Call quality inspection method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a call quality inspection method and device, electronic equipment and a storage medium, belongs to the technical field of artificial intelligence, and is suitable for the field of financial science and technology. The method comprises the steps that sample call data are acquired, the sample call data comprise sample call voices and sample quality inspection labels of the sample call voices, and the sample quality inspection labels are used for representing that the types of the sample call voices are compliant call types or non-compliant call types; performing text conversion on the sample call voice to obtain a sample call text; performing quality inspection category prediction on the sample call text to obtain sample category prediction probability data; performing text conversion on the target call voice to obtain a target call text; and performing quality inspection on the target call text based on the sample category prediction probability data and the sample quality inspection label to obtain a target quality inspection category, the target quality inspection category being used for representing that the category of the target call voice is a compliant call category or a non-compliant call category. According to the embodiment of the invention, the accuracy of call quality inspection can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, is applicable to the field of financial technology, and in particular to a call quality inspection method and device, an electronic device, and a storage medium. Background Art

[0002] Call quality inspection is an artificial intelligence technology that is used to perform quality checks on call recordings. It can evaluate the service quality provided by customer service personnel during telephone communications and can be applied to multiple scenarios, such as financial scenarios.

[0003] At present, neural network models are mainly used to predict the quality of call content, and the category of the call content is determined based on a preset threshold. However, in actual application scenarios, the prediction values ​​of different categories may be relatively similar, and cannot be simply distinguished by a threshold, which is prone to deviations, thus affecting the accuracy of call quality inspection.

[0004] Therefore, how to improve the accuracy of call quality inspection has become a technical problem that needs to be solved urgently. Summary of the invention

[0005] The main purpose of the embodiments of the present application is to provide a call quality inspection method and device, an electronic device and a storage medium, aiming to improve the accuracy of call quality inspection.

[0006] To achieve the above object, a first aspect of an embodiment of the present application provides a call quality inspection method, the method comprising:

[0007] Acquire sample call data; wherein the sample call data includes sample call voice and a sample quality inspection label of the sample call voice, wherein the sample quality inspection label is used to characterize whether the sample call voice is a compliant call category or a non-compliant call category;

[0008] Convert the sample call speech into text to obtain the sample call text;

[0009] Performing quality inspection category prediction on the sample call text to obtain sample category prediction probability data;

[0010] Convert the pre-acquired target call speech into text to obtain the target call text;

[0011] The target call text is quality-checked based on the sample category prediction probability data and the sample quality inspection label to obtain a target quality inspection category; wherein the target quality inspection category is used to characterize the category of the target call voice as the compliant call category or the non-compliant call category.

[0012] In some embodiments, the step of performing quality inspection on the target call text based on the sample category prediction probability data and the sample quality inspection label to obtain a target quality inspection category includes:

[0013] Performing quality inspection category prediction on the target call text to obtain target category prediction probability data;

[0014] Screening the sample category prediction probability data based on the sample quality inspection label to obtain reference category prediction probability data;

[0015] Clustering is performed based on the target category prediction probability data and the reference category prediction probability data to obtain the target quality inspection category.

[0016] In some embodiments, the reference category prediction probability data includes the same number of compliant call prediction probability data and non-compliant call prediction probability data, the sample quality inspection label of the compliant call prediction probability data represents the category of the compliant call prediction probability data as the compliant call category, and the sample quality inspection label of the non-compliant call prediction probability data represents the category of the non-compliant call prediction probability data as the non-compliant call category; the clustering process based on the target category prediction probability data and the reference category prediction probability data to obtain the target quality inspection category includes:

[0017] Clustering the compliant call prediction probability data, the non-compliant call prediction probability data, and the target category prediction probability data to obtain target clustering data;

[0018] Class voting is performed on the target clustering data to obtain the target quality inspection category.

[0019] In some embodiments, clustering the compliant call prediction probability data, the non-compliant call prediction probability data, and the target category prediction probability data to obtain target clustering data includes:

[0020] Calculating the distance between the target category prediction probability data and each of the compliant call prediction probability data to obtain compliant distance data;

[0021] Calculating the distance between the target category prediction probability data and each of the non-compliant call prediction probability data to obtain non-compliant distance data;

[0022] Sorting the plurality of compliant distance data and the plurality of non-compliant distance data to obtain an original distance sequence;

[0023] The original distance sequence is screened to obtain the target clustering data.

[0024] In some embodiments, the target clustering data includes a plurality of target prediction probability data, and performing category voting on the target clustering data to obtain the target quality inspection category includes:

[0025] Perform compliance category statistics on the sample quality inspection labels of the target prediction probability data to obtain the number of compliance categories;

[0026] Performing statistics on non-compliant categories on the sample quality inspection labels of the target prediction probability data to obtain the number of non-compliant categories;

[0027] The category with the largest number is screened out from the number of compliant categories and the number of non-compliant categories to obtain the target quality inspection category.

[0028] In some embodiments, performing quality inspection category prediction on the sample call text to obtain sample category prediction probability data includes:

[0029] Structuring the sample call text based on a preset question-and-answer template to obtain a sample question-and-answer text;

[0030] Perform category prediction on the sample question and answer text to obtain the sample category prediction probability data.

[0031] In some embodiments, converting the sample call speech into text to obtain the sample call text includes:

[0032] Performing speech recognition on the sample call speech to obtain an original sample text;

[0033] Performing text cleaning on the original sample text to obtain an initial sample text;

[0034] The initial sample text is subjected to text standardization processing to obtain the sample call text.

[0035] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application provides a call quality inspection device, the device comprising:

[0036] A sample data acquisition module, used to acquire sample call data; wherein the sample call data includes sample call voice and sample quality inspection label, wherein the sample quality inspection label is used to characterize the category of the sample call voice as a compliant call category or a non-compliant call category;

[0037] A sample text conversion module is used to convert the sample call speech into text to obtain the sample call text;

[0038] A sample quality inspection category prediction module is used to perform quality inspection category prediction on the sample call text to obtain sample category prediction probability data;

[0039] A target text conversion module is used to convert the pre-acquired target call speech into text to obtain a target call text;

[0040] A target quality inspection module is used to perform quality inspection on the target call text based on the sample category prediction probability data and the sample quality inspection label to obtain a target quality inspection category; wherein the target quality inspection category is used to characterize the category of the target call voice as the compliant call category or the non-compliant call category.

[0041] To achieve the above objectives, a third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method described in the first aspect is implemented.

[0042] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in the first aspect is implemented.

[0043] The call quality inspection method and device, electronic device and storage medium proposed in the present application obtain sample call data; wherein the sample call data includes sample call speech and sample quality inspection labels corresponding to the sample call speech, and the sample quality inspection labels are used to characterize the category of the sample call speech as a compliant call category or a non-compliant call category; the sample call speech is converted into text to obtain a sample call text, and the sample call text is predicted to be a quality inspection category to obtain sample category prediction probability data, thereby learning the corresponding features of the compliant call category and the non-compliant call category, and obtaining sample category prediction probability data; further, the pre-acquired target call speech is converted into text to obtain a target call text, and the target call text is quality inspected based on the sample category prediction probability data and the sample quality inspection labels to obtain a reliable target quality inspection category, which can effectively eliminate the category prediction bias caused by the sample data and improve the accuracy of call quality inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is a flow chart of a call quality inspection method provided in an embodiment of the present application;

[0045] Figure 2 yes Figure 1 Flow chart of step S102 in FIG.

[0046] Figure 3 yes Figure 1 Flow chart of step S103 in FIG.

[0047] Figure 4 yes Figure 1 Flow chart of step S105 in FIG.

[0048] Figure 5 yes Figure 4 Flow chart of step S403 in FIG.

[0049] Figure 6 yes Figure 5 Flow chart of step S501 in FIG.

[0050] Figure 7 yes Figure 5 Flow chart of step S502 in FIG.

[0051] Figure 8 It is a structural diagram of a call quality inspection device provided in an embodiment of the present application;

[0052] Fig. 9 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0054] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0055] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0056] First, some nouns involved in this application are analyzed:

[0057] Artificial intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new type of intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing and expert systems. AI can simulate the information process of human consciousness and thinking. AI is also a 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.

[0058] Natural language processing (NLP): NLP uses computers to process, understand and apply human languages ​​(such as Chinese, English, etc.). NLP is a branch of artificial intelligence and an interdisciplinary subject between computer science and linguistics. It is often referred to as computational linguistics. Natural language processing includes grammatical analysis, semantic analysis, and text understanding. Natural language processing is often used in technical fields such as machine translation, handwritten and printed character recognition, speech recognition and text-to-speech conversion, information intent recognition, information extraction and filtering, text classification and clustering, public opinion analysis and opinion mining. It involves data mining related to language processing, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research, and linguistic research related to language computing.

[0059] Call quality inspection: Call quality inspection is an artificial intelligence technology used to conduct quality inspections on call recordings. It can evaluate the service quality of customer service personnel during telephone communication. The call quality inspection process is mainly achieved by analyzing and evaluating call recordings. It aims to discover problems in the service and find ways to improve it, so as to improve service levels and customer satisfaction. Call quality inspection covers multiple aspects such as customer service personnel's language expression, polite language, service attitude, problem-solving ability, professional ability, response speed, error handling, and communication skills. By establishing and evaluating quality inspection standards, customer service personnel can improve their service levels and enhance their professional capabilities, thereby providing customers with better service guarantees.

[0060] Call quality inspection can be applied to multiple scenarios. For example, in the field of financial technology, call quality inspection can help financial institutions monitor whether customer service representatives comply with relevant laws and policies when communicating with customers, and promptly discover and correct violations. In addition, by analyzing the language expression, attitude, and problem-solving ability of customer service representatives, financial institutions can evaluate service quality, discover deficiencies in services, and conduct targeted training and improvements, thereby providing customers with better, compliant, and safer services.

[0061] At present, neural network models are mainly used to conduct quality inspection and prediction of call content, and determine the category corresponding to the call content based on preset thresholds.

[0062] However, in actual application scenarios, the predicted values ​​of different categories may be relatively close, and the classification based on the predicted values ​​is often affected by the inherent bias during training. Bias refers to the model's oversensitivity or neglect of certain information in the absence of representative or balanced data, which may lead to performance degradation in actual applications. These deviations may come from the imbalance of training data or language characteristics in specific fields, such as frequent word bias and field label bias. Specifically, the output probability value of a specific category is higher, causing the model's classification to tend to a specific category. Therefore, it is not possible to simply use a threshold for quality inspection classification, which affects the accuracy of call quality inspection.

[0063] Based on this, the embodiments of the present application provide a call quality inspection method and device, an electronic device and a storage medium, aiming to improve the accuracy of call quality inspection.

[0064] The call quality inspection method and device, electronic device and storage medium provided in the embodiments of the present application are specifically described through the following embodiments. First, the call quality inspection method in the embodiments of the present application is described.

[0065] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, 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.

[0066] 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.

[0067] The call quality inspection method provided in the embodiment of the present application relates to the field of artificial intelligence technology and is applicable to the field of financial technology. The call quality inspection method provided in the embodiment of the present application can be applied to the terminal, can also be applied to the server side, and can also be software running in the terminal or the server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or it can be configured as a server cluster or distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the call quality inspection method, etc., but is not limited to the above forms.

[0068] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0069] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of these data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0070] Figure 1 is an optional flow chart of the call quality inspection method provided in the embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S105.

[0071] Step S101, obtaining sample call data; wherein the sample call data includes sample call voice and sample quality inspection labels of the sample call voice, and the sample quality inspection labels are used to characterize whether the sample call voice is a compliant call category or a non-compliant call category;

[0072] Step S102, converting the sample call speech into text to obtain a sample call text;

[0073] Step S103, performing quality inspection category prediction on the sample call text to obtain sample category prediction probability data;

[0074] Step S104, converting the pre-acquired target call speech into text to obtain a target call text;

[0075] Step S105, based on the sample category prediction probability data and the sample quality inspection label, the target call text is quality inspected to obtain the target quality inspection category; wherein the target quality inspection category is used to characterize the category of the target call voice as a compliant call category or a non-compliant call category.

[0076] Steps S101 to S105 shown in the embodiment of the present application are performed by acquiring sample call data; wherein the sample call data includes sample call speech and a sample quality inspection label corresponding to the sample call speech, and the sample quality inspection label is used to characterize the category of the sample call speech as a compliant call category or a non-compliant call category; the sample call speech is converted into text to obtain a sample call text, and the sample call text is quality inspected to obtain sample category prediction probability data, thereby learning the features corresponding to the compliant call category and the non-compliant call category, and obtaining sample category prediction probability data; further, the pre-acquired target call speech is converted into text to obtain a target call text, and the target call text is quality inspected based on the sample category prediction probability data and the sample quality inspection label to obtain a reliable target quality inspection category, which can effectively eliminate the category prediction bias caused by the sample data and improve the accuracy of call quality inspection.

[0077] In step S101 of some embodiments, sample call data in a financial application scenario is obtained, specifically including a plurality of sample call voices and a sample quality inspection label corresponding to each sample call voice data, wherein the sample quality inspection label is pre-marked by a technician and includes two categories, namely, a category characterizing the sample call voice as a compliant call category, or a category characterizing the sample call voice as a non-compliant call category;

[0078] It should be noted that the categories of sample call voices are limited according to the regulatory system of specific application scenarios, for example:

[0079] Financial institution employees or other relevant personnel spread false information or jointly manipulate market prices through phone calls in an attempt to affect the normal operation of the financial market. For example, an investor disclosed in a phone call that he would buy a large amount of a certain stock in an attempt to increase the stock price. This behavior constitutes market manipulation and falls into the category of non-compliant calls.

[0080] When selling financial products, financial institutions or their staff provide customers with false information or conceal risks through phone calls, inducing customers to make investment decisions that are not based on the actual situation. Such calls violate the principle of integrity in financial sales and harm the interests of customers. For example, a salesperson exaggerates the benefits of a financial product during a call, but conceals its potential high risks. This behavior falls into the category of non-compliant calls.

[0081] It should be noted that financial application scenarios may include, but are not limited to: outbound phone call scenarios, telephone follow-up scenarios, and other scenarios that require manual telephone services, but are not limited to these.

[0082] See also Figure 2 In some embodiments, step S102 may include but is not limited to steps S201 to S203:

[0083] Step S201, performing speech recognition on the sample call speech to obtain the original sample text;

[0084] Step S202, performing text cleaning on the original sample text to obtain an initial sample text;

[0085] Step S203, performing text standardization processing on the initial sample text to obtain a sample call text.

[0086] Steps S201 to S203 shown in the embodiment of the present application obtain original sample text by performing speech recognition on the sample call speech, and then perform text cleaning and text standardization on the original sample text to obtain sample call text, which can effectively improve the accuracy and consistency of the data and help to improve the reliability and accuracy of call quality inspection in subsequent processing.

[0087] In step S201 of some embodiments, the sample call speech can be recognized and converted into text through a pre-trained neural network model, Transformer model, Gaussian mixture model, etc. to obtain the original sample text. The specific selection needs to be made in combination with the actual application scenario, but is not limited to this.

[0088] In step S202 of some embodiments, text cleaning includes but is not limited to removing punctuation marks, special symbols, stop words, etc., and the specific selection needs to be made in combination with the actual application scenario, but is not limited to this.

[0089] In some embodiments, the original sample text may be cleaned by using regular expressions or NLP libraries (such as NLTK, spaCy).

[0090] In step S203 of some embodiments, the text standardization process includes but is not limited to:

[0091] Case normalization, converting uppercase characters in the text to lowercase characters;

[0092] Word segmentation, which divides a continuous text sequence into characters or words;

[0093] Lemmatization, converting words to their base or dictionary form;

[0094] Unify the formats of numbers, dates, and times;

[0095] It should be noted that by performing text cleaning and text standardization on the original sample text, the accuracy and consistency of the data can be effectively improved, which will help improve the processing speed of the call quality inspection model and thus improve the efficiency of quality inspection.

[0096] Before step S103 in some embodiments, the call quality inspection method further includes pre-training a call quality inspection model, which is used to predict the quality inspection category of the call text. Specifically, the call quality inspection model can be a GPT-2 model, a transformer model, an LSTM model, etc., but is not limited thereto, and needs to be trained in combination with actual reference scenarios.

[0097] See also Figure 3 In some embodiments, step S103 may include but is not limited to steps S301 to S302:

[0098] Step S301, performing structured processing on the sample call text based on a preset question and answer template to obtain a sample question and answer text;

[0099] Step S302: perform category prediction on the sample question and answer text to obtain sample category prediction probability data.

[0100] In steps S301 to S302 shown in the embodiment of the present application, by performing structured processing on the sample call text based on a preset question and answer template to obtain a sample question and answer text, unstructured text data can be efficiently converted into structured question and answer data; and category prediction is performed on the sample question and answer text to obtain sample category prediction probability data, thereby improving the efficiency and accuracy of information extraction and processing.

[0101] In step S301 of some embodiments, the preset question-and-answer template may be: please output the category prediction probability data "Y" corresponding to this paragraph of text "T"; wherein T corresponds to the "sample call text", and Y is the result required for category prediction, namely, the "sample category prediction probability data".

[0102] In step S302 of some embodiments, the sample question and answer text is quality checked by a call quality check model to output a category prediction value of the sample question and answer text, that is, sample category prediction probability data.

[0103] It should be noted that the sample category prediction probability data is a probability value that can characterize the quality inspection result corresponding to the sample call speech, and its value range depends on the output of the call quality inspection model, for example, it can be an interval of 0 to 1.

[0104] In some embodiments, after step S103, the call quality inspection method further includes but is not limited to the following steps:

[0105] A classification reference database is constructed based on sample category prediction probability data and sample quality inspection labels.

[0106] It should be noted that the classification reference database is used for category confirmation during subsequent call quality inspection, and there is no need to repeat steps S101 to S103, which helps to improve the efficiency and accuracy of call quality inspection.

[0107] In some embodiments, step S104 may include but is not limited to the following steps:

[0108] Performing speech recognition on the pre-acquired target call speech to obtain the original call text;

[0109] Perform text cleaning on the original call text to obtain the initial call text;

[0110] The initial call text is subjected to text normalization processing to obtain the target call text.

[0111] It should be noted that the specific implementation of the above step S104 is basically the same as the specific implementation of the above steps S201 to S203, and will not be repeated here.

[0112] See also Figure 4 In some embodiments, step S105 may include but is not limited to steps S401 to S403:

[0113] Step S401, performing quality inspection category prediction on the target call text to obtain target category prediction probability data;

[0114] Step S402, screening the sample category prediction probability data based on the sample quality inspection label to obtain reference category prediction probability data;

[0115] Step S403: clustering is performed based on the target category prediction probability data and the reference category prediction probability data to obtain the target quality inspection category.

[0116] In steps S401 to S403 shown in the embodiment of the present application, target category prediction probability data is obtained by performing quality inspection category prediction on the target call text; sample category prediction probability data is screened based on sample quality inspection labels to obtain reference category prediction probability data; clustering processing is performed based on the target category prediction probability data and the reference category prediction probability data, and reference data with prior knowledge can be used to perform category confirmation to obtain accurate target quality inspection categories, thereby improving the accuracy and stability of call quality inspection.

[0117] In some embodiments, step S401 may include but is not limited to the following steps:

[0118] Structural processing is performed on the target call text based on a preset question-and-answer template to obtain a target question-and-answer text;

[0119] The call quality inspection model is used to predict the category of the target question and answer text and obtain the target category prediction probability data.

[0120] It should be noted that the specific implementation of the above step S401 is basically the same as the specific implementation of the above steps S301 to S302, and will not be repeated here.

[0121] In step S402 of some embodiments, a preset reference quantity of data is screened out from the sample category prediction probability data based on the sample quality inspection label to obtain reference category prediction probability data; wherein the reference category prediction probability data includes the same quantity of compliant call prediction probability data and non-compliant call prediction probability data, the sample quality inspection label of the compliant call prediction probability data represents that the category of the compliant call prediction probability data is a compliant call category, and the sample quality inspection label of the non-compliant call prediction probability data represents that the category of the non-compliant call prediction probability data is a non-compliant call category;

[0122] It should be noted that the preset reference quantity is set based on the actual application scenario, and the preset reference quantity is used to limit the number of data screened from the sample category prediction probability data, that is, the number of reference category prediction probability data;

[0123] For example: the reference category prediction probability data includes prediction probability data of 4 sample quality inspection labels as compliant call categories, and prediction probability data of 4 sample quality inspection labels as non-compliant call categories, that is, the reference category prediction probability data includes 4 compliant call prediction probability data, and 4 non-compliant call prediction probability data.

[0124] It can be understood that the sample category prediction probability data is a probability value, and data with large probability value deviations (i.e., outliers) will not be screened as reference category prediction probability data, which helps to improve the accuracy of quality inspection category classification through clustering processing.

[0125] See also Figure 5 In some embodiments, step S403 may also include but is not limited to steps S501 to S502:

[0126] Step S501, clustering the compliant call prediction probability data, the non-compliant call prediction probability data and the target category prediction probability data to obtain target clustering data;

[0127] Step S502: class voting is performed on the target clustering data to obtain the target quality inspection class.

[0128] Steps S101 to S106 shown in the embodiment of the present application, by clustering the compliant call prediction probability data, the non-compliant call prediction probability data and the target category prediction probability data, obtain the target cluster data, and perform category voting on the target cluster data, can use cluster analysis to identify the intrinsic correlation between the data, and then enhance the robustness of the decision through the voting mechanism, so as to more accurately infer the quality inspection category corresponding to the target call voice, effectively reduce the category prediction deviation that may be caused by the sample data, and improve the accuracy and reliability of call prediction.

[0129] See also Figure 6 In some embodiments, step S501 includes but is not limited to steps S601 to S604:

[0130] Step S601, performing distance calculation between the target category prediction probability data and each compliant call prediction probability data to obtain compliant distance data;

[0131] Step S602, performing distance calculation between the target category prediction probability data and each non-compliant call prediction probability data to obtain non-compliant distance data;

[0132] Step S603, sorting the plurality of compliant distance data and the plurality of non-compliant distance data to obtain an original distance sequence;

[0133] Step S604: Screen the original distance sequence to obtain target clustering data.

[0134] Steps S601 to S604 shown in the embodiment of the present application, by respectively calculating the distance between the target category prediction probability data and each compliant call prediction probability data and each non-compliant call prediction probability data, and sorting and screening, can accurately identify the prediction probability data most relevant to the target category (i.e., the smallest distance), obtain target clustering data, and help improve the accuracy and efficiency of call quality inspection classification.

[0135] In some embodiments, the distance calculation may adopt a cosine distance calculation method, an Euclidean distance calculation method, a Manhattan distance calculation method, etc., and the specific selection needs to be made in combination with the actual application scenario, but is not limited thereto.

[0136] In step S603 of some embodiments, multiple compliant distance data and multiple non-compliant distance data are sorted according to numerical values ​​to obtain an original distance sequence; specifically, the sorting may be from large to small numerical values, or from small to large numerical values, but is not limited thereto.

[0137] In step S604 of some embodiments, target clustering data is obtained by screening from the original distance sequence based on a preset number of targets;

[0138] It should be noted that the preset target number is used to limit the number of data screened from the original distance sequence, that is, the number of target clustering data; the preset target number needs to be set according to the actual application scenario.

[0139] In some embodiments, the preset target number is an odd number, and the preset target number is smaller than the preset reference number, which helps to classify in subsequent category voting, and avoids the situation where the votes for two categories are the same, resulting in failure to classify.

[0140] In some embodiments, the target clustering data includes a plurality of target prediction probability data;

[0141] See also Figure 7 In some embodiments, step S502 may include but is not limited to steps S701 to S703:

[0142] Step S701, performing compliance category statistics on sample quality inspection labels of target prediction probability data to obtain the number of compliance categories;

[0143] Step S702, performing non-compliant category statistics on sample quality inspection labels of target prediction probability data to obtain the number of non-compliant categories;

[0144] Step S703, filtering out the category with the largest number from the number of compliant categories and the number of non-compliant categories to obtain the target quality inspection category.

[0145] In steps S701 to S703 shown in the embodiment of the present application, by counting the compliance and non-compliance categories of the target prediction probability sample quality inspection labels in the clustering data, and screening out the category with the largest number as the target quality inspection category, voting on the clustering results is achieved, which can more objectively reflect the dominant category in the clustering, thereby improving the accuracy and stability of the quality inspection category determination, and further improving the accuracy of call quality inspection.

[0146] It is understandable that since the target clustering data contains an odd number of target prediction probability data, when voting, there is always a largest number between the number of compliant categories and the number of non-compliant categories. Therefore, the category with the largest number is selected from the number of compliant categories and the number of non-compliant categories to obtain the target quality inspection category.

[0147] Specifically, if the number of compliant categories is greater than the number of non-compliant categories, the target quality inspection category represents the category of the target call voice as a compliant call category;

[0148] If the number of compliant categories is less than the number of non-compliant categories, the target quality inspection category represents the category of the target call voice as a non-compliant call category.

[0149] For example:

[0150] The reference category prediction probability data includes 4 compliance call prediction probability data and 4 non-compliance call prediction probability data.

[0151] Calculate the distance between the target category prediction probability data and each compliant call prediction probability data to obtain compliant distance data; calculate the distance between the target category prediction probability data and each non-compliant call prediction probability data to obtain non-compliant distance data; sort multiple compliant distance data and multiple non-compliant distance data to obtain an original distance sequence;

[0152] The first three data with the largest values ​​are selected from the original distance sequence to obtain the target cluster data; wherein the target cluster data includes three target prediction probability data;

[0153] Obtain the sample quality inspection labels corresponding to the three target prediction probability data, and perform statistics on the compliance categories and non-compliance categories. The number of compliance categories is 2, and the number of non-compliance categories is 1;

[0154] Therefore, the target quality inspection category corresponding to the target call voice represents that the category of the target call voice is a compliant call category.

[0155] In some embodiments, the preset reference number, the preset target number, and the calculation method adopted for distance calculation all need to be set in combination with the actual application scenario;

[0156] Therefore, based on the sample call data, the preset reference number, the preset target number and the calculation method for distance calculation can be adjusted by using a five-fold cross-validation method to achieve the optimal model configuration.

[0157] See also Figure 8 The embodiment of the present application further provides a call quality inspection device, which can implement the above-mentioned call quality inspection method, and the device includes:

[0158] The sample data acquisition module 801 is used to acquire sample call data; wherein the sample call data includes sample call voice and sample quality inspection label, and the sample quality inspection label is used to characterize the category of the sample call voice as a compliant call category or a non-compliant call category;

[0159] The sample text conversion module 802 is used to convert the sample call speech into text to obtain the sample call text;

[0160] The sample quality inspection category prediction module 803 is used to perform quality inspection category prediction on the sample call text to obtain sample category prediction probability data;

[0161] The target text conversion module 804 is used to convert the pre-acquired target call speech into text to obtain the target call text;

[0162] The target quality inspection module 805 is used to perform quality inspection on the target call text based on the sample category prediction probability data and the sample quality inspection label to obtain the target quality inspection category; wherein the target quality inspection category is used to characterize the category of the target call voice as a compliant call category or a non-compliant call category.

[0163] The specific implementation of the call quality inspection device is basically the same as the specific implementation of the above-mentioned call quality inspection method, and will not be repeated here.

[0164] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above call quality inspection method when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0165] See also Fig. 9 , Fig. 9 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0166] The processor 901 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0167] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 902, and the processor 901 calls and executes the call quality inspection method of the embodiment of this application;

[0168] Input / output interface 903, used to implement information input and output;

[0169] Communication interface 904, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0170] A bus 905 that transmits information between various components of the device (e.g., the processor 901, the memory 902, the input / output interface 903, and the communication interface 904);

[0171] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0172] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned call quality inspection method is implemented.

[0173] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0174] The call quality inspection method and device, electronic device and storage medium provided in the embodiments of the present application obtain sample call data; wherein the sample call data includes sample call speech and sample quality inspection labels, and the sample quality inspection labels are used to characterize the category of the sample call speech as a compliant call category or a non-compliant call category; the sample call speech is converted into text to obtain a sample call text, and the sample call text is predicted to be a quality inspection category to obtain sample category prediction probability data, thereby learning the features corresponding to the compliant call category and the non-compliant call category respectively, and obtaining sample category prediction probability data; further, the pre-acquired target call speech is converted into text to obtain a target call text, and the target call text is quality inspected based on the sample category prediction probability data and the sample quality inspection labels to obtain a reliable target quality inspection category, which can effectively eliminate the category prediction bias caused by the sample data and improve the accuracy of call quality inspection.

[0175] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0176] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0177] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0178] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0179] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0180] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0181] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0182] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0183] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0184] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0185] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A call quality inspection method, characterized in that: The method comprises: Acquire sample call data; wherein the sample call data includes sample call voice and a sample quality inspection label of the sample call voice, wherein the sample quality inspection label is used to characterize whether the sample call voice is a compliant call category or a non-compliant call category; Convert the sample call speech into text to obtain the sample call text; Performing quality inspection category prediction on the sample call text to obtain sample category prediction probability data; Convert the pre-acquired target call speech into text to obtain the target call text; The target call text is quality-checked based on the sample category prediction probability data and the sample quality inspection label to obtain a target quality inspection category; wherein the target quality inspection category is used to characterize the category of the target call voice as the compliant call category or the non-compliant call category.

2. The method according to claim 1, characterized in that The step of performing quality inspection on the target call text based on the sample category prediction probability data and the sample quality inspection label to obtain a target quality inspection category includes: Performing quality inspection category prediction on the target call text to obtain target category prediction probability data; Screening the sample category prediction probability data based on the sample quality inspection label to obtain reference category prediction probability data; Clustering is performed based on the target category prediction probability data and the reference category prediction probability data to obtain the target quality inspection category.

3. The method according to claim 2, characterized in that The reference category prediction probability data includes the same number of compliant call prediction probability data and non-compliant call prediction probability data, the sample quality inspection labels of the compliant call prediction probability data represent that the category of the compliant call prediction probability data is the compliant call category, and the sample quality inspection labels of the non-compliant call prediction probability data represent that the category of the non-compliant call prediction probability data is the non-compliant call category; The clustering process based on the target category prediction probability data and the reference category prediction probability data to obtain the target quality inspection category includes: Clustering the compliant call prediction probability data, the non-compliant call prediction probability data, and the target category prediction probability data to obtain target clustering data; Class voting is performed on the target clustering data to obtain the target quality inspection category.

4. The method according to claim 3, characterized in that: The clustering of the compliant call prediction probability data, the non-compliant call prediction probability data, and the target category prediction probability data to obtain target clustering data includes: Calculating the distance between the target category prediction probability data and each of the compliant call prediction probability data to obtain compliant distance data; Calculating the distance between the target category prediction probability data and each of the non-compliant call prediction probability data to obtain non-compliant distance data; Sorting the plurality of compliant distance data and the plurality of non-compliant distance data to obtain an original distance sequence; The original distance sequence is screened to obtain the target clustering data.

5. The method according to claim 3, characterized in that: The target clustering data includes a plurality of target prediction probability data, and the class voting for the target clustering data to obtain the target quality inspection class includes: Perform compliance category statistics on the sample quality inspection labels of the target prediction probability data to obtain the number of compliance categories; Performing statistics on non-compliant categories on the sample quality inspection labels of the target prediction probability data to obtain the number of non-compliant categories; The category with the largest number is screened out from the number of compliant categories and the number of non-compliant categories to obtain the target quality inspection category.

6. The method according to any one of claims 1 to 5, characterized in that: The performing quality inspection category prediction on the sample call text to obtain sample category prediction probability data includes: Structuring the sample call text based on a preset question-and-answer template to obtain a sample question-and-answer text; Perform category prediction on the sample question and answer text to obtain the sample category prediction probability data.

7. The method according to any one of claims 1 to 5, characterized in that: The step of converting the sample call speech into text to obtain the sample call text includes: Performing speech recognition on the sample call speech to obtain an original sample text; Performing text cleaning on the original sample text to obtain an initial sample text; The initial sample text is subjected to text standardization processing to obtain the sample call text.

8. A call quality inspection device, characterized in that: The device comprises: A sample data acquisition module, used to acquire sample call data; wherein the sample call data includes sample call voice and sample quality inspection label, wherein the sample quality inspection label is used to characterize the category of the sample call voice as a compliant call category or a non-compliant call category; A sample text conversion module is used to convert the sample call speech into text to obtain the sample call text; A sample quality inspection category prediction module is used to perform quality inspection category prediction on the sample call text to obtain sample category prediction probability data; A target text conversion module is used to convert the pre-acquired target call speech into text to obtain a target call text; A target quality inspection module is used to perform quality inspection on the target call text based on the sample category prediction probability data and the sample quality inspection label to obtain a target quality inspection category; wherein the target quality inspection category is used to characterize the category of the target call voice as the compliant call category or the non-compliant call category.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images