A training device implementation method and system

Abstract

The present invention provides a training device implementation method and system, which relates to the field of data processing, wherein the method includes: obtaining employee training information; processing the employee training information through a new module, saving it in a training material database, and returning storage address information; obtaining user input information; obtaining a matching link path through a link module based on the user input information; entering a model retrieval module through the matching link path to retrieve, obtain matching training materials, and determine the material storage address information according to the matching training materials; and browsing online through an online browsing module according to the material storage address information. The technical problem that the existing technology of carrying out targeted training for operating users of various business management systems is limited by multiple factors, is costly, and has an insignificant effect, and that the data retrieval function of financial institutions has a poor effect in terms of semantics and environmental matching is solved.

Description

Technical Field

[0001] The invention relates to the field of data processing, and in particular to a training device implementation method and system. Background Art

[0002] The business management system of financial institutions such as large banks mainly realizes the information management of various businesses of enterprises, promotes the informatization and systematization of various business processes, and brings economic benefits to enterprises. Generally, the head office or human resources department issues the operation manual of the relevant system, or carries out corresponding training to help users understand and become familiar with the various operations of the system. At present, the head office or human resources department carries out targeted training for the users of various business management systems, which is limited by many factors such as time, position, personnel, location, funds, system version, etc., with high costs and ineffective results. The data retrieval function of banks and other financial institutions generally uses technologies such as words, phrases, and key combinations for retrieval, which is less effective in semantics and environmental matching.

[0003] However, there are many factors in the existing technology that limit the implementation of targeted training, such as high cost and unclear effect. In addition, the data retrieval function of financial institutions has poor technical effects in terms of semantics and environmental matching. Summary of the invention

[0004] This application provides a training device implementation method and system to solve the technical problems in the prior art that there are many factors that restrict the implementation of targeted training, high cost, and ineffectiveness, and the data retrieval function of financial institutions has poor results in semantics and environmental matching. It has achieved the goal of reducing human training costs and improving training efficiency, which is conducive to the informatization, intelligence, openness, and transparency of system operation procedures and specifications. The use of neural network BERT technology can quickly and accurately complete semantic matching tasks, and users can quickly become familiar with the technical effects of the system by viewing all the training guidance materials of the system.

[0005] In view of the above problems, the present application provides a training device implementation method and system.

[0006] In a first aspect, the present application provides a method for implementing a training device, the method comprising: obtaining employee training information, processing the employee training information through a new module, saving the information into a training material database, and returning storage address information, obtaining user input information, obtaining a matching link path through a link module based on the user input information, entering a model retrieval module through the matching link path to perform retrieval, obtaining matching training materials, and determining material storage address information based on the matching training materials, obtaining the matching training materials based on the material storage address information, and browsing the matching training materials online through an online browsing module.

[0007] On the other hand, the present application provides a training device implementation system, which includes: a first acquisition unit, which is used to obtain employee training information; a first execution unit, which is used to process the employee training information through a new module, save it into a training material database, and return storage address information; a second acquisition unit, which is used to obtain user input information; a third acquisition unit, which is used to obtain a matching link path through a link module based on the user input information; a fourth acquisition unit, which is used to enter a model retrieval module through the matching link path to retrieve, obtain matching training materials, and determine the material storage address information based on the matching training materials; a second execution unit, which is used to browse the matching training materials online through an online browsing module based on the material storage address information.

[0008] In a third aspect, the present invention provides a training device implementation system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in the first aspect when executing the program.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] 1. Due to the adoption of collecting employee training information, after processing through the newly added module, it is saved in the training material database and the storage address information is returned; further collecting user input information, obtaining the matching link path through the link module and the user input information, entering the model retrieval module through the matching link path to retrieve and obtain matching training materials; and determining the material storage address information according to the matching training materials, and obtaining the matching training materials according to the material storage address information; the technical solution that employees use the online browsing module to browse the matching training materials online, this application provides a training device implementation method and system, thereby reducing the cost of human training and improving training efficiency, which is conducive to the informatization, intelligence, openness and transparency of system operation procedures and specifications, and uses the neural network BERT technology to quickly and accurately complete semantic matching tasks. Users can quickly become familiar with the technical effects of the system by viewing all the training guidance materials of the system.

[0011] 2. By adopting the methods of converting files to PDF, adding watermarks, and online browsing, the purpose of security and independence of training materials is achieved, which can reduce the demand for external plug-ins and improve the applicability of mobile terminals.

[0012] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A flowchart of a training device implementation method according to an embodiment of the present application;

[0014] Figure 2 A flowchart of a training device implementation method of the present application for processing employee training information by adding a new module;

[0015] Figure 3 A flowchart of a method for implementing a training device to obtain a retrieval model according to an embodiment of the present application;

[0016] Figure 4 A flowchart of a method for implementing a training device in an embodiment of the present application for obtaining a BERT model;

[0017] Figure 5 A schematic diagram of a flow chart of a training device implementation method according to a preset rule for reading and processing data in accordance with an embodiment of the present application;

[0018] Figure 6 This is a structural diagram of a training device implementation system according to an embodiment of the present application;

[0019] Figure 7 It is a schematic diagram of the structure of an exemplary electronic device according to an embodiment of the present application.

[0020] Explanation of the reference numerals: first obtaining unit 11 , first executing unit 12 , second obtaining unit 13 , third obtaining unit 14 , fourth obtaining unit 15 , second executing unit 16 , electronic device 300 , memory 301 , processor 302 , communication interface 303 , bus architecture 304 . DETAILED DESCRIPTION

[0021] This application provides a training device implementation method and system to solve the technical problems in the prior art that there are many factors that restrict the implementation of targeted training, high cost, and ineffectiveness, and the data retrieval function of financial institutions has poor results in semantics and environmental matching. It has achieved the goal of reducing human training costs and improving training efficiency, which is conducive to the informatization, intelligence, openness, and transparency of system operation procedures and specifications. The use of neural network BERT technology can quickly and accurately complete semantic matching tasks, and users can quickly become familiar with the technical effects of the system by viewing all the training guidance materials of the system.

[0022] The acquisition, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0023] The business management system of financial institutions such as large banks mainly realizes the information management of various businesses of enterprises, promotes the informatization and systematization of various business processes, and brings economic benefits to enterprises. Generally, the head office or human resources department issues the operation manual of the relevant system, or carries out corresponding training to help users understand and become familiar with the various operations of the system. At present, the data retrieval function of banks and other financial institutions generally uses technologies such as words, phrases, and key combinations for retrieval. There are many factors in the existing technology that restrict the development of targeted training, which is costly and ineffective. In addition, the data retrieval function of financial institutions has poor technical problems in terms of semantics and environmental matching.

[0024] In response to the above technical problems, the overall idea of ​​the technical solution provided by this application is as follows:

[0025] The present application provides a method for implementing a training device, the method comprising: collecting employee training information, processing it through a new module, saving it in a training material database, and returning storage address information; further collecting user input information, obtaining a matching link path through a link module and user input information, entering a model retrieval module through the matching link path to perform retrieval and obtain matching training materials; and determining material storage address information based on the matching training materials, and obtaining matching training materials based on the material storage address information; employees use an online browsing module to browse the matching training materials online.

[0026] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically described below in conjunction with the drawings in the specification.

[0027] Embodiment 1

[0028] like Figure 1 As shown, an embodiment of the present application provides a method for implementing a training device, wherein the method is applied to a business management system, the business management system is embeddedly connected to the training device, the training device includes a new module, a link module, a model retrieval module, and an online browsing module, and the method includes:

[0029] S100: Obtain employee training information;

[0030] S200: After processing the employee training information through the newly added module, save it into the training material database and return the storage address information;

[0031] Specifically, embedded generally refers to embedded systems. Embedded systems are application-centric, based on computer technology, with tailorable software and hardware, and are suitable for special-purpose computer systems with strict requirements on functions, reliability, cost, volume, and power consumption. They have strong flexibility and customizability and can be used for a variety of purposes. In other words, embedded means porting an operating system to an existing hardware platform and doing corresponding application development. The training device is embeddedly connected to the business management system, which can be understood as carrying out corresponding training application development on the hardware platform of the existing business management system. The training device includes a new module, a link module, a model retrieval module, and an online browsing module. The new module is used to process the function name (including menu name, button name, etc.), function path, function description (introduction), and uploaded guidance attachments entered by the information publisher.

[0032] Employee training information includes guidance document information and other descriptive texts for employee training. Other descriptive texts include, but are not limited to, function names, function paths, and function descriptions entered by the information publisher. The function names include menu names, button names, etc., such as the name of a training program. The descriptive information includes a brief introduction to the training program, a topic introduction, etc. The function path includes a directory, an index, etc.

[0033] Employee training information is collected and uploaded to the new module by the information publisher, who is generally a demand analyst, system designer, head office manager, etc. In order to ensure the safety and independence of training materials, the employee training information is processed using the new module. Processing includes but is not limited to methods such as conversion to PDF and adding watermarks. After processing, the processed attachments are uploaded to the platform and stored in the training material database, and the storage address is returned. The information publisher will also undertake tasks such as information modification and deletion to modify and delete the stored information. From the perspective of information release, it greatly reduces the cost of human training, improves training efficiency, and is conducive to the informatization, intelligence, openness, and transparency of system operation procedures and specifications.

[0034] S300: Obtaining user input information;

[0035] S400: Obtaining a matching link path through a link module based on the user input information;

[0036] Specifically, the user is the information receiver, that is, the general user of the system, who is the main training audience, such as employees of financial institutions. The user searches for training documents by entering information, and the user-entered information includes but is not limited to search keywords and phrases, user-clicked menu names, etc.

[0037] The link module is an important channel connecting general users and training materials. There are two ways for users to enter the training guidance file device: one is to directly enter the training guidance search page and enter the keywords and phrases of interest to search; the other is to enter the business operation page and click on the guidance operation related link specially set for each operation menu. The system automatically captures the current menu name and jumps to the training guidance page. The search keyword jumps to the menu name by default.

[0038] Based on the user input information, including but not limited to the search keywords entered by the user, the business function menu link clicked by the user, etc., a matching link path is obtained through the two types of access to the training guidance file device of the connection module. The user is supported to enter the learning module in multiple ways, and the matching link path can enable the user to enter the matching learning module quickly and accurately.

[0039] S500: Entering the model retrieval module through the matching link path to perform retrieval, obtaining matching training materials, and determining material storage address information according to the matching training materials;

[0040] S600: According to the material storage address information, the matching training material is browsed online through an online browsing module.

[0041] Specifically, the model retrieval module is entered through the matching link path for retrieval. The model retrieval module is a key module for users to accurately and quickly find corresponding results through key words and sentences. The retrieval model used is preferably a BERT model (Bidirectional Encoder Representations from Transformers). Compared with the RNN model (Recurrent Neural Networks, cyclic neural network model) and the LSTM model (Long short-term Memory, long short-term memory neural network model), the BERT model can perform concurrent execution and extract the relationship features of words in sentences at the same time, and can extract relationship features at multiple different levels, thereby more comprehensively reflecting the semantics of sentences. Compared with the Word2Vec model, BERT can obtain the meaning of words according to the context of sentences, thereby avoiding ambiguity. The Word2Vec model is a model for learning semantic knowledge in an unsupervised manner from a large amount of text predictions, and is widely used in natural language processing.

[0042] After the model retrieval module, the matching training materials are retrieved, and when the training materials are stored in the local training material database, there will be a corresponding storage address. After obtaining the storage address information of the materials based on the matching training materials, the system can use the HTTP protocol to obtain the attachments through the file storage address link. Then, for different attachment formats, different display forms are used for users to browse. Attachment formats include but are not limited to videos such as MP4 format and documents such as PDF format. For video files, it is preferred to use HTML5 technology, and as long as the client browser supports HTML5, it can support online video playback. For document files in PDF format, it is preferred to call the PDF component, obtain the file through the data address, and then render the PDF to each page, so that the display file can be seen directly on the web page.

[0043] The online browsing module provides users with the functions of online video playback and online document preview. Through the online browsing module, users obtain corresponding display modes, including online video playback and online document preview. From the perspective of information reception, it is simple and easy to operate. Users can view all training guidance materials of the system through the present invention, so as to quickly become familiar with the system.

[0044] Further, such as Figure 2 As shown, after the employee training information is processed by the new module, it is saved in the training material database and the storage address information is returned. Step S200 includes:

[0045] S210: Obtaining function name, function path, and description information according to the employee training information;

[0046] S220: Get employee training information format;

[0047] S230: Matching processing logic according to the employee training information format;

[0048] S240: After processing the employee training information according to the matching processing logic, save it into the training material database to obtain the storage address information;

[0049] S250: Return the storage address information, and save the function name, function path, description information, and storage address information locally.

[0050] Specifically, the employee training information includes guidance document information and other descriptive texts for employee training. Employee training guidance documents include training documents, training videos, etc. Other descriptive texts include but are not limited to function names, function paths, and function descriptions entered by the information publisher. Therefore, according to the employee training information, the function name, function path, and description information are obtained. The function name includes menu name, button name, etc., such as the name of the training project. The descriptive information includes a brief introduction to the training project, a theme introduction, etc. The function path includes a directory, an index, etc.

[0051] According to the employee training information, the format information of the employee guidance attachments is obtained. Different attachment processing logics are used for attachments of different formats. For Office documents (such as Word, Excel, PowerPoint), the COM interface provided by Office is called to save the document as PDF; for plain text documents, technical components such as iTextSharp are called to specify fonts and then converted into PDF; from the perspective of information security, watermarks are added to video documents (MP4, WebM, HLS, etc.) and uploaded or converted PDF documents.

[0052] After processing the employee's instruction attachment according to the matching processing logic, it is saved in the training material database, and the storage address information of the instruction attachment is obtained. Finally, the storage address information is returned, and the function name, function path, description information, and storage address information are saved in the local database. This can improve the training efficiency and achieve the technical effect of informatization, intelligence, and openness of system operation procedures and specifications.

[0053] Furthermore, the matching link path is obtained through a link module based on the user input information, and step S400 includes:

[0054] S410: Determine whether to enter the training guidance search page based on the user input information;

[0055] S420: when entering the training guidance search page, obtaining search keywords;

[0056] S430: Based on the search keyword, a matching link path is obtained, where the matching link path is to enter the model search module for search through the search keyword.

[0057] Specifically, multiple access channels are provided for users to obtain training guidance documents. In addition to the unified search distribution center, users can quickly jump to the corresponding operation guidance material page on the general business function menu page. Therefore, it is necessary to judge the user's input information to determine whether the user chooses to enter the training guidance search page or the business function menu page. When it is determined that the user enters the training guidance search page, the keywords searched by the user are collected, and a matching link path is obtained based on the search keywords. The matching link path is to enter the model search module for retrieval through the search keywords. It can achieve the technical effect of effectively seeking solutions through keyword retrieval in a timely manner when the user encounters an operation problem.

[0058] Further, such as Figure 3 As shown, the embodiment of the present application includes:

[0059] S510: Obtain a data set;

[0060] S520: reading the data set according to a preset rule, and converting the read data to obtain an input feature value set;

[0061] S530: Determine a training feature value based on the input feature value set, and perform model training using the training feature value, wherein the model is a BERT model;

[0062] S540: Calculate the loss function value of the output result of the BERT model using the cross entropy loss function until the loss function value meets the training requirements, thereby obtaining a BERT training model;

[0063] S550: Performing model adjustment on the trained BERT training model by using the Adam algorithm to determine an optimized model, and obtaining a retrieval model based on the optimized model;

[0064] S560: Input the user input information into the retrieval model to obtain the output result of the retrieval model, wherein the output result includes matching training materials.

[0065] Further, such as Figure 4 As shown, the data set is read according to a preset rule, and the read data is converted, and step S520 also includes:

[0066] S521: Read data according to the preset rules to obtain data information, wherein the preset rules include removing trailing spaces, separating tags from sentence content, and adding special symbols and separator marks;

[0067] S522: Convert the data information into digital information;

[0068] S523: Setting fixed length information;

[0069] S524: Determine whether the length of the digital information satisfies the fixed length information;

[0070] S525: When the length is less than the fixed length information, the length is padded according to the fixed length information;

[0071] S526: When the length exceeds the fixed length information, the length is truncated according to the fixed length information.

[0072] Specifically, a data set is obtained based on local historical data, and data sampling is performed. The data set is further extracted according to preset rules, wherein the preset rules are set according to requirements, and the preset rules include removing trailing spaces, separating tags from sentence content, adding special symbols, and separator marks. When reading data, first remove the leading and trailing spaces, separate tags from sentence content, add a special symbol [CLS] in front of each input example, and use a special mark [SEP] as a separator (for example, to separate questions / answers). Among them, [CLS] indicates that the feature is used for the classification model. [SEP] represents a sentence symbol, which is used to disconnect two sentences in the input corpus. Then, the data is converted into digital information, and fixed length information is set to determine whether the length of the digital information meets the set fixed length information. If the length is insufficient, 0 is used as padding, and if the length exceeds the fixed length, it is truncated according to the fixed length information. Since the BERT model can be connected to the context, in the Chinese BERT model, Chinese word segmentation is based on characters rather than words, so there is no need for additional word segmentation operations.

[0073] The converted feature values ​​can be used as input to obtain the input feature value set, which is used for model training and testing. The input feature value set is divided into a training set, a validation set, and a test set according to a certain ratio, and the division ratio is preferably 6:2:2.

[0074] The pre-trained BERT model can be fine-tuned with an additional output layer to create advanced models for a wide range of tasks (such as question answering and language reasoning) without making substantial modifications to the task-specific architecture. The fine-tuning model is preferably based on the Chinese pre-trained model bert-base-chinese officially released by Google. It is recommended to perform fine-tuning training based on this model. During the fine-tuning process, all parameters are fine-tuned. The BERT model is trained using the training feature values. The model training uses the cross-entropy loss function, which not only allows for easy derivative calculations and fast convergence, but also can measure the effect of the model well. In the case of binary division, there are only two cases where the model needs to predict the final result. The probabilities predicted for each category are p and 1-p. At this time, the cross-entropy expression is:

[0075]

[0076] Where: y i Indicates the label of sample i, the positive class is 1 and the negative class is 0; p i It represents the probability that sample i is predicted to be a positive class.

[0077] The cross entropy loss function is used to calculate the loss function value until convergence to obtain the BERT training model. The model training algorithm uses the Adam algorithm (Adaptive moment Estimation) dedicated to BERT. Set appropriate learning rate, exponential decay rate of first-order moment estimation, and exponential decay rate of second-order moment estimation. Adam relies on momentum and adaptive learning rate to accelerate convergence. The model is adjusted by the Adam algorithm to determine the optimization model, and the retrieval model is obtained based on the optimization model. User retrieval stage. Load the model, and then convert the search keywords entered by the user into feature values ​​according to the above-mentioned feature extraction method, and obtain the output results through model retrieval, and the output results include matching training materials. Complete data reading, feature conversion, confirm the loss function, fine-tune the model, and perform model training through the model training algorithm Adam algorithm to obtain a scientific and highly applicable retrieval model.

[0078] Further, such as Figure 5 As shown, the embodiment of the present application also includes:

[0079] S561: Obtain training data set;

[0080] S562: performing joint adjustment on the training data set to obtain a joint adjustment feature value, wherein the joint adjustment is adjusting the left and right contexts;

[0081] S563: Based on the training data set and the joint adjustment feature value, obtain a first training set and a second training set, wherein the first training set is unlabeled data and the second training set is labeled data;

[0082] S564: Perform unsupervised model training using the first training set to obtain a first language model;

[0083] S565: Adding a neural network layer based on the first language model, and performing supervised model training on the first language model with the added neural network layer using the second training set to obtain the BERT model;

[0084] S566: During the unsupervised model training and the supervised model training, an intermediate model result is obtained, and the intermediate model result is added to the training features for model training.

[0085] Specifically, based on the training data set obtained from big data, when learning features, the BERT model jointly adjusts the left and right contexts through bidirectional training to obtain jointly adjusted feature values. At the same time, the intermediate results of the language model are used as additional features and introduced into the model of the original task, thereby achieving the effect of effectively learning context and context.

[0086] Based on the training set and the joint adjustment feature value, a first training set and a second training set are generated. The first training set is unlabeled data, and the second training set is labeled data. Two labels are defined for labeling, which respectively represent retrieval matching (label=1) and retrieval mismatch (label=0), and the label and the text are separated by a Tab symbol.

[0087] BERT feature training mainly includes two parts: first, unsupervised training of the language model on the first training set, and obtaining the first language model after training; then supervised training of the model using the labeled second training set, adding a small number of neural network layers on the basis of the first language model to complete specific tasks such as sequence labeling and classification to obtain the BERT model. As one of the most advanced natural language processing tools currently available, the BERT model can play a good technical role in connecting context, understanding semantics, and predicting results.

[0088] Furthermore, the matching training materials are browsed online through an online browsing module according to the material storage address information, and step S600 further includes:

[0089] S610: Acquire the matching training data according to the data storage address information;

[0090] S620: Determine whether the matching training material is in video format;

[0091] S630: when the matching training material is in the video format, obtaining a preset player, and playing the matched training material online through the preset player;

[0092] S640: When the matching training material is not in the video format, obtain a preset converter, and use the preset converter to convert the format of the matching training material for online browsing.

[0093] Specifically, according to the data storage address information, the system uses the HTTP protocol to obtain the attachment through the file storage address link, the attachment includes the matching training data, and then uses different display forms for different matching training data formats for users to browse. It is determined whether the matching training data is in video format. If it is in video format, a preset network video player is obtained to play the video online.

[0094] Before the emergence of HTML5 technology, most web video playback was achieved through browser plug-ins such as Adobe Flash, which required customers to ensure that the corresponding components were installed before watching the video. In order to be independent of the device and reduce the need for external plug-ins (such as Flash, which has stopped updating), using a video network video player no longer relies on specific third-party technology. As long as the client browser supports HTML5, it can support online video playback. Currently, most mainstream browsers already support HTML5, such as IE, Firefox, Google, etc. In addition to the computer side, the video player is also well suited for the mobile side.

[0095] When the matching training materials are not in the video format, in other words, when the matching training materials are in non-video formats such as Office documents (Word, Excel, PowerPoint), a preset converter is obtained to convert the format of the matching training materials to enable online browsing. For example, for PDF files, the browser does not need to install an additional PDF reader plug-in, call the PDF component, obtain the file through the data address, and then render the PDF to each page, so that the display file can be seen directly on the web page. The use of technical means such as online file browsing can reduce the need for external plug-ins and improve the applicability of mobile terminals.

[0096] Embodiment 2

[0097] Based on the same inventive concept as the method for implementing a training device in the aforementioned embodiment, Figure 6 As shown, an embodiment of the present application provides a training device implementation system, wherein the system includes:

[0098] A first obtaining unit 11, wherein the first obtaining unit 11 is used to obtain employee training information;

[0099] A first execution unit 12, which is used to process the employee training information through a new module, save it into a training material database, and return storage address information;

[0100] A second obtaining unit 13, the second obtaining unit 13 is used to obtain user input information;

[0101] A third obtaining unit 14, the third obtaining unit 14 is used to obtain a matching link path through a link module based on the user input information;

[0102] A fourth obtaining unit 15, the fourth obtaining unit 15 is used to enter the model retrieval module through the matching link path to perform retrieval, obtain matching training data, and determine data storage address information according to the matching training data;

[0103] The second execution unit 16 is used for browsing the matching training materials online through an online browsing module according to the material storage address information.

[0104] Furthermore, the system comprises:

[0105] A fifth obtaining unit, the fifth obtaining unit is used to obtain a function name, a function path, and description information according to the employee training information;

[0106] A sixth obtaining unit, the sixth obtaining unit is used to obtain the employee training information format;

[0107] A third execution unit, the third execution unit is used to match the processing logic according to the employee training information format;

[0108] A seventh obtaining unit, the seventh obtaining unit is used to process the employee training information according to the matching processing logic, save it into the training material database, and obtain the storage address information;

[0109] The fourth execution unit is used to return the storage address information and save the function name, function path, description information, and storage address information locally.

[0110] Furthermore, the system comprises:

[0111] A first judgment unit, the first judgment unit is used to judge whether to enter the training guidance search page based on the user input information;

[0112] An eighth obtaining unit, the eighth obtaining unit is used to obtain a search keyword when entering the training guidance search page;

[0113] A ninth obtaining unit, the ninth obtaining unit is used to obtain a matching link path based on the search keyword, wherein the matching link path is to enter the model retrieval module for retrieval through the search keyword.

[0114] Furthermore, the system comprises:

[0115] A fifth execution unit, the fifth execution unit is used to enter the business operation page when not entering the training guidance search page;

[0116] a tenth obtaining unit, the tenth obtaining unit being used to obtain a guidance operation link based on the business operation page;

[0117] an eleventh obtaining unit, the eleventh obtaining unit being used to obtain a menu name according to the guiding operation link;

[0118] The twelfth obtaining unit is used to jump into the training guidance search page based on the menu name, wherein the search keyword is the menu name, and based on the menu name, a matching link path is obtained.

[0119] Furthermore, the system comprises:

[0120] a thirteenth obtaining unit, the thirteenth obtaining unit being used to obtain a data set;

[0121] A fourteenth obtaining unit, the fourteenth obtaining unit is used to read data from the data set according to a preset rule, and convert the read data to obtain an input feature value set;

[0122] A sixth execution unit, the sixth execution unit is used to determine a training feature value based on the input feature value set, and perform model training using the training feature value, wherein the model is a BERT model;

[0123] A seventh execution unit, wherein the seventh execution unit is used to calculate a loss function value on an output result of the BERT model using a cross entropy loss function until the loss function value meets the training requirements, thereby obtaining a BERT training model;

[0124] An eighth execution unit, the eighth execution unit is used to perform model adjustment on the trained BERT training model through the Adam algorithm, determine an optimization model, and obtain a retrieval model based on the optimization model;

[0125] A fifteenth obtaining unit is used to input the user input information into the retrieval model to obtain an output result of the retrieval model, wherein the output result includes matching training materials.

[0126] Furthermore, the system comprises:

[0127] A sixteenth obtaining unit, the sixteenth obtaining unit is used to obtain a training data set;

[0128] A seventeenth obtaining unit, the seventeenth obtaining unit is used to perform joint adjustment on the training data set to obtain a joint adjustment feature value, wherein the joint adjustment is to adjust the left and right contexts;

[0129] An eighteenth obtaining unit, the eighteenth obtaining unit being used to obtain a first training set and a second training set based on the training data set and the joint adjustment feature value, wherein the first training set is unlabeled data and the second training set is labeled data;

[0130] a nineteenth obtaining unit, the nineteenth obtaining unit being configured to perform unsupervised model training using the first training set to obtain a first language model;

[0131] a twentieth obtaining unit, wherein the twentieth obtaining unit is used to add a neural network layer based on the first language model, and perform supervised model training on the first language model with the added neural network layer by using the second training set, so as to obtain the BERT model;

[0132] The twenty-first acquisition unit is used to obtain the model intermediate result during the unsupervised model training and the supervised model training, and add the model intermediate result to the training feature for model training.

[0133] Furthermore, the system comprises:

[0134] A twenty-second obtaining unit, the twenty-second obtaining unit is used to read the data according to the preset rules to obtain data information, the preset rules including removing trailing spaces, separating tags from sentence content, adding special symbols, and separator marks;

[0135] a ninth execution unit, the ninth execution unit being configured to convert the data information into digital information;

[0136] a tenth execution unit, the tenth execution unit being used to set fixed length information;

[0137] A second judging unit, the second judging unit is used to judge whether the length of the digital information meets the fixed length information;

[0138] an eleventh execution unit, the eleventh execution unit being configured to pad the length according to the fixed length information when the length is less than the fixed length information;

[0139] A twelfth execution unit, the twelfth execution unit is used to truncate the length according to the fixed length information when the length exceeds the fixed length information.

[0140] Furthermore, the system comprises:

[0141] A twenty-third obtaining unit, the twenty-third obtaining unit is used to obtain the matching training data according to the data storage address information;

[0142] a third judging unit, the third judging unit being used to judge whether the matching training material is in a video format;

[0143] A twenty-fourth obtaining unit, the twenty-fourth obtaining unit is used to obtain a preset player when the matching training material is in the video format, and play it online through the preset player;

[0144] The twenty-fifth obtaining unit is used to obtain a preset converter when the matching training material is not in the video format, and perform format conversion on the matching training material through the preset converter for online browsing.

[0145] Exemplary Electronic Devices

[0146] Reference below Figure 7 To describe the electronic device of the embodiment of the present application.

[0147] Based on the same inventive concept as a training device implementation method in the aforementioned embodiment, an embodiment of the present application also provides a training device implementation system, including: a processor, the processor is coupled to a memory, the memory is used to store a program, when the program is executed by the processor, the system executes any one of the methods described in the first aspect.

[0148] The electronic device 300 includes: a processor 302, a communication interface 303, and a memory 301. Optionally, the electronic device 300 may also include a bus architecture 304. The communication interface 303, the processor 302, and the memory 301 may be interconnected via the bus architecture 304; the bus architecture 304 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The bus architecture 304 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0149] The processor 302 may be a CPU, a microprocessor, an ASIC, or one or more integrated circuits for controlling the execution of the program of the present application.

[0150] The communication interface 303 uses any transceiver-like device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), wired access networks, etc.

[0151] The memory 301 may be a ROM or other type of static storage device that can store static information and instructions, a RAM or other type of dynamic storage device that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory may exist independently and be connected to the processor via the bus architecture 304. The memory may also be integrated with the processor.

[0152] The memory 301 is used to store computer-executable instructions for executing the solution of the present application, and the execution is controlled by the processor 302. The processor 302 is used to execute the computer-executable instructions stored in the memory 301, thereby realizing a training device implementation method provided in the above embodiment of the present application.

[0153] Optionally, the computer-executable instructions in the embodiments of the present application may also be referred to as application code, which is not specifically limited in the embodiments of the present application.

[0154] An embodiment of the present application provides a method for implementing a training device, wherein the method includes: collecting employee training information, processing it through a new module, saving it in a training material database, and returning storage address information; further collecting user input information, obtaining a matching link path through a link module and user input information, entering a model retrieval module through the matching link path to perform retrieval and obtain matching training materials; and determining material storage address information based on the matching training materials, and obtaining matching training materials based on the material storage address information; employees use an online browsing module to browse the matching training materials online.

[0155] Those of ordinary skill in the art will appreciate that the various digital numbers such as the first and second involved in the present application are only for the convenience of description, and are not used to limit the scope of the embodiments of the present application, nor do they represent the order of precedence. "And / or" describes the association relationship of the associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects before and after are in an "or" relationship. "At least one" refers to one or more. At least two refers to two or more. "At least one", "any one" or similar expressions refer to any combination of these items, including any combination of single items (individuals) or plural items (individuals). For example, at least one item (individual, kind) of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0156] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server, data center, etc. that includes one or more available media integrated. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)), etc.

[0157] The various illustrative logic units and circuits described in the embodiments of the present application can be implemented or operated by a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic, a discrete hardware component, or the design of any combination of the above functions. The general-purpose processor can be a microprocessor, and optionally, the general-purpose processor can also be any traditional processor, controller, microcontroller or state machine. The processor can also be implemented by a combination of computing devices, such as a digital signal processor and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a digital signal processor core, or any other similar configuration to implement.

[0158] The steps of the method or algorithm described in the embodiments of the present application can be directly embedded in hardware, a software unit executed by a processor, or a combination of the two. The software unit can be stored in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium in the art. Exemplarily, the storage medium can be connected to the processor so that the processor can read information from the storage medium and can write information to the storage medium. Optionally, the storage medium can also be integrated into the processor. The processor and the storage medium can be arranged in an ASIC, and the ASIC can be arranged in a terminal. Optionally, the processor and the storage medium can also be arranged in different components in the terminal. These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are performed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0159] Although the present application has been described in conjunction with specific features and embodiments thereof, it is apparent that various modifications and combinations may be made thereto without departing from the spirit and scope of the present application. Accordingly, this specification and the drawings are merely exemplary illustrations of the present application as defined by the appended claims, and are deemed to have covered any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application is intended to include these modifications and variations.

Claims

1. A training device implementation method, characterized in that: The method is applied to a business management system, the business management system is embeddedly connected with the training device, the training device includes a new module, a link module, a model retrieval module, and an online browsing module, and the method includes: Obtain employee training information; After processing the employee training information through the newly added module, it is saved in the training material database and the storage address information is returned; Get user input information; Obtaining a matching link path through a link module based on the user input information; Entering the model retrieval module through the matching link path to perform retrieval, obtaining matching training data, and determining data storage address information according to the matching training data; According to the data storage address information, the matching training data is browsed online through an online browsing module; After the employee training information is processed by the new module, it is saved in the training material database and the storage address information is returned, including: According to the employee training information, obtain function name, function path, and description information; Obtain employee training information format; Matching processing logic according to the employee training information format; After processing the employee training information according to the matching processing logic, the information is saved in the training material database to obtain the storage address information; Return the storage address information, and save the function name, function path, description information, and storage address information locally; The method further comprises: Get the dataset; Reading the data set according to a preset rule, and converting the read data to obtain an input feature value set; Based on the input feature value set, determine a training feature value, and use the training feature value to perform model training, wherein the model is a BERT model; Using the cross entropy loss function to calculate the loss function value of the output result of the BERT model until the loss function value meets the training requirements, thereby obtaining a BERT training model; The trained BERT training model is adjusted by using the Adam algorithm to determine an optimized model, and a retrieval model is obtained based on the optimized model; Inputting the user input information into the retrieval model to obtain an output result of the retrieval model, wherein the output result includes matching training materials; The method further comprises: Get a training dataset; Performing joint adjustment on the training data set to obtain a joint adjustment feature value, wherein the joint adjustment is adjusting the left and right contexts; Based on the training data set and the joint adjustment feature value, a first training set and a second training set are obtained, wherein the first training set is unlabeled data and the second training set is labeled data; Using the first training set to perform unsupervised model training to obtain a first language model; Adding a neural network layer based on the first language model, and performing supervised model training on the first language model with the added neural network layer using the second training set to obtain the BERT model; Wherein, during the unsupervised model training and the supervised model training, an intermediate model result is obtained, and the intermediate model result is added to the training feature for model training; The step of reading the data set according to a preset rule and converting the read data includes: Read the data according to the preset rules to obtain data information, wherein the preset rules include removing trailing spaces, separating tags from sentence content, adding special symbols, and separator marks; Converting the data information into digital information; Set fixed length information; Determining whether the length of the digital information satisfies the fixed length information; When the length is less than the fixed length information, the length is padded according to the fixed length information; When the length exceeds the fixed length information, the length is truncated according to the fixed length information.

2. The method according to claim 1, characterized in that The obtaining of a matching link path through a link module based on the user input information includes: Determining whether to enter the training guidance search page based on the user input information; When entering the training guide search page, obtaining search keywords; Based on the search keyword, a matching link path is obtained, and the matching link path is to enter the model search module through the search keyword for search.

3. The method according to claim 2, characterized in that After determining whether to enter the training guidance search page based on the user input information, the method further includes: When not entering the training guidance search page, entering the business operation page; Based on the business operation page, obtain a guidance operation link; According to the guide operation link, obtain the menu name; Based on the menu name, jump to the training guidance search page, wherein the search keyword is the menu name, and based on the menu name, obtain the matching link path.

4. The method according to claim 1, characterized in that The step of browsing the matching training materials online through an online browsing module according to the material storage address information includes: Acquiring the matching training data according to the data storage address information; Determining whether the matching training material is in video format; When the matching training material is in the video format, obtaining a preset player, and playing the matching training material online through the preset player; When the matching training material is not in the video format, a preset converter is obtained, and the matching training material is converted into a format by the preset converter for online browsing.

5. A training device implementation system, characterized in that: The system comprises: A first obtaining unit, the first obtaining unit is used to obtain employee training information; A first execution unit, the first execution unit is used to process the employee training information through a new module, save it into a training material database, and return storage address information; A second obtaining unit, the second obtaining unit is used to obtain user input information; A third obtaining unit, the third obtaining unit is used to obtain a matching link path through a link module based on the user input information; A fourth obtaining unit, the fourth obtaining unit is used to enter the model retrieval module through the matching link path to perform retrieval, obtain matching training data, and determine data storage address information according to the matching training data; A second execution unit, the second execution unit is used to browse the matching training materials online through an online browsing module according to the material storage address information; After the employee training information is processed by the new module, it is saved in the training material database and the storage address information is returned, including: According to the employee training information, obtain function name, function path, and description information; Obtain employee training information format; Matching processing logic according to the employee training information format; After processing the employee training information according to the matching processing logic, the information is saved in the training material database to obtain the storage address information; Return the storage address information, and save the function name, function path, description information, and storage address information locally; The system further comprises: Get the dataset; Reading the data set according to a preset rule, and converting the read data to obtain an input feature value set; Based on the input feature value set, determine a training feature value, and use the training feature value to perform model training, wherein the model is a BERT model; Using the cross entropy loss function to calculate the loss function value of the output result of the BERT model until the loss function value meets the training requirements, thereby obtaining a BERT training model; The trained BERT training model is adjusted by using the Adam algorithm to determine an optimized model, and a retrieval model is obtained based on the optimized model; Inputting the user input information into the retrieval model to obtain an output result of the retrieval model, wherein the output result includes matching training materials; The system further comprises: Get a training dataset; Performing joint adjustment on the training data set to obtain a joint adjustment feature value, wherein the joint adjustment is adjusting the left and right contexts; Based on the training data set and the joint adjustment feature value, a first training set and a second training set are obtained, wherein the first training set is unlabeled data and the second training set is labeled data; Using the first training set to perform unsupervised model training to obtain a first language model; Adding a neural network layer based on the first language model, and performing supervised model training on the first language model with the added neural network layer using the second training set to obtain the BERT model; Wherein, during the unsupervised model training and the supervised model training, an intermediate model result is obtained, and the intermediate model result is added to the training feature for model training; The step of reading the data set according to a preset rule and converting the read data includes: Read the data according to the preset rules to obtain data information, wherein the preset rules include removing trailing spaces, separating tags from sentence content, adding special symbols, and separator marks; Converting the data information into digital information; Set fixed length information; Determining whether the length of the digital information satisfies the fixed length information; When the length is less than the fixed length information, the length is padded according to the fixed length information; When the length exceeds the fixed length information, the length is truncated according to the fixed length information.

6. A computer program product comprising a computer program and / or instructions, characterized in that: When the computer program and / or the instructions are executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images