Performance performance quality inspection method and device for installation and maintenance personnel, and medium

Abstract

The invention provides a performance condition quality inspection method and device for installation and maintenance personnel and a medium. The method comprises the following steps: acquiring a corresponding quality inspection call text according to quality inspection call records of the installation and maintenance personnel and a user; using a pre-trained time information extraction model to extract the reservation door-to-door time of the installation and maintenance personnel and the time difference acceptable by the user from the quality inspection call text; and calculating the performance condition quality inspection score of the installation and maintenance personnel according to the reservation door-to-door time of the installation and maintenance personnel, the time difference which can be accepted by the user and the actual door-to-door time of the installation and maintenance personnel. According to the method, the device and the medium, the problems that quality inspection for the performance condition of the installation and maintenance personnel mostly depends on manual recording listening, the efficiency is low, the data analysis capability is insufficient, and a quality inspection result is easily inconsistent with an actual result in the prior art can be solved.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a method, device and medium for quality inspection of the performance of installation and maintenance personnel. Background Art

[0002] In the modern communications industry, the performance of installation and maintenance personnel directly impacts user satisfaction and the operator's reputation. However, existing technologies often neglect the monitoring and evaluation of installation and maintenance personnel's performance on time. While some companies have implemented quality inspection mechanisms, most focus solely on service quality and fail to effectively track personnel's timely performance.

[0003] With the development of technology, real-time monitoring and analysis of installation and maintenance personnel's call logs has become possible. By incorporating these logs into the quality inspection process, operators can more accurately assess the performance of installation and maintenance personnel. However, existing technologies often rely on manual listening to recordings, which is inefficient and lacks data analysis capabilities, easily leading to inconsistent quality inspection results with actual results. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to address the above-mentioned shortcomings of the existing technology and provide a method, device and medium for quality inspection of the performance of installation and maintenance personnel, so as to solve the problem that the quality inspection of the performance of installation and maintenance personnel in the existing technology mostly relies on manual listening to recordings, which is inefficient and has insufficient data analysis capabilities, which easily leads to inconsistency between quality inspection results and actual results.

[0005] In a first aspect, the present invention provides a method for quality inspection of the performance of installation and maintenance personnel, comprising:

[0006] Obtain the corresponding quality inspection call text based on the quality inspection call recording between the installation and maintenance personnel and the user;

[0007] Utilize a pre-trained time information extraction model to extract the scheduled visit time of the installation and maintenance personnel and the time difference acceptable to the user from the quality inspection call text;

[0008] The quality inspection score of the installation and maintenance personnel's performance is calculated based on the scheduled visit time of the installation and maintenance personnel, the time difference that the user can accept, and the actual visit time of the installation and maintenance personnel.

[0009] Furthermore, obtaining the corresponding quality inspection call text based on the quality inspection call recording between the installation and maintenance personnel and the user specifically includes:

[0010] Disassembling the recording of the quality inspection call between the installation and maintenance personnel and the user, and performing Fourier transform on the disassembled recording of the quality inspection call between the installation and maintenance personnel and the user to construct an image corresponding to the recording of the quality inspection call between the installation and maintenance personnel and the user;

[0011] Input the image corresponding to the quality inspection call recording between the installation and maintenance personnel and the user into a pre-trained deep fully-sequential convolutional neural network to output the quality inspection call text, where the pre-trained deep fully-sequential convolutional neural network is trained based on images corresponding to quality inspection call recordings of multiple historical installation and maintenance personnel and users and the corresponding quality inspection call texts.

[0012] Further, using the pre-trained time information extraction model to extract the scheduled on-site visit time of the installation and maintenance personnel and the acceptable time difference of the user from the quality inspection call text, specifically including:

[0013] Perform preprocessing and feature normalization on the quality inspection call text;

[0014] Input the preprocessed and feature-normalized quality inspection call text into the trained time information extraction model to output the entity label corresponding to each word in the quality inspection call text, where the trained time information extraction model is trained based on historically annotated quality inspection call texts;

[0015] According to the entity label corresponding to each word, extract time entities from the quality inspection call text to obtain the scheduled on-site visit time of the installation and maintenance personnel and the acceptable time difference of the user.

[0016] Further, the time information extraction model is an ERNIE model. The step of inputting the preprocessed and feature-normalized quality inspection call text into the trained time information extraction model to output the entity label corresponding to each word in the quality inspection call text specifically includes:

[0017] Input the preprocessed and feature-normalized quality inspection call text into the trained ERNIE model, predict the probability of each word in the quality inspection call text belonging to different entity categories through the trained ERNIE model, and determine the entity label corresponding to each word using maximum likelihood decoding according to the probability of each word belonging to different entity categories.

[0018] Further, calculating the quality inspection score of the fulfillment situation of the installation and maintenance personnel according to the scheduled on-site visit time of the installation and maintenance personnel, the acceptable time difference of the user, and the actual on-site visit time of the installation and maintenance personnel, specifically including:

[0019] Calculate the quality inspection score of the fulfillment situation of the installation and maintenance personnel through the following formula:

[0020]

[0021] Among them, defence is the quality inspection score of the performance of the installation and maintenance personnel, timeA is the scheduled arrival time of the installation and maintenance personnel, timeB is the actual arrival time of the installation and maintenance personnel, timeC is the acceptable time difference for the user, n is the service quality impact coefficient, k is the adjustment coefficient related to the user acceptance, m is the weight coefficient affecting the service attitude, p is the weight coefficient affected by the time difference, and ∈ is a small positive number.

[0022] Further, after calculating the quality inspection score of the performance of the installation and maintenance personnel according to the scheduled arrival time of the installation and maintenance personnel, the acceptable time difference for the user, and the actual arrival time of the installation and maintenance personnel, the method further includes:

[0023] Obtain the quality inspection result of the performance of the installation and maintenance personnel according to the quality inspection score of the performance of the installation and maintenance personnel, the scheduled arrival time, the actual arrival time, and the acceptable time difference for the user.

[0024] Further, the obtaining the quality inspection result of the performance of the installation and maintenance personnel according to the quality inspection score of the performance of the installation and maintenance personnel, the scheduled arrival time, the actual arrival time, and the acceptable time difference for the user specifically includes:

[0025] If the quality inspection score of the performance of the installation and maintenance personnel is higher than the preset standard value and the scheduled arrival time is the same as the actual arrival time, or the quality inspection score of the performance of the installation and maintenance personnel is higher than the preset standard value and the scheduled arrival time is different from the actual arrival time but within the acceptable time difference for the user, then the quality inspection result of the performance of the installation and maintenance personnel is qualified;

[0026] If the quality inspection score of the performance of the installation and maintenance personnel is lower than the preset standard value, or the scheduled arrival time of the installation and maintenance personnel is different from the actual arrival time and not within the acceptable time difference for the user, then the quality inspection result of the performance of the installation and maintenance personnel is unqualified.

[0027] In a second aspect, the present invention provides a quality inspection device for the performance of installation and maintenance personnel, including:

[0028] An acquisition module, configured to obtain a corresponding quality inspection call text according to the quality inspection call recording between the installation and maintenance personnel and the user;

[0029] An extraction module, connected to the acquisition module, configured to extract the scheduled arrival time of the installation and maintenance personnel and the acceptable time difference for the user from the quality inspection call text by using a pre-trained time information extraction model;

[0030] A quality inspection module, connected to the extraction module, configured to calculate the quality inspection score of the performance of the installation and maintenance personnel according to the scheduled arrival time of the installation and maintenance personnel, the acceptable time difference for the user, and the actual arrival time of the installation and maintenance personnel.

[0031] In a third aspect, the present invention provides a quality inspection device for the performance of installation and maintenance personnel, including a memory and a processor. A computer program is stored in the memory, and the processor is configured to run the computer program to implement the quality inspection method for the performance of installation and maintenance personnel described in the first aspect above.

[0032] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the quality inspection method for the performance of installation and maintenance personnel described in the first aspect above.

[0033] The quality inspection method, device and medium for the performance of installation and maintenance personnel provided by the present invention first obtain the corresponding quality inspection call text according to the quality inspection call recording between the installation and maintenance personnel and the user; then use a pre-trained time information extraction model to extract the scheduled on-site time of the installation and maintenance personnel and the time difference acceptable to the user from the quality inspection call text; finally, according to the scheduled on-site time of the installation and maintenance personnel, the time difference acceptable to the user, and the actual on-site time of the installation and maintenance personnel, calculate the quality inspection score for the performance of the installation and maintenance personnel. The present invention performs intelligent analysis on the quality inspection call text corresponding to the quality inspection call recording between the installation and maintenance personnel and the user through a pre-trained time information extraction model, real-time extracts the scheduled on-site time of the installation and maintenance personnel and the performance time information of the time difference acceptable to the user, and calculates the quality inspection score for the performance of the installation and maintenance personnel based on the scheduled on-site time of the installation and maintenance personnel, the time difference acceptable to the user, and the actual on-site time of the installation and maintenance personnel. While improving the quality inspection efficiency, it avoids the situation where the quality inspection result is inconsistent with the actual result, ensures that the installation and maintenance personnel can fulfill their obligations on time, improves user satisfaction, and solves the problem in the prior art that the quality inspection of the performance of installation and maintenance personnel mostly relies on manual listening to recordings, with low efficiency and insufficient data analysis ability, which easily leads to the inconsistency between the quality inspection result and the actual result. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is a flowchart of a quality inspection method for the performance of installation and maintenance personnel according to Embodiment 1 of the present invention;

[0035] Figure 2 is a flowchart of another quality inspection method for the performance of installation and maintenance personnel according to an embodiment of the present invention;

[0036] Figure 3 is a flowchart of the installation and maintenance personnel providing installation and maintenance services according to an embodiment of the present invention;

[0037] Figure 4 is a schematic structural diagram of a quality inspection device for the performance of installation and maintenance personnel according to Embodiment 2 of the present invention;

[0038] Figure 5 This is a schematic structural diagram of a quality inspection device for the performance compliance of installation and maintenance personnel in Embodiment 3 of the present invention. Detailed implementation manners

[0039] To enable those skilled in the art to better understand the technical solutions of the present invention, the following will further describe the embodiments of the present invention in detail with reference to the accompanying drawings.

[0040] It can be understood that the specific embodiments and accompanying drawings described herein are only used to explain the present invention, rather than limiting the present invention.

[0041] It can be understood that, without conflict, the various embodiments in the present invention and the features in the embodiments can be combined with each other.

[0042] It can be understood that for the convenience of description, only the parts related to the present invention are shown in the accompanying drawings of the present invention, and the parts unrelated to the present invention are not shown in the accompanying drawings.

[0043] It can be understood that each unit and module involved in the embodiments of the present invention may correspond to only one entity structure, or may be composed of multiple entity structures, or multiple units and modules may also be integrated into one entity structure.

[0044] It can be understood that, without conflict, the functions and steps marked in the flowcharts and block diagrams of the present invention may occur in a different order from that marked in the accompanying drawings.

[0045] It can be understood that in the flowcharts and block diagrams of the present invention, the possible system architectures, functions, and operations of the systems, devices, equipment, and methods according to the embodiments of the present invention are shown. Among them, each block in the flowchart or block diagram may represent a unit, module, program segment, or code, which contains executable instructions for implementing the specified function. Moreover, each block or combination of blocks in the block diagram and flowchart can be implemented by a hardware-based system for implementing the specified function, or by a combination of hardware and computer instructions.

[0046] It can be understood that the units and modules involved in the embodiments of the present invention can be implemented in software or in hardware. For example, the units and modules can be located in the processor.

[0047] Application overview

[0048] The existing quality inspection process for the performance compliance of installation and maintenance personnel has the following disadvantages:

[0049] 1. Lack of real-time performance: Traditional quality inspection methods mostly rely on manual listening to recordings, with low efficiency and difficulty in providing real-time feedback.

[0050] 2. Insufficient data analysis: The existing technologies are limited in data processing and analysis and cannot deeply mine the key information in the call content.

[0051] 3. Strong subjectivity: The results of manual evaluation are often affected by the subjective factors of the evaluators, which easily leads to inconsistent results.

[0052] In view of the above technical problems, the present application provides a quality inspection method, device and medium for the performance compliance of installation and maintenance personnel. By using a pre-trained time information extraction model to intelligently analyze the quality inspection call text corresponding to the quality inspection call recording of the installation and maintenance personnel and the user, the performance compliance time information such as the scheduled door-to-door time of the installation and maintenance personnel and the time difference acceptable to the user is extracted in real time, and a quality inspection score for the performance compliance of the installation and maintenance personnel is calculated based on the scheduled door-to-door time of the installation and maintenance personnel, the time difference acceptable to the user, and the actual door-to-door time of the installation and maintenance personnel. While improving the quality inspection efficiency, it avoids the situation where the quality inspection result is inconsistent with the actual result, ensures that the installation and maintenance personnel can fulfill their obligations on time, improves user satisfaction, and at least solves the problem that the quality inspection of the performance compliance of installation and maintenance personnel in the existing technology mostly relies on manual listening to recordings, with low efficiency, insufficient data analysis ability, and easy to cause the inconsistency between the quality inspection result and the actual result.

[0053] After introducing the basic principle of the present application, the various non-limiting embodiments of the present application will be specifically introduced with reference to the accompanying drawings.

[0054] Embodiment 1:

[0055] This embodiment provides a quality inspection method for the performance compliance of installation and maintenance personnel, as Figure 1 shown, the method includes:

[0056] Step S101: Obtain the corresponding quality inspection call text according to the quality inspection call recording of the installation and maintenance personnel and the user.

[0057] It should be noted that the quality inspection call recording of the installation and maintenance personnel and the user refers to the call recording in which the installation and maintenance personnel make an appointment for the door-to-door time with the user and ask about the time difference between the scheduled door-to-door time and the actual door-to-door time acceptable to the user before the installation and maintenance personnel go to the user's home for installation and maintenance services.

[0058] Specifically, in the quality inspection process, the quality inspection call recording of the installation and maintenance personnel and the user needs to be collected through a high-quality recording device. In order to protect the privacy of the installation and maintenance personnel's mobile phone numbers, the call initiated by the user on the platform is recorded in the form of IVR (Interactive Voice Response) double call. The recording format preferably uses wav, 16-bit mono, and the sampling rate is preferably 16 kHz to ensure clear sound quality. Use a cloud storage solution to ensure the security and reliability of the data and facilitate subsequent data analysis and processing.

[0059] In an alternative embodiment, obtaining the corresponding quality inspection call text according to the quality inspection call recording between the installation and maintenance personnel and the user specifically includes:

[0060] Decompose the quality inspection call recording between the installation and maintenance personnel and the user, perform Fourier transform on the decomposed quality inspection call recording between the installation and maintenance personnel and the user, and construct an image corresponding to the quality inspection call recording between the installation and maintenance personnel and the user;

[0061] Input the image corresponding to the quality inspection call recording between the installation and maintenance personnel and the user into a pre-trained deep full-sequence convolutional neural network, and output the quality inspection call text. Among them, the trained deep full-sequence convolutional neural network is trained based on images corresponding to multiple historical quality inspection call recordings between installation and maintenance personnel and users and the corresponding quality inspection call texts.

[0062] Specifically, the audio of the quality inspection call recording between the installation and maintenance personnel and the user is decomposed into frame-by-frame data. Fourier transform is performed on each frame of data to convert the time-domain signal of the audio to the frequency domain, and time and frequency are used as two dimensions to construct an image corresponding to the quality inspection call recording between the installation and maintenance personnel and the user. This image becomes the input of the pre-trained deep full-sequence convolutional neural network. Inside the network, through layers of convolutional layers and pooling layers, the features of the entire audio are extracted and analyzed for comprehensive modeling. After a series of complex operations, the model directly outputs the quality inspection call text corresponding to the audio content. Among them, the quality inspection call text can be the text composed of Chinese characters in the audio.

[0063] It should be noted that the trained deep full-sequence convolutional neural network is trained based on images corresponding to multiple historical quality inspection call recordings between installation and maintenance personnel and users and the corresponding quality inspection call texts. Its training process is a known prior art, and the present invention does not make any regulations.

[0064] Step S102: Use a pre-trained time information extraction model to extract the scheduled on-site visit time of the installation and maintenance personnel and the time difference acceptable to the user from the quality inspection call text.

[0065] It should be noted that the pre-trained time information extraction model is used to extract the scheduled on-site visit time of the installation and maintenance personnel and the time difference acceptable to the user from the quality inspection call text. It can be a pre-trained traditional machine learning model or a pre-trained large language model, preferably the ERNIE model.

[0066] In an alternative embodiment, the using a pre-trained time information extraction model to extract the scheduled on-site visit time of the installation and maintenance personnel and the time difference acceptable to the user from the quality inspection call text specifically includes:

[0067] Preprocess the quality inspection call text and normalize the features;

[0068] Input the preprocessed and feature-normalized quality inspection call text into the trained time information extraction model, and output the entity label corresponding to each word in the quality inspection call text, where the trained time information extraction model is trained based on the historically annotated quality inspection call text;

[0069] Extract time entities from the quality inspection call text according to the entity label corresponding to each word, and obtain the scheduled on-site visit time of the installation and maintenance personnel and the time difference acceptable to the user.

[0070] Specifically, entity extraction is performed on the quality inspection call text to obtain the scheduled on-site visit time of the installation and maintenance personnel and the time difference acceptable to the user. In this link, the quality inspection call text is first preprocessed: the dialogue is separated by roles, possible errors are corrected, special characters and stop words are removed, and then the text is segmented and the part-of-speech is annotated. Then feature normalization is performed: the preprocessed text is converted into word vectors using Word2Vec, the sentence is converted into a fixed-length vector using LSTM (Long Short-Term Memory), and the feature vector is standardized using Z-score. Then the trained time information extraction model is used for entity extraction: the preprocessed and feature-normalized quality inspection call text is input into the trained time information extraction model according to the required format, and the model outputs the entity label corresponding to each word in the quality inspection call text. Finally, entity decoding is performed: the label output by the model is converted into an entity, the time entity is extracted corresponding to the quality inspection call text, the extracted entities are de-duplicated and unified in format, and output in JSON to obtain the scheduled on-site visit time of the installation and maintenance personnel and the time difference acceptable to the user for subsequent use.

[0071] It should be noted that the trained time information extraction model is trained based on the quality inspection call text with historically annotated entity labels. During the model training process, the model is optimized to better perform the extraction task.

[0072] In an optional embodiment, the time information extraction model is an ERNIE model. The inputting the preprocessed and feature-normalized quality inspection call text into the trained time information extraction model and outputting the entity label corresponding to each word in the quality inspection call text specifically includes:

[0073] Input the preprocessed and feature-normalized quality inspection call text into the trained ERNIE model, predict the probability that each word in the quality inspection call text belongs to different entity categories through the trained ERNIE model, and determine the entity label corresponding to each word using maximum likelihood decoding according to the probability that each word belongs to different entity categories.

[0074] Specifically, the quality inspection call text after preprocessing and feature normalization is input into the pre-trained ERNIE model. The model predicts and outputs the probability that each word belongs to different entity categories, and uses maximum likelihood decoding to determine the entity label corresponding to each word.

[0075] Step S103: Calculate the quality inspection score of the fulfillment situation of the installation and maintenance personnel according to the scheduled door-to-door time of the installation and maintenance personnel, the acceptable time difference of the user, and the actual door-to-door time of the installation and maintenance personnel.

[0076] It should be noted that the actual door-to-door time of the installation and maintenance personnel refers to the time when the installation and maintenance personnel actually carry out the installation and maintenance service, and the actual door-to-door time can be manually recorded by the installation and maintenance personnel.

[0077] Specifically, the quality inspection score of the fulfillment situation of the installation and maintenance personnel is calculated through the following formula:

[0078]

[0079] Among them, defence is the quality inspection score of the fulfillment situation of the installation and maintenance personnel, timeA is the scheduled door-to-door time of the installation and maintenance personnel, timeB is the actual door-to-door time of the installation and maintenance personnel, timeC is the acceptable time difference of the user, n is the service quality impact coefficient, k is the adjustment coefficient related to the user acceptance, m is the weight coefficient affecting the service attitude, p is the weight coefficient affected by the time difference, and ∈ is a small positive number.

[0080] It is worth mentioning that the present invention has the following advantages: dynamic threshold adjustment. Through the above calculation formula, the present invention introduces a dynamic threshold adjustment mechanism to adjust the relationship between the actual scenario and the fulfillment situation of the installation and maintenance in real time to adapt to different scenario requirements.

[0081] In an optional embodiment, after calculating the quality inspection score of the fulfillment situation of the installation and maintenance personnel according to the scheduled door-to-door time of the installation and maintenance personnel, the acceptable time difference of the user, and the actual door-to-door time of the installation and maintenance personnel, the method further includes:

[0082] Obtain the quality inspection result of the fulfillment situation of the installation and maintenance personnel according to the quality inspection score of the fulfillment situation of the installation and maintenance personnel, the scheduled door-to-door time, the actual door-to-door time, and the acceptable time difference of the user.

[0083] Specifically, if the quality inspection score of the installation and maintenance personnel's performance compliance is higher than the preset standard value and the scheduled door-to-door time is the same as the actual door-to-door time, or the quality inspection score of the installation and maintenance personnel's performance compliance is higher than the preset standard value and the scheduled door-to-door time is different from the actual door-to-door time but within the acceptable time difference of the user, then the quality inspection result of the installation and maintenance personnel's performance compliance is qualified; if the quality inspection score of the installation and maintenance personnel's performance compliance is lower than the preset standard value, or the scheduled door-to-door time of the installation and maintenance personnel is different from the actual door-to-door time and not within the acceptable time difference of the user, then the quality inspection result of the installation and maintenance personnel's performance compliance is unqualified.

[0084] In a specific embodiment, the specific steps of the quality inspection method for the performance compliance of installation and maintenance personnel are as follows:

[0085] Step 1: Recording acquisition: In the quality inspection system, the call content of the installation and maintenance personnel needs to be collected through high-quality recording equipment. To protect the privacy of the installation and maintenance personnel's mobile phone numbers, the calls initiated by users on the platform are recorded in the form of IVR double calls. The recording format is wav, 16-bit mono, and the sampling rate is 16 kHz to ensure clear sound quality. Use cloud storage solutions to ensure the security and reliability of data, facilitating subsequent data analysis and processing.

[0086] Step 2: Audio-to-text conversion: The input audio is disassembled into frame-by-frame data. Each frame of audio undergoes a Fourier transform, which converts the time-domain signal of the audio to the frequency domain, taking time and frequency as two dimensions to construct an image. This image becomes the input to the deep full-sequence convolutional neural network. Inside the network, through layers upon layers of convolutional layers and pooling layers, the features of the entire audio are extracted and analyzed for comprehensive modeling. Finally, after a series of complex operations, the model directly outputs the text data corresponding to the audio content, which can be Chinese characters in the audio.

[0087] Step 3: Time extraction from complex text: Based on the steady-state data finally output in the previous step, a named entity recognition task is performed. The main purpose is to identify entity names with specific meanings from the text and classify them into predefined categories. Here, it mainly refers to time. Specifically, the ERNIE model in the PaddleNLP framework is used. The accurate extraction of time information in dialogue text is achieved. This ability is particularly prominent in the scenario of extracting installation and maintenance time because it overcomes the limitation of traditional models that require a large amount of data for training and can efficiently extract time information from dialogue text.

[0088] Step 4: Delimitation and diagnosis of the performance compliance of installation and maintenance: Calculate the score of the installation and maintenance personnel (i.e., the quality inspection score of the performance compliance of the installation and maintenance personnel) through the following formula to accurately determine the location and cause of the fault and provide targeted information for the operation and maintenance team:

[0089]

[0090] The parameter explanations are as follows: defence: the score of the installation and maintenance personnel, timeA: the specified arrival time (i.e., the reserved arrival time), timeB: the actual arrival time, timeC: the acceptable time difference for the user, n: the service quality impact coefficient, which can be customized and affects the range of the final score, k: the adjustment coefficient related to the user acceptance, which affects the sensitivity of the score change, m: the weight coefficient affecting the service attitude, which adjusts the influence degree of the service attitude on the score, p: the weight coefficient of the time difference impact, which adjusts the impact of the time difference on the final score, ∈: a small positive number to avoid the situation of division by zero.

[0091] The specific analysis of the formula is as follows:

[0092] 1. Logistic growth function:

[0093] This part reflects the relationship between the actual time, the specified time, and the acceptable time difference. In the form of a Logistic growth function, it can better describe the situation where the score drops rapidly when the time difference exceeds the acceptable range.

[0094] 2. Square root and denominator:

[0095] It reflects the influence of the time difference. The larger the time difference, the lower the score. At the same time, the square root and denominator settings are used to avoid excessive punishment.

[0096] 3. Inverse proportion influence:

[0097] It reflects the inverse proportion influence of the time difference. As the time difference increases, the score gradually decreases, reflecting the negative correlation between the service attitude and time.

[0098] It should be noted that by adjusting the values of the coefficients k, m, and p, the sensitivity of the formula and the influence of each parameter on the final score can be customized to make the formula more in line with the specific application scenario.

[0099] Step 5: Feedback and optimization: The installation and maintenance personnel assist in self-feedback on the quality inspection situation to ensure the continuous effectiveness of the model and calculation method. Real-time feedback module: Model optimization module: As time goes by, network behavior may change, so it is necessary to update and optimize periodically.

[0100] It is worth mentioning that the main advantages of the present invention include: 1. Massive data feedback: A system can be built to allow installation and maintenance personnel to review their own work orders. When there is a discrepancy with the actual situation, the data can be fed back to assist in adjusting the algorithm and calculation method. 2. Feature extraction: The optimized ERNIE model can effectively extract key features from complex texts, helping to reduce manual workload and improve the accuracy of subsequent anomaly detection. 3. Flexibility: The innovative calculation method comprehensively considers various input factors, and combined with manually adjustable parameters, it can easily demarcate the difference between abnormal situations and actual situations.

[0101] In another specific embodiment, in view of the challenges in defining the performance time of existing quality inspection installation and maintenance personnel and the limitations of existing quality inspection tools, the present invention provides a quality inspection method for the performance of installation and maintenance personnel based on the named entity recognition (NER) ability and innovative calculation method in natural language processing (NLP), as Figure 2 shown, the specific steps of using the ERNIE model and innovative calculation method to demarcate the quality inspection of the installation and maintenance service process are as follows:

[0102] Step 1: The call center platform creates a call for quality inspection between the user and the installation and maintenance personnel.

[0103] Step 2: Record the content of the call and store it in the cloud environment for quality inspection use.

[0104] Step 3: Use the engine to transcribe the voice into text.

[0105] Step 4: Perform entity extraction on the transcribed text. In this link, the text is first preprocessed. Separate the roles of the dialogue, correct possible errors, and remove special characters and stop words, then segment the text and label the part-of-speech. Then perform feature numerical normalization (i.e., feature normalization), convert the processed text into word vectors using Word2Vec, convert the sentence into a fixed-length vector using LSTM, and standardize the feature vector using Z-score. Then use ERNIE for entity extraction, select the appropriate version according to the task complexity and download the weights for loading. Input the processed data in the required format, where the text entities have been labeled. The model predicts and outputs the probability that each word belongs to different entity categories, and uses the maximum likelihood decoding to determine the entity label. During model training, based on learning rate adjustment, accuracy selection, and regularization parameter optimization, a specific method is used to optimize the model to better perform the extraction task. Finally, entity decoding is performed, converting the model output label into an entity. For example, the text corresponding to "O, O, B-PERSON, I-PERSON" extracts the time entity, de-duplicates and unifies the format of the extracted entities, and outputs them in JSON format for subsequent use.

[0106] Step 5: Define the parameters in the above formula according to the production scenario. By performing quality inspection on the conversation recordings and extracting the time entities in the text (i.e., the scheduled on-site service time of the installation and maintenance personnel and the time difference acceptable to the user), input the time entities and the known actual on-site service time into the formula to calculate the service score of the installation and maintenance. If the times are the same and the score is higher than the standard value, or the times are different but the time difference is within the specified range (i.e., the time difference is within the range acceptable to the user) and the score is higher than the standard value, it is determined to be qualified. If the times are different and the time difference is not within the specified range, or the score is lower than the standard value, it is determined to be unqualified.

[0107] It should be noted that, as Figure 3 shown, the specific process of the installation and maintenance personnel for installation and maintenance services is as follows: 1. Make an appointment with the user to clarify the on-site service time, and record the content of the call at this time. 2. Provide on-site services according to the actual situation. 3. Feedback and adjustment, and perform quality inspection on the compliance of the installation and maintenance personnel according to the recording at this time.

[0108] It is worth mentioning that the present invention collects and quality-inspects the conversation records, work order information, log data, etc. of the installation and maintenance personnel, and uses advanced NER (Named Entity Recognition) capabilities to comprehensively analyze a large amount of data, so as to detect problems such as user complaints or service quality decline caused by abnormal fluctuations in the compliance time of the installation and maintenance personnel. The present invention also has the following advantages: Specific task optimization: Compared with traditional models, the model performance is optimized in the following three ways: Learning rate adjustment, gradually reducing the learning rate (Learning Rate), and ensuring the stability and efficiency of model training by finely adjusting the learning rate value. Precision adjustment, during the model inference process, there are two precision values, fp16 and fp32. Among them, the fp16 precision has a faster text inference speed and is suitable for scenarios with high real-time requirements. By selecting the appropriate precision, the overall operation efficiency of the system can be improved while ensuring the model performance. Regularization adjustment, a key means to prevent model overfitting. According to the complexity and generalization ability requirements of the model, adjust the intensity of the regularization parameter to achieve the best balance state. This helps to reduce the interference of other entity information on the accuracy of time extraction and improve the robustness of the model. Multi-source data analysis: The present invention provides a more comprehensive perspective for analyzing service conditions by focusing on the on-site service time of the installation and maintenance personnel.

[0109] The quality inspection method for the fulfillment situation of installation and maintenance personnel provided by the embodiments of the present invention first obtains the corresponding quality inspection call text according to the quality inspection call recording between the installation and maintenance personnel and the user; then uses a pre-trained time information extraction model to extract the scheduled on-site visit time of the installation and maintenance personnel and the acceptable time difference of the user from the quality inspection call text; finally, calculates the quality inspection score for the fulfillment situation of the installation and maintenance personnel according to the scheduled on-site visit time of the installation and maintenance personnel, the acceptable time difference of the user, and the actual on-site visit time of the installation and maintenance personnel. The present invention intelligently analyzes the quality inspection call text corresponding to the quality inspection call recording between the installation and maintenance personnel and the user through a pre-trained time information extraction model, real-time extracts the fulfillment time information of the scheduled on-site visit time of the installation and maintenance personnel and the acceptable time difference of the user, and calculates the quality inspection score for the fulfillment situation of the installation and maintenance personnel based on the scheduled on-site visit time of the installation and maintenance personnel, the acceptable time difference of the user, and the actual on-site visit time of the installation and maintenance personnel. While improving the quality inspection efficiency, it avoids the situation where the quality inspection result is inconsistent with the actual result, ensures that the installation and maintenance personnel can fulfill their obligations on time, improves user satisfaction, and solves the problem in the prior art that the quality inspection of the fulfillment situation of installation and maintenance personnel mostly relies on manual listening to recordings, with low efficiency and insufficient data analysis ability, which easily leads to the inconsistency between the quality inspection result and the actual result.

[0110] Embodiment 2:

[0111] As Figure 4 shown, the present embodiment provides a quality inspection device for the fulfillment situation of installation and maintenance personnel, which is used to execute the above-mentioned quality inspection method for the fulfillment situation of installation and maintenance personnel, and includes:

[0112] An acquisition module 11, configured to obtain the corresponding quality inspection call text according to the quality inspection call recording between the installation and maintenance personnel and the user;

[0113] An extraction module 12, connected to the acquisition module 11, and configured to extract the scheduled on-site visit time of the installation and maintenance personnel and the acceptable time difference of the user from the quality inspection call text by using a pre-trained time information extraction model;

[0114] A quality inspection module 13, connected to the extraction module 12, and configured to calculate the quality inspection score for the fulfillment situation of the installation and maintenance personnel according to the scheduled on-site visit time of the installation and maintenance personnel, the acceptable time difference of the user, and the actual on-site visit time of the installation and maintenance personnel.

[0115] Further, the acquisition module 11 specifically includes:

[0116] A disassembly and construction unit, configured to disassemble the quality inspection call recording between the installation and maintenance personnel and the user, perform Fourier transform on the disassembled quality inspection call recording between the installation and maintenance personnel and the user, and construct an image corresponding to the quality inspection call recording between the installation and maintenance personnel and the user;

[0117] A first input / output unit, configured to input an image corresponding to a quality inspection call recording between the installation and maintenance personnel and the user into a pre-trained deep fully-sequential convolutional neural network, and output the quality inspection call text, wherein the trained deep fully-sequential convolutional neural network is trained based on images corresponding to quality inspection call recordings of multiple historical installation and maintenance personnel and users and corresponding quality inspection call texts.

[0118] Further, the extraction module 12 specifically includes:

[0119] A processing and normalization unit, configured to perform preprocessing and feature normalization on the quality inspection call text;

[0120] A second input / output unit, configured to input the preprocessed and feature-normalized quality inspection call text into a trained time information extraction model, and output an entity label corresponding to each word in the quality inspection call text, wherein the trained time information extraction model is trained based on historically annotated quality inspection call texts;

[0121] An extraction unit, configured to extract time entities from the quality inspection call text according to the entity label corresponding to each word, and obtain the scheduled on-site time of the installation and maintenance personnel and the time difference acceptable to the user.

[0122] Further, the time information extraction model is an ERNIE model, and the second input / output unit specifically is configured to:

[0123] Input the preprocessed and feature-normalized quality inspection call text into the trained ERNIE model, predict the probability that each word in the quality inspection call text belongs to different entity categories through the trained ERNIE model, and determine the entity label corresponding to each word by using maximum likelihood decoding according to the probability that each word belongs to different entity categories.

[0124] Further, the quality inspection module 13 specifically includes:

[0125] Calculate the quality inspection score of the fulfillment situation of the installation and maintenance personnel through the following formula:

[0126]

[0127] where defence is the quality inspection score of the fulfillment situation of the installation and maintenance personnel, timeA is the scheduled on-site time of the installation and maintenance personnel, timeB is the actual on-site time of the installation and maintenance personnel, timeC is the time difference acceptable to the user, n is the service quality impact coefficient, k is the adjustment coefficient related to user acceptance, m is the weight coefficient affecting service attitude, p is the weight coefficient affected by the time difference, and ∈ is a small positive number.

[0128] Further, the device further includes:

[0129] The module is used to obtain the performance quality inspection results of the installation and maintenance personnel based on the performance quality inspection score of the installation and maintenance personnel, the scheduled door-to-door time, the actual door-to-door time and the time difference acceptable to the user.

[0130] Furthermore, the obtaining module specifically includes:

[0131] The first result unit is used to determine that the performance quality inspection result of the installation and maintenance personnel is qualified if the performance quality inspection score of the installation and maintenance personnel is higher than the preset standard value and the scheduled on-site visit time is consistent with the actual on-site visit time, or if the performance quality inspection score of the installation and maintenance personnel is higher than the preset standard value and the scheduled on-site visit time is inconsistent with the actual on-site visit time but within a time difference acceptable to the user;

[0132] The second result unit is used to determine that if the quality inspection score of the installation and maintenance personnel's performance is lower than the preset standard value or the scheduled on-site time of the installation and maintenance personnel is inconsistent with the actual on-site time and is not within the time difference acceptable to the user, the quality inspection result of the installation and maintenance personnel's performance is unqualified.

[0133] Example 3:

[0134] refer to Figure 5 This embodiment provides a device for quality inspection of the performance of installation and maintenance personnel, including a memory 21 and a processor 22. The memory 21 stores a computer program, and the processor 22 is configured to run the computer program to execute the quality inspection method for the performance of installation and maintenance personnel in Example 1.

[0135] The memory 21 is connected to the processor 22 . The memory 21 may be a flash memory, a read-only memory, or other memory. The processor 22 may be a central processing unit or a single-chip microcomputer.

[0136] Example 4:

[0137] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for quality inspection of the performance of installation and maintenance personnel in the above-mentioned embodiment 1 is implemented.

[0138] The computer-readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information such as computer-readable instructions, data structures, computer program modules, or other data. The computer-readable storage medium includes, but is not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read Only Memory), flash memory or other memory technologies, CD-ROM (Compact Disc Read-Only Memory), digital versatile discs (DVDs) or other optical disc storage, magnetic cassettes, tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer.

[0139] In summary, the quality inspection method, device and medium for the fulfillment situation of the installation and maintenance personnel provided by the embodiments of the present invention first obtain the corresponding quality inspection call text according to the quality inspection call recording between the installation and maintenance personnel and the user; then use the pre-trained time information extraction model to extract the scheduled on-site time of the installation and maintenance personnel and the time difference acceptable to the user from the quality inspection call text; finally, calculate the quality inspection score for the fulfillment situation of the installation and maintenance personnel according to the scheduled on-site time of the installation and maintenance personnel, the time difference acceptable to the user, and the actual on-site time of the installation and maintenance personnel. Through the intelligent analysis of the quality inspection call text corresponding to the quality inspection call recording between the installation and maintenance personnel and the user by the pre-trained time information extraction model, the present invention can extract in real time the fulfillment time information such as the scheduled on-site time of the installation and maintenance personnel and the time difference acceptable to the user, and calculate the quality inspection score for the fulfillment situation of the installation and maintenance personnel based on the scheduled on-site time of the installation and maintenance personnel, the time difference acceptable to the user, and the actual on-site time of the installation and maintenance personnel. While improving the quality inspection efficiency, it avoids the situation where the quality inspection result is inconsistent with the actual result, ensures that the installation and maintenance personnel can fulfill their obligations on time, improves user satisfaction, and solves the problem that the quality inspection of the fulfillment situation of the installation and maintenance personnel in the prior art mostly relies on manual listening to recordings, with low efficiency and insufficient data analysis ability, which easily leads to the inconsistency between the quality inspection result and the actual result.

[0140] It can be understood that the above embodiments are merely exemplary embodiments adopted to illustrate the principles of the present invention, and the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also regarded as the protection scope of the present invention.

Claims

1. A quality inspection method for the performance of installation and maintenance personnel, characterized in that, The method includes: Obtaining a corresponding quality inspection call text based on the quality inspection call recording between the installation and maintenance personnel and the user; Using a pre-trained time information extraction model to extract the scheduled on-site time of the installation and maintenance personnel and the acceptable time difference of the user from the quality inspection call text; Calculating the quality inspection score of the performance compliance of the installation and maintenance personnel according to the scheduled on-site time of the installation and maintenance personnel, the acceptable time difference of the user, and the actual on-site time of the installation and maintenance personnel.

2. The method according to claim 1, wherein The obtaining a corresponding quality inspection call text based on the quality inspection call recording between the installation and maintenance personnel and the user specifically includes: Decomposing the quality inspection call recording between the installation and maintenance personnel and the user, performing Fourier transform on the decomposed quality inspection call recording between the installation and maintenance personnel and the user, and constructing an image corresponding to the quality inspection call recording between the installation and maintenance personnel and the user; Inputting the image corresponding to the quality inspection call recording between the installation and maintenance personnel and the user into a pre-trained deep fully-sequential convolutional neural network to output the quality inspection call text, where the trained deep fully-sequential convolutional neural network is trained based on images corresponding to multiple historical quality inspection call recordings between installation and maintenance personnel and users and the corresponding quality inspection call texts.

3. The method according to claim 1, wherein The using a pre-trained time information extraction model to extract the scheduled on-site time of the installation and maintenance personnel and the acceptable time difference of the user from the quality inspection call text specifically includes: Performing preprocessing and feature normalization on the quality inspection call text; Inputting the preprocessed and feature-normalized quality inspection call text into a trained time information extraction model to output the entity label corresponding to each word in the quality inspection call text, where the trained time information extraction model is trained based on historically annotated quality inspection call texts; Extracting time entities from the quality inspection call text according to the entity label corresponding to each word to obtain the scheduled on-site time of the installation and maintenance personnel and the acceptable time difference of the user.

4. The method according to claim 3, wherein The time information extraction model is an ERNIE model. The inputting the preprocessed and feature-normalized quality inspection call text into a trained time information extraction model to output the entity label corresponding to each word in the quality inspection call text specifically includes: Inputting the preprocessed and feature-normalized quality inspection call text into a trained ERNIE model, predicting the probability of each word in the quality inspection call text belonging to different entity categories through the trained ERNIE model, and determining the entity label corresponding to each word using maximum likelihood decoding according to the probability of each word belonging to different entity categories.

5. The method according to claim 1, characterized in that, The calculating the quality inspection score of the performance compliance of the installation and maintenance personnel according to the scheduled on-site time of the installation and maintenance personnel, the acceptable time difference of the user, and the actual on-site time of the installation and maintenance personnel specifically includes: Calculating the quality inspection score of the performance compliance of the installation and maintenance personnel through the following formula: Among them, defence is the quality inspection score of the fulfillment of the installation and maintenance personnel, timeA is the scheduled on-site visit time of the installation and maintenance personnel, timeB is the actual on-site visit time of the installation and maintenance personnel, timeC is the acceptable time difference for the user, n is the service quality impact coefficient, k is the adjustment coefficient related to the user acceptance, m is the weight coefficient affecting the service attitude, p is the weight coefficient affected by the time difference, and ∈ is a small positive number.

6. The method according to claim 1, wherein After calculating the quality inspection score of the fulfillment of the installation and maintenance personnel according to the scheduled on-site visit time of the installation and maintenance personnel, the acceptable time difference for the user, and the actual on-site visit time of the installation and maintenance personnel, the method further includes: Obtaining the quality inspection result of the fulfillment of the installation and maintenance personnel according to the quality inspection score of the fulfillment of the installation and maintenance personnel, the scheduled on-site visit time, the actual on-site visit time, and the acceptable time difference for the user.

7. The method according to claim 6, wherein The obtaining the quality inspection result of the fulfillment of the installation and maintenance personnel according to the quality inspection score of the fulfillment of the installation and maintenance personnel, the scheduled on-site visit time, the actual on-site visit time, and the acceptable time difference for the user specifically includes: If the quality inspection score of the fulfillment of the installation and maintenance personnel is higher than the preset standard value and the scheduled on-site visit time is the same as the actual on-site visit time, or the quality inspection score of the fulfillment of the installation and maintenance personnel is higher than the preset standard value and the scheduled on-site visit time is different from the actual on-site visit time but within the acceptable time difference for the user, then the quality inspection result of the fulfillment of the installation and maintenance personnel is qualified; If the quality inspection score of the fulfillment of the installation and maintenance personnel is lower than the preset standard value, or the scheduled on-site visit time of the installation and maintenance personnel is different from the actual on-site visit time and not within the acceptable time difference for the user, then the quality inspection result of the fulfillment of the installation and maintenance personnel is unqualified.

8. A quality inspection device for the performance of installation and maintenance personnel, characterized in that, Including: An acquisition module, configured to obtain the corresponding quality inspection call text according to the quality inspection call recording between the installation and maintenance personnel and the user; An extraction module, connected to the acquisition module, configured to extract the scheduled on-site visit time of the installation and maintenance personnel and the acceptable time difference for the user from the quality inspection call text by using a pre-trained time information extraction model; A quality inspection module, connected to the extraction module, configured to calculate the quality inspection score of the fulfillment of the installation and maintenance personnel according to the scheduled on-site visit time of the installation and maintenance personnel, the acceptable time difference for the user, and the actual on-site visit time of the installation and maintenance personnel.

9. A quality inspection device for the performance of installation and maintenance personnel, characterized in that, Including a memory and a processor, where a computer program is stored in the memory, and the processor is configured to run the computer program to implement the method for quality inspection of the fulfillment of the installation and maintenance personnel according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is executed by the processor, it implements the method for quality inspection of the fulfillment of the installation and maintenance personnel according to any one of claims 1-7.

Images