An intelligent one-click order processing system for after-sales service

Abstract

The present invention discloses an after-sales one-key order intelligent processing system, including: a user information processing module, used to verify and filter valid after-sales demand single user information; a product information merging module, used to extract product unique values, and obtain a product information list; an after-sales demand description text analysis module, used to determine whether the user has independently attempted to repair the product; a non-independent repair fault analysis module, used to determine the maintenance measures for which the user has not independently attempted to repair the product; an independent repair fault analysis module, used to determine the compound fault maintenance measures for which the user has independently attempted to repair the product; a maintenance instruction video generation module, used to generate a maintenance instruction video; a maintenance task order generation module, used to match the after-sales service area, and generate a maintenance task order. The present invention solves the problems of information asymmetry, low processing efficiency, and slow response speed in traditional after-sales service systems through information standardization, intelligent text processing, and intelligent scheduling.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to an after-sales one-key ordering intelligent processing system. Background Art

[0002] In today's e-commerce platforms and after-sales service systems, with the rapid development of e-commerce, the demand for after-sales service continues to increase, and enterprises are facing many challenges. Although the importance of after-sales service is becoming increasingly prominent, there are still many problems that need to be solved in the traditional after-sales processing process, which seriously affect the efficiency of after-sales service and user satisfaction. E-commerce platforms and joint guarantee after-sales companies often have independent product databases, with inconsistent information standards, making it difficult to effectively integrate. The traditional product information processing method lacks automation means, and the data integration efficiency is low, resulting in information asymmetry in the after-sales processing process, which is easy to cause misjudgment and delay service. When users submit after-sales needs, the processing of after-sales demand description text is another key link. The after-sales needs described by users usually contain a large amount of natural language text. The traditional text processing method relies on manual review, which is inefficient and prone to errors. Manual review is not only time-consuming, but also unable to quickly identify key fault information in the text and whether the user has made independent repair attempts, resulting in slow response and poor user experience. In particular, the traditional system lacks effective identification means for whether the user has made independent repair attempts. The user's independent repair attempt directly affects the subsequent maintenance measures and task arrangements, but the existing system is difficult to distinguish between the user's specific operations and fault phenomena, and cannot accurately determine the cause of the fault and provide personalized maintenance solutions. In this case, different fault handling methods are difficult to reasonably distinguish, which affects the maintenance effect and user satisfaction. In addition, the traditional maintenance task order generation method mainly relies on manual scheduling, lacks intelligent means, and is difficult to efficiently match after-sales service areas. Therefore, there is an urgent need for an efficient and intelligent after-sales service processing system to solve the above problems and improve the overall level of after-sales service. Summary of the invention

[0003] The present invention solves the problems existing in the above-mentioned prior art and provides an after-sales one-key order intelligent processing system, which mainly includes:

[0004] The user information processing module is used to obtain the user information dataset of the after-sales demand order, verify and filter the valid contact information, match the valid users to generate the address copy list, and convert it into regional information and latitude and longitude coordinates through the geocoding API;

[0005] The product information merging module is used to extract the unique value of the product according to the after-sales demand order database, match the product information of the e-commerce platform and the product database data of the joint guarantee after-sales enterprise through association mapping, and merge the product information to obtain a product information list;

[0006] After-sales demand description text analysis module is used to perform K-means clustering based on the after-sales demand description text of the product, generate repair operation labels and fault phenomenon labels for the text segments, and determine whether the user has made independent repair attempts on the product;

[0007] The non-independent repair fault analysis module is used to match the fault code and the fault cause according to the fault phenomenon label of the after-sales demand description text that is judged as the user has not made an independent repair attempt on the product, and determine the repair measures to be obtained;

[0008] The independent repair fault analysis module is used to annotate the text segments in time sequence according to the user repair operation and fault phenomenon labels of the after-sales demand description text that is judged to be the user's independent repair attempt on the product, determine the operation-fault chain and the root cause of the fault, and obtain the repair measures for the compound fault through the maintenance knowledge base;

[0009] A maintenance guidance video generation module is used to determine after-sales demand orders that can be repaired by themselves according to the complexity of the fault repair measures or the complexity of the complex fault repair measures, generate a maintenance guidance video, and send it to the user of the after-sales demand order that can be repaired by themselves;

[0010] The maintenance task order generation module is used to formulate area matching rules according to the user's longitude and latitude, match the after-sales service area, and generate a maintenance task order.

[0011] Furthermore, the user information processing module is used to obtain the user information data set of the after-sales demand form, verify and screen the valid contact information, match the valid user to generate the address copy list, and convert it into regional information and latitude and longitude coordinates through the geocoding API, including:

[0012] According to the API interface provided by the e-commerce platform, the user information data set of the after-sales demand form is obtained, including the user's name, contact information, and address; by cleaning the data, removing redundant information, correcting format errors, and using regular expressions to standardize the user name, contact information, and address format, a standardized user information data set is obtained; by traversing the user information data set, the mobile phone number field of each user is extracted to obtain a list containing all user mobile phone numbers; the mobile phone number list is passed to the verification API to determine the availability and validity of each number, and the status of each number is returned through the API, including valid numbers, empty numbers, or invalid numbers; invalid numbers and empty numbers are eliminated, valid numbers are screened out, and a list of valid numbers is generated; by matching valid mobile phone numbers with the user information data set, all users with valid contact information and their demand forms are screened out; user information of invalid numbers is extracted, and a warning prompt is sent through the message system of the e-commerce platform to notify users to update their contact information; by traversing the after-sales demand forms of valid users, the address field is extracted to generate a text list containing all valid user addresses; by calling the geocoding API, the text list of valid user addresses is converted into regional information and latitude and longitude coordinates.

[0013] Furthermore, the product information merging module is used to extract the unique value of the product according to the after-sales demand order database, match the e-commerce platform product information and the joint guarantee after-sales enterprise product database data through association mapping, and merge the product information to obtain a product information list, including:

[0014] By querying the after-sales demand order database, the corresponding product unique value in each after-sales demand order is extracted, and the unique value is the SKU number; using the association mapping technology, according to the product unique value, it is matched with the product database of the e-commerce platform to obtain the product information of the e-commerce platform, including the manufacturer, category, specification, brand and model information; through the joint guarantee after-sales enterprise product database, according to the product unique value, the products that meet the joint guarantee after-sales enterprise product database are determined, and the product information that meets the joint guarantee after-sales enterprise product database is obtained, including warranty status, repair history, fault code, after-sales support agreement; data standardization is performed on the product information from the e-commerce platform and the joint guarantee after-sales enterprise product database, and the field name and data format are unified; matching rules are defined based on the product unique value as the primary key, and the hash mapping algorithm is used to find the e-commerce platform product information and the joint guarantee after-sales enterprise product database product information with the same product unique value, and the matched product information in the e-commerce platform and the joint guarantee after-sales enterprise product database are merged to obtain a product information list.

[0015] Furthermore, the after-sales demand description text analysis module is used to perform K-means clustering based on the after-sales demand description text of the product, generate repair operation labels and fault phenomenon labels of the text segments, and determine whether the user has made an independent repair attempt on the product, including:

[0016] According to the after-sales demand form, obtain the after-sales demand description text of the product; use regular expressions to segment the after-sales demand description text to obtain a vocabulary list for each text segment; use a predefined stop word list to filter out irrelevant words to obtain a streamlined vocabulary list; use SpaCy to perform part-of-speech tagging on the filtered words, mark the part of speech of each word, and extract nouns, verbs, and adjectives; obtain the vocabulary index representation of each text segment according to the streamlined vocabulary list, and use the word segmenter of the BERT model to map each word to an index representation in the BERT vocabulary table; input the index representation of each text segment into the pre-trained BERT model for forward propagation to obtain the embedding vector of each word in the context; extract the embedding vector corresponding to the CLS tag output by the BERT model as the feature vector of the entire text segment, and obtain the text segment feature vector containing context information; according to the text feature vector, use the K-means clustering algorithm to cluster the text segment to obtain the cluster label of the text segment, and the cluster label includes the repair operation label and the fault phenomenon label; according to the repair operation label of the text segment, determine whether the user has made an independent repair attempt on the product.

[0017] Furthermore, the non-independent repair fault analysis module is used to match the fault code and the fault cause according to the fault phenomenon label of the after-sales demand description text that is determined to be the user's no independent repair attempt on the product, and determine to obtain the repair measures, including:

[0018] If the user does not attempt to repair the product independently, the fault mode text corresponding to the fault code is obtained through the joint warranty after-sales enterprise product database API; the fault phenomenon label and the fault mode text corresponding to the fault code are converted into feature vectors using the TF-IDF algorithm, and the similarity between the fault phenomenon label feature vector and the fault mode feature vector corresponding to the fault code is obtained using the Pearson correlation coefficient calculation method; the fault code corresponding to each fault phenomenon label is determined according to a preset similarity threshold, and if the similarity between the fault phenomenon label feature vector and the fault mode feature vector corresponding to the fault code is greater than the preset similarity threshold, the fault phenomenon label is judged to match the fault code, and the cause of the fault is obtained based on the fault code; by calling the maintenance knowledge base API, the cause of the fault is used as the query condition to obtain the fault repair measures and the complexity of the fault repair measures, which include simple, medium, and complex.

[0019] Furthermore, the independent repair fault analysis module is used to perform time sequence annotation on text segments according to user repair operations and fault phenomenon labels of after-sales demand description texts determined as independent repair attempts by users on products, determine operation-fault chains and root causes of faults, and obtain complex fault repair measures through a repair knowledge base, including:

[0020] If the user attempts to repair the product independently, the text segments are time-sequence labeled according to the repair operation labels and fault phenomenon labels, predefined conjunctions and predefined time description words of the text segments to determine the order of the text segments; based on the time-sequence labeled text segments, a long short-term memory network model is used to perform contextual semantic analysis to determine the repair operation of each repair attempt and the resulting fault change phenomenon, identify the initial fault phenomenon and the fault change phenomenon caused by each repair attempt, and the fault change phenomenon includes new faults and fault conversions; obtain historical maintenance feedback text, and annotate the fault change phenomenon caused by each repair attempt and the corresponding fault phenomenon label, as well as the user's erroneous repair attempt operation, use a recurrent neural network to build a user erroneous repair attempt operation prediction model, predict the user's erroneous repair attempt operation, and determine the root cause of the fault change phenomenon; by calling the maintenance knowledge base API, use the root cause of the fault as the query condition, determine the maintenance measures for each fault change phenomenon, and obtain the composite fault maintenance measures and the complexity of the composite fault maintenance measures.

[0021] Furthermore, the maintenance instruction video generation module is used to determine the after-sales demand form that can be repaired by oneself according to the complexity of the fault repair measures or the complexity of the compound fault repair measures, generate a maintenance instruction video, and send it to the user of the after-sales demand form that can be repaired by oneself, including:

[0022] Obtain after-sales demand orders with complexity levels of simple fault repair measures and complex fault repair measures, and mark the after-sales demand orders as after-sales demand orders that can be repaired by oneself; determine the repair parts according to the fault repair measures and complex fault repair measures; obtain parts images directly connected to the repair parts from the product part image database, mark the obtained parts images, and indicate the operation positions; determine the repair operation steps of the repair measures according to the maintenance manual, and arrange the parts images in sequence according to the maintenance operation steps to form a picture sequence that shows the operation process; generate text descriptions of each repair operation step according to the maintenance manual, and use text-to-speech conversion technology to convert the text descriptions of each repair operation step into voice; synchronize the marked pictures and the generated voice, and use the video editing software Adobe to synthesize the picture sequence and voice into a repair guidance video; send the repair guidance video to the user account and mobile phone of the after-sales demand order that can be repaired by oneself.

[0023] Furthermore, the maintenance task order generation module is used to formulate a region matching rule according to the user's longitude and latitude, match the after-sales service area, and generate a maintenance task order, including:

[0024] Based on the user's latitude and longitude information, regional matching rules are formulated to match the user's latitude and longitude information with the after-sales service area of ​​the after-sales company to determine the after-sales service area to which the user belongs; based on the fault phenomenon identification results and fault repair measures in the after-sales demand description text that cannot be repaired by oneself, a maintenance task order is generated, and the generated maintenance task order is assigned to the after-sales service area to which the user belongs. The maintenance task order content includes the fault labeling of the initial fault phenomenon and the fault change phenomenon caused by each repair attempt; the maintenance task assignment results are sent to the user account and mobile phone, and the maintenance time is arranged; through the real-time monitoring system, the execution progress of each maintenance task is tracked and monitored; the user feedback information after the maintenance task is completed is obtained to conduct quality evaluation and improvement of after-sales maintenance services.

[0025] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0026] The present invention provides an intelligent processing system for after-sales one-key ordering. Through the effective integration of data between an e-commerce platform and a joint guarantee after-sales enterprise, information islands are eliminated, and data consistency and processing efficiency are improved. The present invention can intelligently process the after-sales demand text of users, can quickly generate repair operations and fault phenomenon labels, and effectively determine whether the user has made independent repair attempts, thereby improving the efficiency and accuracy of text processing. For different types of faults, the present invention can quickly match fault codes and determine the cause of the fault, and provide targeted repair measures. For the user's independent repair attempt, the present invention accurately identifies the root cause of the fault by analyzing the operation and the fault chain, and provides a personalized composite fault repair plan to ensure the pertinence and effectiveness of the repair plan. The present invention can also generate personalized repair guidance videos according to the complexity of the repair measures to help users perform self-repair, improve user experience, and reduce unnecessary waste of after-sales resources. The present invention ensures the efficient allocation and execution of maintenance tasks through precise regional matching rules and intelligent task order generation. The present invention solves the problems of information asymmetry, low processing efficiency and slow response speed in traditional after-sales service systems through information standardization, intelligent text processing and intelligent scheduling, and significantly improves the efficiency, accuracy and user satisfaction of after-sales services. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flow chart of an after-sales one-key ordering intelligent processing system of the present invention;

[0028] Figure 2 A schematic diagram of an after-sales one-key ordering intelligent processing system of the present invention;

[0029] Figure 3 It is another schematic diagram of an after-sales one-key ordering intelligent processing system of the present invention; DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0031] like Figure 1-3 In this embodiment, an after-sales one-key order intelligent processing system may specifically include:

[0032] Step S101, the user information processing module is used to obtain the after-sales demand form user information data set, verify and filter the valid contact information, match the valid users to generate an address copy list, and convert it into regional information and latitude and longitude coordinates through the geocoding API.

[0033] According to the API interface provided by the e-commerce platform, the user information dataset of the after-sales demand form is obtained, including the user's name, contact information, and address. By cleaning the data, removing redundant information, correcting format errors, and using regular expressions to standardize the user name, contact information, and address format, a standardized user information dataset is obtained. By traversing the user information dataset, the mobile phone number field of each user is extracted to obtain a list of all user mobile phone numbers. The mobile phone number list is passed to the verification API to determine the availability and validity of each number, and the status of each number is returned through the API, including valid numbers, empty numbers, or invalid numbers. Invalid numbers and empty numbers are removed, valid numbers are screened out, and a list of valid numbers is generated. By matching valid mobile phone numbers with user information datasets, all users with valid contact information and their demand forms are screened out. User information of invalid numbers is extracted, and warning prompts are sent through the e-commerce platform's message system to notify users to update their contact information. By traversing the after-sales demand forms of valid users, the address field is extracted to generate a copy list containing all valid user addresses. By calling the geocoding API, the copy list of valid user addresses is converted into regional information and latitude and longitude coordinates.

[0034] Exemplarily, the after-sales demand form user information data set obtained from the e-commerce platform includes user A, phone number 001, address 100 I Street, H District, G City. User B, phone number 002, address 200 L Street, K District, J City. User C, phone number 003, address 300 O Street, N District, M City. User D, phone number 004, address 400 R Street, Q District, P City. Clean these data, remove redundant information, and ensure that necessary fields such as user name, contact information, and address are retained. Correct format errors through regular expressions to ensure that the phone number is the standard number of phone numbers and the address contains province, city, and district information. The standardized user information data set obtained after cleaning. Extract the mobile phone number field of each user, pass these mobile phone numbers to the verification API, and return the status of each number. Phone number 001 is valid, 002 is valid, and 003 is valid. The list of valid numbers screened out is: 001, 002, 003. Match valid mobile phone numbers with user information datasets, and filter out all users with valid contact information and their demand sheets. User A, phone number is 001, and address is No. 100, I Street, H District, G City. User B, phone number is 002, and address is No. 200, L Street, K District, J City. User C, phone number is 003, and address is No. 300, O Street, N District, M City. User D's phone number is 004, which is an invalid number. A reminder is sent to user D through the e-commerce platform's message system. Extract the address field of the after-sales demand form of valid users and generate a list containing all valid user addresses, including No. 100 I Street, H District, G City, No. 200 L Street, K District, J City, and No. 300 O Street, N District, M City. Call the geocoding API to convert the valid user addresses into regional information and longitude and latitude coordinates. The returned results are 116.3245 East longitude, 39.9857 North latitude, 121.5646 East longitude, 31.2295 North latitude, 104.0677 East longitude, 30.5745 North latitude.

[0035] Step S102, a product information merging module is used to extract a unique product value according to the after-sales demand order database, match the e-commerce platform product information and the joint guarantee after-sales enterprise product database data through association mapping, and merge the product information to obtain a product information list.

[0036] By querying the after-sales demand order database, the corresponding product unique value in each after-sales demand order is extracted, and the unique value is the SKU number. Using the association mapping technology, according to the product unique value, it is matched with the product database of the e-commerce platform to obtain the product information of the e-commerce platform, including the manufacturer, category, specification, brand and model information. Through the joint guarantee after-sales enterprise product database, according to the product unique value, the products that meet the joint guarantee after-sales enterprise product database are determined, and the product information that meets the joint guarantee after-sales enterprise product database is obtained, including warranty status, repair history, fault code, and after-sales support agreement. Data standardization is performed on product information from the e-commerce platform and the joint guarantee after-sales enterprise product database, and the field name and data format are unified. Based on the product unique value as the primary key, the matching rule is defined, and the hash mapping algorithm is used to find the e-commerce platform product information and the joint guarantee after-sales enterprise product database product information with the same product unique value, and the matching product information in the e-commerce platform and the joint guarantee after-sales enterprise product database is merged to obtain a product information list.

[0037] For example, user C reports that the camera of SKU789 is blurry. The corresponding product information is obtained from the product database of the e-commerce platform. The manufacturer of SKU123 is manufacturer 1, the category is smart phone, the specification is 128GB, the brand is brand 1, and the model is model A. The manufacturer of SKU456 is manufacturer 2, the category is tablet computer, the specification is 256GB, the brand is brand 2, and the model is model B. The manufacturer of SKU789 is manufacturer 3, the category is smart watch, the specification is 64GB, the brand is brand 3, and the model is model C. Next, through the joint warranty after-sales enterprise product database, according to the product SKU, the repair status of SKU123 is obtained as under warranty, the repair history is 1 time, the fault code is E001, and the after-sales support agreement is standard warranty. The warranty status of SKU456 is out of warranty, the repair history is 2 times, the fault code is E002, and the after-sales support agreement is extended warranty. The warranty status of SKU789 is in warranty, the repair history is 0 times, the fault code is E003, and the after-sales support agreement is standard warranty. Data standardization is performed on product information from e-commerce platforms and joint warranty after-sales enterprise product databases, and field names and data formats are unified. Use product SKU as the primary key to define matching rules, and use the hash mapping algorithm to find and merge product information with the same SKU.

[0038] Step S103, the after-sales demand description text analysis module is used to perform K-means clustering based on the after-sales demand description text of the product, generate repair operation labels and fault phenomenon labels for the text segments, and determine whether the user has made an independent repair attempt on the product.

[0039] According to the after-sales demand form, obtain the after-sales demand description text of the product. Use regular expressions to segment the after-sales demand description text and obtain the vocabulary list of each text segment. Use the predefined stop word list to filter out irrelevant words and obtain a streamlined vocabulary list. Use SpaCy to perform part-of-speech tagging on the filtered words, annotate the part of speech of each word, and extract nouns, verbs, and adjectives. According to the streamlined vocabulary list, obtain the vocabulary index representation of each text segment, and use the word segmenter of the BERT model to map each word to the index representation in the BERT vocabulary. Input the index representation of each text segment into the pre-trained BERT model for forward propagation to obtain the embedding vector of each word in the context. By extracting the embedding vector corresponding to the CLS tag output by the BERT model as the feature vector of the entire text segment, obtain the feature vector of the text segment containing context information. According to the text feature vector, use the K-means clustering algorithm to cluster the text segment to obtain the cluster label of the text segment, which includes the repair operation label and the fault phenomenon label. According to the repair operation label of the text segment, determine whether the user has made an independent repair attempt on the product.

[0040] For example, a user's after-sales demand form includes a product after-sales demand description text, which contains the following content: My phone cannot be turned on. I tried to restart and replace the battery, but it still did not respond. Use regular expressions to segment the text, and the resulting vocabulary list includes my, phone, cannot be turned on, tried, restarted, replaced, battery, but, still, no response. Use a predefined stop word list to filter out irrelevant words, such as my, still, and but, and obtain a streamlined vocabulary list of phone, cannot be turned on, tried, restarted, replaced, battery, and still, no response. Use SpaCy to perform part-of-speech tagging on these words, extract nouns, verbs, and adjectives, and the results are nouns: phone, battery, response. Verbs: try to restart, replace, adjectives: cannot be turned on, no response. Based on these streamlined vocabulary lists, obtain the vocabulary index representation of each text fragment, and use the BERT model's tokenizer to map each word to an index representation in the BERT vocabulary. Input these index representations into the pre-trained BERT model for forward propagation to obtain the embedding vector of each word in the context. By extracting the embedding vector corresponding to the CLS tag output by the BERT model as the feature vector of the entire text segment, the feature vector of the text segment containing contextual information is obtained. According to the text feature vector, the K-means clustering algorithm is used to cluster the text segment to obtain the cluster label of the text segment. The cluster label includes the repair operation label and the fault phenomenon label. For example, attempts to restart and replace the battery are marked as repair operation labels, while failure to start and no response are marked as fault phenomenon labels. According to the repair operation label of the text segment, it is judged whether the user has made an independent repair attempt on the product. Since the user reported attempts to restart and replace the battery, it can be judged that the user did make an independent repair attempt.

[0041] Step S104, the non-independent repair fault analysis module is used to match the fault code and the fault cause according to the fault phenomenon label of the after-sales demand description text that indicates that the user has not made an independent repair attempt on the product, and determine to obtain repair measures.

[0042] If the user does not attempt to repair the product independently, the fault mode text corresponding to the fault code is obtained through the joint warranty after-sales enterprise product database API; the fault phenomenon label and the fault mode text corresponding to the fault code are converted into feature vectors using the TF-IDF algorithm, and the similarity between the fault phenomenon label feature vector and the fault mode feature vector corresponding to the fault code is obtained using the Pearson correlation coefficient calculation method; the fault code corresponding to each fault phenomenon label is determined according to a preset similarity threshold, and if the similarity between the fault phenomenon label feature vector and the fault mode feature vector corresponding to the fault code is greater than the preset similarity threshold, the fault phenomenon label is judged to match the fault code, and the cause of the fault is obtained based on the fault code; by calling the maintenance knowledge base API, the cause of the fault is used as the query condition to obtain the fault repair measures and the complexity of the fault repair measures, which include simple, medium, and complex.

[0043] For example, a user's after-sales demand description text content is that the device displays error code E104, but no repair attempt is made. The fault mode text corresponding to fault code E104 is obtained through the product manual. The fault mode text describing E104 in the manual is "battery failure, cannot be charged, it may be a battery connection problem or battery aging." Use the TF-IDF algorithm to convert the fault phenomenon label battery problem and the fault mode text corresponding to the fault code E104 "battery failure, cannot be charged, it may be a battery connection problem or battery aging" into feature vectors. After feature vectorization, the Pearson correlation coefficient calculation method is used to calculate the similarity between the fault phenomenon label feature vector and the fault mode feature vector corresponding to the fault code E104. After calculation, the fault phenomenon label battery is obtained. The similarity of the feature vectors of the fault mode text corresponding to the battery problem and fault code E104 is 0.85. According to the preset similarity threshold of 0.8, since 0.85 is greater than 0.8, it is determined that the fault phenomenon label battery problem matches the fault code E104. Based on this matching result, it is confirmed that the fault cause of the fault code E104 is a battery connection problem or battery aging. By calling the maintenance knowledge base API, using battery connection problem or battery aging as the query condition, the corresponding maintenance measures are obtained. If the maintenance measures returned by the API include checking and reconnecting the battery interface, and if the problem still exists, replacing the battery, the complexity of the maintenance measures is simple.

[0044] Step S105, an independent repair fault analysis module is used to time-sequence the text segments according to the user repair operations and fault phenomenon labels of the after-sales demand description text that is judged to be the user's independent repair attempt on the product, determine the operation-fault chain and the root cause of the fault, and obtain the complex fault repair measures through the maintenance knowledge base.

[0045] If the user attempts to repair the product independently, the text segments are time-sequence labeled according to the repair operation labels and fault phenomenon labels, predefined conjunctions and predefined time description words of the text segments to determine the order of the text segments; based on the time-sequence labeled text segments, a long short-term memory network model is used to perform contextual semantic analysis to determine the repair operation of each repair attempt and the resulting fault change phenomenon, identify the initial fault phenomenon and the fault change phenomenon caused by each repair attempt, and the fault change phenomenon includes new faults and fault conversions; obtain historical maintenance feedback text, and annotate the fault change phenomenon caused by each repair attempt and the corresponding fault phenomenon label, as well as the user's erroneous repair attempt operation, use a recurrent neural network to build a user erroneous repair attempt operation prediction model, predict the user's erroneous repair attempt operation, and determine the root cause of the fault change phenomenon; by calling the maintenance knowledge base API, use the root cause of the fault as the query condition, determine the maintenance measures for each fault change phenomenon, and obtain the composite fault maintenance measures and the complexity of the composite fault maintenance measures.

[0046] For example, a user encountered a problem when using a smart TV and tried to repair it independently. The after-sales demand description submitted by the user mentioned that there was no sound after the TV was turned on, but the image was displayed normally. The user first checked the audio settings and found that the volume had been adjusted to the maximum, but there was still no sound. Then the user restarted the TV, and the problem still existed. Then the user replaced the HDMI cable, and the sound returned to normal, but the screen flickered. In order to fix the screen flicker, the user adjusted the display settings, and the screen flicker phenomenon was alleviated, but the sound disappeared again. Through time series annotation, the order relationship of these text fragments can be determined, and the repair operation labels and fault phenomenon labels can be extracted. The initial fault phenomenon is no sound, the first and second repair attempts have no change, the third repair attempt to replace the HDMI cable causes the sound to be restored but the screen flickering fault is added, and the fourth repair attempt adjusts the display settings to alleviate the screen flickering but the sound disappears again. The long short-term memory network model is used to perform contextual semantic analysis on the text fragments after time series annotation, determine the repair operation and the resulting fault change phenomenon of each repair attempt, and identify the user's wrong repair attempt operation, such as replacing an incompatible HDMI cable. The prediction model of user error repair attempts built by recurrent neural network and the annotation data in historical repair feedback texts predicts that the user's replacement of incompatible HDMI cables is the main cause of new faults. By calling the repair knowledge base API and using the root cause of the fault as the query condition, the best repair measure is determined, and the user is advised to use a compatible HDMI cable and check and update the audio and display drivers, and a detailed composite fault repair plan is generated, and the repair measure of the composite fault repair plan is medium.

[0047] Step S106, a maintenance instruction video generation module is used to determine the after-sales demand form that can be repaired by oneself according to the complexity of the fault repair measures or the complexity of the compound fault repair measures, generate a maintenance instruction video, and send it to the user of the after-sales demand form that can be repaired by oneself.

[0048] Obtain after-sales demand orders with simple fault repair measures and compound fault repair measures, and mark the after-sales demand orders as self-repairable after-sales demand orders. Determine the repair parts according to the fault repair measures and compound fault repair measures. Obtain the parts images directly connected to the repair parts from the product parts image database, mark the obtained parts images, and indicate the operation positions. Determine the repair operation steps of the repair measures according to the maintenance manual, and arrange the parts images in sequence according to the maintenance operation steps to form a picture sequence showing the operation process. Generate a text description of each repair operation step according to the maintenance manual, and use text-to-speech conversion technology to convert the text description of each repair operation step into speech. Synchronize the marked pictures and the generated speech, and use the video editing software Adobe to synthesize the picture sequence and speech into a repair guidance video. Send the repair guidance video to the user account and mobile phone of the after-sales demand order that can be repaired by oneself.

[0049] For example, if the after-sales demand description text content in the after-sales demand form provided by a user is that the laptop screen is not bright, the maintenance measure is to replace the screen, and the complexity of the maintenance measure is simple maintenance, then the after-sales demand form is marked as an after-sales demand form that can be repaired by oneself. The screen and its directly connected parts pictures are obtained through the product part picture database, including the screen screw position and the screen connection line, and the obtained pictures are annotated to indicate the specific operation positions that need to be disassembled and connected, such as marking the position of the screws and the connection line. According to the maintenance manual, the operation steps for replacing the screen include using a screwdriver to remove the four screws on the edge of the screen, disconnecting the screen connection line, removing the old screen and installing the new screen, reconnecting the screen connection line, and tightening the screws to complete the replacement. According to these steps, the annotated parts pictures are arranged in sequence to form a picture sequence showing the operation process. The text description of each maintenance operation step is generated according to the maintenance manual. The text description of each maintenance operation step is converted into a voice file using text-to-speech conversion technology. Then, the annotated images and the generated voice are synchronized, and the image sequence and voice are synthesized into a maintenance instruction video using the video editing software Adobe Premiere. The generated maintenance instruction video is sent to the user account and mobile phone of the after-sales demand form that can be repaired by themselves.

[0050] Step S107, a maintenance task order generation module is used to formulate a region matching rule according to the user's longitude and latitude, match the after-sales service area, and generate a maintenance task order.

[0051] Based on the user's latitude and longitude information, regional matching rules are formulated to match the user's latitude and longitude information with the after-sales service area of ​​the after-sales company to determine the after-sales service area to which the user belongs; based on the fault phenomenon identification results and fault repair measures in the after-sales demand description text that cannot be repaired by oneself, a maintenance task order is generated, and the generated maintenance task order is assigned to the after-sales service area to which the user belongs. The maintenance task order content includes the fault labeling of the initial fault phenomenon and the fault change phenomenon caused by each repair attempt; the maintenance task assignment results are sent to the user account and mobile phone, and the maintenance time is arranged; through the real-time monitoring system, the execution progress of each maintenance task is tracked and monitored; the user feedback information after the maintenance task is completed is obtained to conduct quality evaluation and improvement of after-sales maintenance services.

[0052] Exemplarily, the user is located in city U, whose longitude and latitude are 118.2437 east longitude and 34.0522 north latitude, and this area belongs to the after-sales service area V of the after-sales enterprise. The after-sales demand description text submitted by the user is that the smart washing machine cannot start and the display screen flickers. The user first checked the power plug and found that the plug was connected normally, but the problem still existed. Then the user restarted the washing machine, and the display screen flickered and there was no change. Then, the user replaced the power plug, and the display screen flickering problem was alleviated, but the washing machine still could not start. Finally, the user checked the internal circuit connection, and the washing machine was able to start, but the drainage pump did not work. According to the after-sales demand description text, the fault phenomenon was identified as a power failure, and the fault repair measure was determined to be replacing the power control module, and a maintenance task order was generated. The maintenance task order content includes the initial fault phenomenon of being unable to start, the first repair attempt to check the power plug did not change, the second repair attempt to restart the washing machine did not change, the third repair attempt to replace the power plug after the display screen flickering problem was alleviated but the washing machine still could not start, and the fourth repair attempt to check the internal circuit connection after the washing machine was able to start but the fault of the drainage pump not working was added. The generated maintenance task order is assigned to the after-sales service area V to which the user belongs, and the maintenance task assignment result is sent to the user's account and mobile phone, and the maintenance time is arranged at the same time. The execution progress of the maintenance task is tracked and monitored through the real-time monitoring system. After the maintenance task is completed, the user's feedback information is obtained, and the quality of the after-sales maintenance service is evaluated and improved.

[0053] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the concept of the present application. The above exemplary features are replaced with the technical features with similar functions disclosed in the present application (but not limited to) to form a technical solution.

Claims

1. An after-sales one-key ordering intelligent processing system, characterized in that: The system comprises: The user information processing module is used to obtain the user information dataset of the after-sales demand order, verify and filter the valid contact information, match the valid users to generate the address copy list, and convert it into regional information and latitude and longitude coordinates through the geocoding API; The product information merging module is used to extract the unique value of the product according to the after-sales demand order database, match the product information of the e-commerce platform and the product database data of the joint guarantee after-sales enterprise through association mapping, and merge the product information to obtain a product information list; After-sales demand description text analysis module is used to perform K-means clustering based on the after-sales demand description text of the product, generate repair operation labels and fault phenomenon labels for the text segments, and determine whether the user has made independent repair attempts on the product; The non-independent repair fault analysis module is used to match the fault code and the fault cause according to the fault phenomenon label of the after-sales demand description text that is judged as the user has not made an independent repair attempt on the product, and determine the repair measures to be obtained; The independent repair fault analysis module is used to annotate the text segments in time sequence according to the user repair operation and fault phenomenon labels of the after-sales demand description text that is judged to be the user's independent repair attempt on the product, determine the operation-fault chain and the root cause of the fault, and obtain the repair measures for the compound fault through the maintenance knowledge base; A maintenance guidance video generation module is used to determine after-sales demand orders that can be repaired by themselves according to the complexity of the fault repair measures or the complexity of the complex fault repair measures, generate a maintenance guidance video, and send it to the user of the after-sales demand order that can be repaired by themselves; The maintenance task order generation module is used to formulate regional matching rules according to the user's longitude and latitude, match the after-sales service area, and generate a maintenance task order; The independent repair fault analysis module is used to perform time sequence annotation on text segments according to the user repair operation and fault phenomenon label of the after-sales demand description text determined as the user's independent repair attempt on the product, determine the operation-fault chain and the root cause of the fault, and obtain the complex fault repair measures through the maintenance knowledge base, including: If the user attempts to repair the product independently, the text segments are time-sequence labeled according to the repair operation labels and fault phenomenon labels, predefined conjunctions and predefined time description words of the text segments to determine the order of the text segments; based on the time-sequence labeled text segments, a long short-term memory network model is used to perform contextual semantic analysis to determine the repair operation of each repair attempt and the resulting fault change phenomenon, identify the initial fault phenomenon and the fault change phenomenon caused by each repair attempt, and the fault change phenomenon includes new faults and fault conversions; obtain historical maintenance feedback text, and annotate the fault change phenomenon caused by each repair attempt and the corresponding fault phenomenon label, as well as the user's erroneous repair attempt operation, use a recurrent neural network to build a user erroneous repair attempt operation prediction model, predict the user's erroneous repair attempt operation, and determine the root cause of the fault change phenomenon; by calling the maintenance knowledge base API, use the root cause of the fault as the query condition, determine the maintenance measures for each fault change phenomenon, and obtain the composite fault maintenance measures and the complexity of the composite fault maintenance measures.

2. The system according to claim 1, wherein: The user information processing module is used to obtain the user information data set of the after-sales demand form, verify and screen the valid contact information, match the valid user to generate the address copy list, and convert it into regional information and latitude and longitude coordinates through the geocoding API. It is characterized by: According to the API interface provided by the e-commerce platform, the user information data set of the after-sales demand form is obtained, including the user's name, contact information, and address; by cleaning the data, removing redundant information, correcting format errors, and using regular expressions to standardize the user name, contact information, and address format, a standardized user information data set is obtained; by traversing the user information data set, the mobile phone number field of each user is extracted to obtain a list containing all user mobile phone numbers; the mobile phone number list is passed to the verification API to determine the availability and validity of each number, and the status of each number is returned through the API, including valid numbers, empty numbers, or invalid numbers; invalid numbers and empty numbers are eliminated, valid numbers are screened out, and a list of valid numbers is generated; by matching valid mobile phone numbers with the user information data set, all users with valid contact information and their demand forms are screened out; user information of invalid numbers is extracted, and a warning prompt is sent through the message system of the e-commerce platform to notify users to update their contact information; by traversing the after-sales demand forms of valid users, the address field is extracted to generate a text list containing all valid user addresses; by calling the geocoding API, the text list of valid user addresses is converted into regional information and latitude and longitude coordinates.

3. The system according to claim 1, wherein: The product information merging module is used to extract the unique value of the product according to the after-sales demand order database, match the product information of the e-commerce platform and the data of the joint guarantee after-sales enterprise commodity database through association mapping, and merge the product information to obtain a product information list, which is characterized by: By querying the after-sales demand order database, the corresponding product unique value in each after-sales demand order is extracted, and the unique value is the SKU number; using the association mapping technology, according to the product unique value, it is matched with the product database of the e-commerce platform to obtain the product information of the e-commerce platform, including the manufacturer, category, specification, brand and model information; through the joint guarantee after-sales enterprise product database, according to the product unique value, the products that meet the joint guarantee after-sales enterprise product database are determined, and the product information that meets the joint guarantee after-sales enterprise product database is obtained, including warranty status, repair history, fault code, after-sales support agreement; data standardization is performed on the product information from the e-commerce platform and the joint guarantee after-sales enterprise product database, and the field name and data format are unified; matching rules are defined based on the product unique value as the primary key, and the hash mapping algorithm is used to find the e-commerce platform product information and the joint guarantee after-sales enterprise product database product information with the same product unique value, and the matched product information in the e-commerce platform and the joint guarantee after-sales enterprise product database are merged to obtain a product information list.

4. The system according to claim 1, wherein: The after-sales demand description text analysis module is used to perform K-means clustering based on the after-sales demand description text of the product, generate repair operation labels and fault phenomenon labels for the text segments, and determine whether the user has made an independent repair attempt on the product: According to the after-sales demand form, obtain the after-sales demand description text of the product; use regular expressions to segment the after-sales demand description text to obtain a vocabulary list for each text segment; use a predefined stop word list to filter out irrelevant words to obtain a streamlined vocabulary list; use SpaCy to perform part-of-speech tagging on the filtered words, mark the part of speech of each word, and extract nouns, verbs, and adjectives; obtain the vocabulary index representation of each text segment according to the streamlined vocabulary list, and use the word segmenter of the BERT model to map each word to an index representation in the BERT vocabulary table; input the index representation of each text segment into the pre-trained BERT model for forward propagation to obtain the embedding vector of each word in the context; extract the embedding vector corresponding to the CLS tag output by the BERT model as the feature vector of the entire text segment, and obtain the text segment feature vector containing context information; according to the text feature vector, use the K-means clustering algorithm to cluster the text segment to obtain the cluster label of the text segment, and the cluster label includes the repair operation label and the fault phenomenon label; according to the repair operation label of the text segment, determine whether the user has made an independent repair attempt on the product.

5. The system according to claim 1, wherein: The non-independent repair fault analysis module is used to match the fault code and the fault cause according to the fault phenomenon label of the after-sales demand description text that the user has not attempted to independently repair the product, and determine the maintenance measures, and is characterized by: If the user does not attempt to repair the product independently, the fault mode text corresponding to the fault code is obtained through the joint warranty after-sales enterprise product database API; the fault phenomenon label and the fault mode text corresponding to the fault code are converted into feature vectors using the TF-IDF algorithm, and the similarity between the fault phenomenon label feature vector and the fault mode feature vector corresponding to the fault code is obtained using the Pearson correlation coefficient calculation method; the fault code corresponding to each fault phenomenon label is determined according to a preset similarity threshold, and if the similarity between the fault phenomenon label feature vector and the fault mode feature vector corresponding to the fault code is greater than the preset similarity threshold, the fault phenomenon label is judged to match the fault code, and the cause of the fault is obtained based on the fault code; by calling the maintenance knowledge base API, the cause of the fault is used as the query condition to obtain the fault repair measures and the complexity of the fault repair measures, which include simple, medium, and complex.

6. The system according to claim 1, wherein: The maintenance instruction video generation module is used to determine the after-sales demand form that can be repaired by oneself according to the complexity of the fault repair measures or the complexity of the complex fault repair measures, generate a maintenance instruction video, and send it to the user of the after-sales demand form that can be repaired by oneself, and is characterized by: Obtain after-sales demand orders with complexity levels of simple fault repair measures and complex fault repair measures, and mark the after-sales demand orders as after-sales demand orders that can be repaired by oneself; determine the repair parts according to the fault repair measures and complex fault repair measures; obtain parts images directly connected to the repair parts from the product part image database, mark the obtained parts images, and indicate the operation positions; determine the repair operation steps of the repair measures according to the maintenance manual, and arrange the parts images in sequence according to the maintenance operation steps to form a picture sequence that shows the operation process; generate text descriptions of each repair operation step according to the maintenance manual, and use text-to-speech conversion technology to convert the text descriptions of each repair operation step into voice; synchronize the marked pictures and the generated voice, and use the video editing software Adobe to synthesize the picture sequence and voice into a repair guidance video; send the repair guidance video to the user account and mobile phone of the after-sales demand order that can be repaired by oneself.

7. The system according to claim 1, wherein: The maintenance task order generation module is used to formulate area matching rules according to the user's longitude and latitude, match the after-sales service area, and generate a maintenance task order, and is characterized by: Based on the user's latitude and longitude information, regional matching rules are formulated to match the user's latitude and longitude information with the after-sales service area of ​​the after-sales company to determine the after-sales service area to which the user belongs; based on the fault phenomenon identification results and fault repair measures in the after-sales demand description text that cannot be repaired by oneself, a maintenance task order is generated, and the generated maintenance task order is assigned to the after-sales service area to which the user belongs. The maintenance task order content includes the fault labeling of the initial fault phenomenon and the fault change phenomenon caused by each repair attempt; the maintenance task assignment results are sent to the user account and mobile phone, and the maintenance time is arranged; through the real-time monitoring system, the execution progress of each maintenance task is tracked and monitored; the user feedback information after the maintenance task is completed is obtained to conduct quality evaluation and improvement of after-sales maintenance services.

Images