspection efficiency and accuracy, but it requires certain technical investment and data training.

[0085] As an implementation manner, a flowchart of step S60 provided in an embodiment of the present invention is as follows: Figure 6 As shown, the steps of performing a speech key point quality check on the second key text according to the preset speech rule and obtaining the speech quality check result include:

[0086] S61: Segment processing is performed on the second key text;

[0087] S62: Perform a speech quality inspection operation on the second key text after sentence segmentation using the preset speech rules to obtain the speech quality inspection result.

[0088] In the embodiment of the present invention, the second key text is firstly segmented into multiple separate sentences, wherein the preset speech rules include key words of each speech point, and then semantic understanding is performed on each separate sentence and the key words of the speech point. If the semantics of the two are the same, it is considered that the speech point is correct and the speech quality inspection result is passed; otherwise, it is considered that the speech point is wrong and the speech quality inspection result is failed.

[0089] S70: Determine the final quality inspection result of the audio to be quality inspected based on the emotion quality inspection result, the business quality inspection result and the speech quality inspection result.

[0090] The embodiment of the present invention comprehensively evaluates the audio to be inspected based on the three dimensions of the above-mentioned emotion quality inspection results, business quality inspection results and speech quality inspection results, thereby obtaining the final quality inspection results. The embodiment of the present invention combines emotion quality inspection, business process quality inspection and speech key point quality inspection to judge the quality inspection results of voice calls from multiple aspects, which can further improve the accuracy of the voice call quality inspection results.

[0091] As an example, the steps of determining the final quality inspection result of the audio to be quality inspected according to the emotion quality inspection result, the business quality inspection result and the speech quality inspection result include:

[0092] Based on preset weights, the emotion quality inspection result, the business quality inspection result and the speech quality inspection result are weighted and summed to determine the final quality inspection result of the audio to be quality inspected.

[0093] In the embodiment of the present invention, the emotional quality inspection results, business quality inspection results, and speech quality inspection results are all pre-set with weights. Different quality inspection results correspond to different weights, which can be pre-set and can be determined according to actual conditions. The embodiment of the present invention does not specifically limit this. According to the preset weight corresponding to the emotional quality inspection result, the preset weight corresponding to the business quality inspection result, and the preset weight corresponding to the speech quality inspection result, a weighted sum is performed to obtain the final quality inspection result of the audio to be inspected.

[0094] It can be seen that in the above scheme, for complex insurance entities such as insurance business, in the scheme implemented by the voice call quality inspection method based on artificial intelligence, after receiving the quality inspection request, the audio to be inspected is extracted, and the audio to be inspected is processed by splitting the tracks to obtain the customer service audio; then the first key text of the audio to be inspected and the second key text of the customer service audio are extracted; it is determined whether the second key text contains preset banned words. If there are preset banned words, it means that the service quality of the customer service is poor, and the final quality inspection result of the customer service can be directly determined. In the embodiment of the present invention, by judging the preset banned words, the final quality inspection results of a part of voice calls with obviously poor service quality can be determined first, without the need for subsequent emotional quality inspection, business process quality inspection and speech key points quality inspection, thereby improving the efficiency of voice call quality inspection and enabling batch voice call quality inspection.

[0095] If there are no preset banned words, the customer service quality needs to be further evaluated. First, the customer service audio and the first key text are combined with the emotion recognition model to obtain the emotion quality inspection result; then the business quality inspection result and the speech quality inspection result are combined to determine the final quality inspection result. In the embodiment of the present invention, multimodal fusion of voice and text is performed through customer service audio and the first key text, so that the emotion recognition result is more accurate and the accuracy of the voice call quality inspection result is improved; in addition, the embodiment of the present invention combines emotion quality inspection, business process quality inspection and speech key point quality inspection to judge the quality inspection results of voice calls from multiple aspects, which can further improve the accuracy of the voice call quality inspection results.

[0096] 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.

[0097] In one embodiment, a voice call quality inspection device based on artificial intelligence is provided, and the voice call quality inspection device based on artificial intelligence corresponds to the voice call quality inspection method based on artificial intelligence in the above embodiment. Figure 7 As shown, the intelligent question-answering processing device includes an audio extraction module 701, a key text module 702, a prohibited judgment module 703, an emotion quality inspection module 704, a business quality inspection module 705, a speech quality inspection module 706 and a result determination module 707. The functional modules are described in detail as follows:

[0098] The audio extraction module 701 is used to obtain the audio to be inspected according to the quality inspection request, and perform track processing on the audio to be inspected to obtain the customer service audio;

[0099] A key text module 702, used to obtain a first key text corresponding to the audio to be inspected and a second key text corresponding to the customer service audio;

[0100] The prohibited word determination module 703 is used to determine whether there is a preset prohibited word in the second key text, and if the preset prohibited word exists in the second key text, determine the final quality inspection result of the audio to be inspected;

[0101] The emotion quality inspection module 704 is used to input the customer service audio and the first key text into a trained emotion recognition model to obtain an emotion quality inspection result if the preset prohibited words do not exist in the second key text, wherein the emotion recognition model is trained by samples and labels;

[0102] A business quality inspection module 705 is used to perform a business process quality inspection on the second key text according to preset business rules to obtain a business quality inspection result;

[0103] A speech quality inspection module 706 is used to perform a speech key point quality inspection on the second key text according to a preset speech rule to obtain a speech quality inspection result;

[0104] The result determination module 707 is used to determine the final quality inspection result of the audio to be quality inspected based on the emotion quality inspection result, the business quality inspection result and the speech quality inspection result.

[0105] In one embodiment, the sentiment quality inspection module 704 is specifically used to:

[0106] Extracting frequency spectrum features of the customer service audio and word vector features of the first key text;

[0107] Concatenating the spectral feature with the word vector to obtain a concatenated feature;

[0108] The splicing features are input into a trained emotion recognition model to obtain the emotion quality inspection result.

[0109] In one embodiment, the service quality inspection module 705 is specifically used to:

[0110] Performing sentence processing on the second key text;

[0111] The preset business rules are used to perform business process quality inspection on the second key text after the sentence processing to obtain the business quality inspection result.

[0112] In one embodiment, the speech quality inspection module 706 is further used to:

[0113] Performing sentence processing on the second key text;

[0114] The preset speech rules are used to perform speech quality inspection on the second key text after sentence segmentation to obtain the speech quality inspection result.

[0115] In one embodiment, the result determination module 707 is specifically configured to:

[0116] Based on preset weights, the emotion quality inspection result, the business quality inspection result and the speech quality inspection result are weighted and summed to determine the final quality inspection result of the audio to be quality inspected.

[0117] In one embodiment, the key text module 702 is specifically used to:

[0118] Using automatic speech recognition technology to respectively recognize the texts corresponding to the audio to be inspected and the customer service audio, to obtain the text corresponding to the audio to be inspected and the text corresponding to the customer service audio;

[0119] Keyword extraction is performed on the text corresponding to the audio to be quality inspected and the text corresponding to the customer service audio to obtain the first key text and the second key text.

[0120] The present invention provides a voice call quality inspection device. In the embodiment of the present invention, by pre-setting the judgment of banned words, the final quality inspection results of a part of voice calls with obviously poor service quality can be determined first, without the need for subsequent emotion quality inspection, business process quality inspection and speech key point quality inspection, thereby improving the efficiency of voice call quality inspection and enabling batch voice call quality inspection; in the embodiment of the present invention, multimodal fusion of voice text is performed through customer service audio and a first key text, thereby making the result of emotion recognition more accurate and improving the accuracy of the voice call quality inspection result; in addition, the embodiment of the present invention combines emotion quality inspection, business process quality inspection and speech key point quality inspection to judge the quality inspection result of voice calls from multiple aspects, thereby further improving the accuracy of the voice call quality inspection result.

[0121] For the specific definition of the voice call quality inspection device, please refer to the definition of the intelligent question and answer method above, which will not be repeated here. Each module in the above-mentioned voice call quality inspection device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0122] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 8 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client through a network connection. When the computer program is executed by the processor, it realizes the functions or steps of a service-side of a voice call quality inspection method based on artificial intelligence.

[0123] In one embodiment, a computer device is provided. The computer device may be a client, and its internal structure diagram may be as follows: Fig. 9As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps on the client side of a voice call quality inspection method based on artificial intelligence

[0124] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0125] Acquire the audio to be inspected according to the quality inspection request, and perform track splitting on the audio to be inspected to obtain customer service audio;

[0126] Obtain the first key text corresponding to the audio to be inspected and the second key text corresponding to the customer service audio;

[0127] Determine whether there is a preset prohibited word in the second key text, and if the preset prohibited word exists in the second key text, determine the final quality inspection result of the audio to be inspected;

[0128] If the preset prohibited words do not exist in the second key text, input the customer service audio and the first key text into a trained emotion recognition model to obtain an emotion quality inspection result, wherein the emotion recognition model is trained by samples and labels;

[0129] According to preset business rules, a business process quality inspection is performed on the second key text to obtain a business quality inspection result;

[0130] According to the preset speech rules, the second key text is quality checked on the key points of the speech to obtain the speech quality check result;

[0131] The final quality inspection result of the audio to be quality inspected is determined based on the emotion quality inspection result, the business quality inspection result and the speech quality inspection result.

[0132] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0133] Acquire the audio to be inspected according to the quality inspection request, and perform track splitting on the audio to be inspected to obtain customer service audio;

[0134] Obtain the first key text corresponding to the audio to be inspected and the second key text corresponding to the customer service audio;

[0135] Determine whether there is a preset prohibited word in the second key text, and if the preset prohibited word exists in the second key text, determine the final quality inspection result of the audio to be inspected;

[0136] If the preset prohibited words do not exist in the second key text, input the customer service audio and the first key text into a trained emotion recognition model to obtain an emotion quality inspection result, wherein the emotion recognition model is trained by samples and labels;

[0137] According to preset business rules, a business process quality inspection is performed on the second key text to obtain a business quality inspection result;

[0138] According to the preset speech rules, the second key text is quality checked on the key points of the speech to obtain the speech quality check result;

[0139] The final quality inspection result of the audio to be quality inspected is determined based on the emotion quality inspection result, the business quality inspection result and the speech quality inspection result.

[0140] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can refer to the relevant descriptions on the server side and the client side in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.

[0141] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0142] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0143] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A voice call quality inspection method based on artificial intelligence, characterized in that: include: Acquire the audio to be inspected according to the quality inspection request, and perform track splitting on the audio to be inspected to obtain customer service audio; Obtain the first key text corresponding to the audio to be inspected and the second key text corresponding to the customer service audio; Determine whether there is a preset prohibited word in the second key text, and if the preset prohibited word exists in the second key text, determine the final quality inspection result of the audio to be inspected; If the preset prohibited words do not exist in the second key text, input the customer service audio and the first key text into a trained emotion recognition model to obtain an emotion quality inspection result, wherein the emotion recognition model is trained by samples and labels; According to preset business rules, a business process quality inspection is performed on the second key text to obtain a business quality inspection result; According to the preset speech rules, the second key text is quality checked on the key points of the speech to obtain the speech quality check result; The final quality inspection result of the audio to be quality inspected is determined based on the emotion quality inspection result, the business quality inspection result and the speech quality inspection result.

2. The method for voice call quality inspection based on artificial intelligence according to claim 1, characterized in that: The step of inputting the semantic information and the first key text into a trained emotion recognition model to obtain an emotion quality inspection result comprises: Extracting frequency spectrum features of the customer service audio and word vector features of the first key text; Concatenating the spectral feature with the word vector to obtain a concatenated feature; The splicing features are input into a trained emotion recognition model to obtain the emotion quality inspection result.

3. The method for voice call quality inspection based on artificial intelligence according to claim 2, characterized in that: The emotion recognition model is an LSTM model.

4. The method for voice call quality inspection based on artificial intelligence according to claim 1, characterized in that: The step of performing a business process quality inspection on the second key text according to the preset business rules to obtain a business quality inspection result includes: Performing sentence processing on the second key text; The preset business rules are used to perform business process quality inspection on the second key text after the sentence processing to obtain the business quality inspection result.

5. The method for voice call quality inspection based on artificial intelligence according to claim 1, characterized in that: The step of performing a speech key point quality check on the second key text according to a preset speech rule to obtain a speech quality check result includes: Performing sentence processing on the second key text; The preset speech rules are used to perform speech quality inspection on the second key text after sentence segmentation to obtain the speech quality inspection result.

6. The method for voice call quality inspection based on artificial intelligence according to claim 1, characterized in that: The step of determining the final quality inspection result of the audio to be quality inspected according to the emotion quality inspection result, the business quality inspection result and the speech quality inspection result comprises: Based on preset weights, the emotion quality inspection result, the business quality inspection result and the speech quality inspection result are weighted and summed to determine the final quality inspection result of the audio to be quality inspected.

7. The method for voice call quality inspection based on artificial intelligence according to any one of claims 1 to 5, characterized in that: The step of extracting the first key text corresponding to the audio to be inspected and the second key text corresponding to the customer service audio includes: Using automatic speech recognition technology to respectively recognize the texts corresponding to the audio to be inspected and the customer service audio, to obtain the text corresponding to the audio to be inspected and the text corresponding to the customer service audio; Keyword extraction is performed on the text corresponding to the audio to be quality inspected and the text corresponding to the customer service audio to obtain the first key text and the second key text.

8. A voice call quality inspection device based on artificial intelligence, characterized in that: include: An audio extraction module is used to obtain the audio to be inspected according to the quality inspection request, and perform track processing on the audio to be inspected to obtain customer service audio; A key text module, used to obtain a first key text corresponding to the audio to be quality inspected and a second key text corresponding to the customer service audio; a prohibited word determination module, configured to determine whether there is a preset prohibited word in the second key text, and if the preset prohibited word exists in the second key text, determine a final quality inspection result of the audio to be inspected; An emotion quality inspection module, configured to input the customer service audio and the first key text into a trained emotion recognition model to obtain an emotion quality inspection result if the preset prohibited words do not exist in the second key text, wherein the emotion recognition model is trained by samples and labels; A business quality inspection module, used to perform business process quality inspection on the second key text according to preset business rules and obtain business quality inspection results; A speech quality inspection module, used to perform speech key point quality inspection on the second key text according to preset speech rules, and obtain speech quality inspection results; A result determination module is used to determine the final quality inspection result of the audio to be quality inspected based on the emotion quality inspection result, the business quality inspection result and the speech quality inspection result.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the artificial intelligence-based voice call quality inspection method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, the steps of the artificial intelligence-based voice call quality inspection method as described in any one of claims 1 to 7 are implemented.

Images