Training data acquisition method and device for virtual training effect evaluation

Abstract

The invention provides a training data acquisition method and device for virtual training effect evaluation, and the method comprises the steps: caching training process data when the data transmission of the training process data of a virtual training system is carried out, and deleting the cached training process data when the data transmission is completed; performing data cleaning on the transmitted training process data to obtain target training data; the method comprises the following steps: collecting target training data, performing aggregation processing on unstructured data in the target training data to obtain collected training process data, performing data reorganization by adopting a data form of a data cube, and performing virtual training effect evaluation on the reorganized training process data. According to the virtual training system, the defects that in the prior art, training process data collection of the virtual training system depends on manual operation, the training process data is numerous, the system load requirement is high, the data collection efficiency is low, and the real-time training evaluation requirement is difficult to meet are overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a training data collection method and device for virtual training effect evaluation. Background Art

[0002] The virtual training system is mainly based on digital technology, relying on computers to build a virtual training environment, simulating real training scenes and situations, so that trainees can obtain training effects similar to actual operations. Training effect evaluation is the review and estimation of training effects in the virtual training system, and is one of the important links in training. Accurate and scientific training evaluation methods help to improve the effectiveness and reliability of virtual training. The training evaluation process mainly evaluates the training effect based on the training data of the trainees. Real-time and reliable training data collection can effectively support training evaluation.

[0003] At present, most virtual training systems are designed with the function of evaluating training effects. The data sources of the training process used for evaluation mainly rely on manual data collection or expert scoring recommendations, but this is inefficient, time-consuming and labor-intensive. In addition, some virtual training systems use traditional relational databases such as MySQL and Oracle to record training data. Due to the large amount of training process data, the system load requirements are high and it is difficult to meet the real-time training evaluation needs. Summary of the invention

[0004] The present invention provides a training data collection method and device for virtual training effect evaluation, which is used to solve the technical problem that the training process data collection of the virtual training system in the prior art relies on manual operation, and the training process data is large, which places high requirements on the system load, resulting in low data collection efficiency and difficulty in meeting real-time training evaluation needs.

[0005] The present invention provides a training data collection method for virtual training effect evaluation, the method comprising the following steps: When data transmission is performed on the training process data of the virtual training system, the training process data is cached, and when the data transmission is completed, the cached training process data is deleted; Perform data cleaning on the transmitted training process data to obtain target training data; Aggregating the unstructured data in the target training data to obtain the collected training process data; The collected training process data is reorganized in the form of data cubes, wherein the reorganized training process data is used for evaluating the effect of virtual training.

[0006] In some embodiments, the data transmission of the training process data is provided with transmission conditions, wherein the transmission conditions include constrained input conditions and standardized output conditions; The constraint input conditions include: applying the data input format defined in the data normalization model; The standard output conditions include: outputting data in a format defined in a formulated data dictionary.

[0007] In some embodiments, data cleaning is performed on the transmitted training process data, including: For the transmitted training process data, perform at least one of the following: Clean up the training process data with missing data: Clean the training process data that contains duplicate data: Clean up the training process data with data errors: Clean out the training process data where there is unavailable data.

[0008] In some embodiments, the unstructured data includes text training data of a text type, and the aggregating process is performed on the unstructured data in the target training data to obtain the collected training process data, including: Performing word segmentation on a text segment corresponding to the text training data to obtain a plurality of words in the text segment; Perform word embedding processing on the word to obtain an embedding numerical vector corresponding to the word; Calling a pre-trained language model to perform semantic analysis on the embedded numerical vector corresponding to the word to obtain aggregated target training data; Entity features are extracted from the aggregated target training data to obtain corresponding feature items, and the feature items and the target training data are used as the collected training process data.

[0009] In some embodiments, extracting entity features from the aggregated target training data to obtain corresponding feature items includes: Extracting sentiment polarity features from the aggregated target training data; or extracting training topic features from the aggregated target training data; The sentiment polarity feature and the training topic feature are used as corresponding feature items.

[0010] In some embodiments, the training process data of the virtual training system includes state data without data change and stateless data with data change, and the method further includes: When data transmission is performed, the transmitted training process data is stored through the training database; The storage process of the transmitted training process data includes: For state data that does not have data changes, a full copy method is used for storage; For stateless data with data changes, a differentiated approach is used for storage.

[0011] The present invention also provides a training data acquisition device for virtual training effect evaluation, the device comprising the following modules: A transmission module, used for caching the training process data of the virtual training system when the training process data is transmitted, and deleting the cached training process data when the data transmission is completed; A cleaning module is used to clean the transmitted training process data to obtain target training data; An aggregation module is used to aggregate the unstructured data in the target training data to obtain the collected training process data; The compilation module is used to compile the collected training process data in the form of data cubes, wherein the compiled training process data is used to evaluate the effect of virtual training.

[0012] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the training data collection method for evaluating the effect of virtual training as described in any one of the above is implemented.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the training data collection method for virtual training effect evaluation as described in any one of the above is implemented.

[0014] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-described training data collection methods for virtual training effect evaluation.

[0015] The training data collection method and device for virtual training effect evaluation provided by the present invention, when completing the data transmission of the training process data of the virtual training system, adopts two methods of transmission and caching to complete the collection of the training process data. Then, the transmitted training process data is cleaned, and the unstructured data therein is aggregated to obtain the collected training process data, and finally the data is reorganized in the form of data cubes for virtual training effect evaluation. Thus, by providing an efficient data collection means, it supports receiving training process data from the virtual training system and real-time aggregation and efficient storage to meet the real-time training evaluation needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced one by one below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is a flow chart of a training data collection method for virtual training effect evaluation provided by the present invention.

[0018] Figure 2 It is a structural schematic diagram of a training data acquisition device for virtual training effect evaluation provided by the present invention.

[0019] Figure 3 It is a schematic diagram of the physical structure of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Combine the following Figure 1-Figure 2 The present invention describes a training data collection method and device for evaluating the effect of virtual training. Figure 1 , Figure 1 is a flow chart of a training data collection method for virtual training effect evaluation provided by the present invention, such as Figure 1 As shown, the method includes the following steps 101 to 104, which are described one by one below.

[0022] Step 101: When data transmission is performed on the training process data of the virtual training system, the training process data is cached, and when the data transmission is completed, the cached training process data is deleted.

[0023] First of all, it should be noted that in the embodiments of the present invention, the virtual training system can be an artificial intelligence virtual character in a virtual computer system as a trainee to simulate a real person to perform physical training, body training or other sports training. In other embodiments, it can also be in a real scene, in real time, to map the real scene of real people performing physical training, body training or other sports training to the computer system, so that the computer system acts as a virtual training system.

[0024] The core of training process data collection is how to collect training process data from the virtual training system and save it in the training database for evaluation of virtual training effects. Data collection involves the data transmission process, and data transmission can use technologies such as Flume as data collection and transmission tools. In addition, due to the large scale of training process data generated by the virtual training system, data loss or even transmission failure is inevitable. Therefore, when the training process data of the virtual training system is transmitted, the training process data is cached, and when the data transmission is completed, the cached training process data is deleted.

[0025] When data is transmitted here, the training process data is cached before it reaches the destination, and the cached training process data is deleted when the training process data actually reaches the destination, that is, when the data transmission is completed. The destination here refers to the training database. In this way, even if there is a problem in the data transmission process, it can be remedied through the cached training process data, thereby ensuring the success rate of data transmission.

[0026] In order to facilitate the subsequent preprocessing of the training process data, the data transmission process needs to be strictly limited according to the set transmission rules. The embodiment of the present invention sets transmission conditions for the data transmission of the training process data, and the transmission conditions include constrained input conditions and standardized output conditions, so as to constrain the input of data and standardize the output of data.

[0027] Among them, the constraint input conditions are the input conditions that need to be met when extracting the training process data from the virtual training data, specifically including the data input format specified in the application data normalization model. The data normalization model is a model that uses the planning standards of data engineering to plan the input of the training process data to unify the input of the constrained data. Among them, the processing process of the data normalization model uses the specified unified dimension, data input format, mandatory items, data preservation verification and other means to constrain the input of the data, thereby ensuring that the input training process data has the specified unified dimension, unified input format, contains mandatory items, and can pass the data preservation verification.

[0028] The standard output conditions are the output conditions that must be met when the training process data is transferred to the destination for storage, including the data output format specified in the data dictionary. The data dictionary can be manually formulated according to the data storage format of the training database. The data dictionary defines the data items, data structures, data flows, data storage, processing logic, external entities and other attributes of the data. When the training process data is output and stored, the standardized data interaction must be carried out in strict accordance with the format content of the data dictionary. If the data interaction is cross-system, the underlying database operation of the training database is directly replaced by subscription and publishing.

[0029] In some embodiments, when the training database transmits and receives external training process data, the disk I / O (input / output) performance of the virtual training system deployed is an order of magnitude different from memory access. When the amount of data recorded by the virtual training system is large, the collection and storage of training process data may become a bottleneck for evaluating the operating performance of the subsystem. Therefore, when facing a large amount of training process data, in addition to improving the storage I / O speed, it is also necessary to design the mechanism, content, storage format, etc. of data collection during simulation training runtime.

[0030] The training process data recorded in the virtual training system can be distinguished according to the data type, such as training assessment completion data, training personnel status data, etc. In different virtual training scenarios, the goals and contents of training process data collection are different, so the data collection requests are also different. In order to better receive, aggregate and process training process data from different virtual training systems, the training database of the embodiment of the present invention technically adopts an extensible data collection mechanism. On the one hand, the most basic data collection function requirements need to be met, and on the other hand, in order to meet the ever-changing data collection requests, the training database can configure the data collection requests in real time.

[0031] In addition, in the virtual training system, more than 90% of the training process data that needs to be recorded comes from the state data of the training entity objects that change over time, that is, the training process data includes state data without data changes and stateless data with data changes, which will be explained below. These two types of training process data are needed for data analysis, replay, and simulation state preservation and recovery in the subsequent virtual training effect evaluation.

[0032] Therefore, when recording the training process data in the virtual training system, the present invention adopts a data I / O mechanism based on object serialization, that is, the training process data is object serialized, written into the I / O stream and saved on the disk, and the disk data can be transferred to the training database for subsequent training process data preservation and recovery and data playback. This can effectively reduce the complexity of data I / O and improve the efficiency of disk data I / O, while also improving the efficiency of data transmission.

[0033] In some embodiments, when the training process data is transmitted, the transmitted training process data is stored in the training database, that is, the training process data is stored in the training database after being transmitted. In the process of storing the transmitted training data, in order to save storage resources, different data storage methods need to be used for different types of training process data.

[0034] The training process data of the virtual training system includes state data without data changes and stateless data with data changes. Among them, state data without data changes are static data, such as some entity names, project names, etc. Stateless data with data changes are dynamic data. The data changes here can be condition changes, such as the performance status of the moving entity, whether a fault occurs during normal operation or normal operation is restored after a fault occurs, or periodic changes, such as changes in the position and action of the moving entity within a time period.

[0035] Based on this, when storing the transmitted training process data, the state data without data changes is stored in a full copy mode. The full copy mode is to store all state data periodically. The periodicity can be in units of data frames. Regardless of whether the subsequent data will change, it is directly copied and stored. This is used when computing power is tight and storage resources are abundant. The stored state data does not need to consider the context when it is used, ensuring the integrity of each frame of data.

[0036] For stateless data with data changes, a differentiated storage method is used. Differential storage is to count the changes in the data in real time during the storage process, record the complete data in the initial data frame, and only record the changed data when the data changes. For example, at the initial moment, the initial position and initial action of the moving entity are fully recorded first. As time changes (for example, within five minutes), the moving entity changes its action, and the action change data is recorded and stored in real time. If the initial position of the moving entity does not change within five minutes, there is no need to record and store any position data within these five minutes. Only when the moving entity starts to move in the sixth minute, the changed position data is recorded and stored. The differentiated storage method needs to count and compare the changes in the data in real time when storing data, and then determine whether to record and store it, so it takes up too much computing resources. But from another perspective, this targeted data storage can save storage resources to a certain extent compared to directly storing all data.

[0037] Step 102: clean the transmitted training process data to obtain target training data.

[0038] However, considering that the training process data sources of the virtual training system are complex and diverse, there are business data already available in various business systems, and there are also data manually collected and entered to help the virtual training system. Due to the different data recording and storage mechanisms, there are also certain data errors in the actual use process. Therefore, it is necessary to clean the transmitted training process data to obtain the target training data. The data errors of the training process data are generally manifested in four categories: missing data, duplicate data, data errors, and unavailable data, which are explained one by one below.

[0039] Missing data is generally manifested as missing recorded data, empty values ​​in the recorded data, or both. For example, the training process data may not record the training name (such as physical training, aerobic exercise, etc.), or the corresponding project name is not recorded and stored in the training name at a certain moment (such as aerobic exercise does not record running).

[0040] Duplicate data usually manifests itself as multiple identical data. For example, these data may be collected manually to help a virtual training system. The virtual training system and manual data collection are performed simultaneously, so multiple entries are repeated during recording. You only need to keep one of them.

[0041] Data errors generally manifest themselves as the recorded data not being recorded strictly in accordance with the specifications, resulting in abnormal values. For example, the position data of a moving entity should be recorded in the format of two digital coordinates, but only one digital coordinate is actually recorded, and the position data is determined to be an abnormal value.

[0042] The unavailable data usually manifests itself as the data not being recorded in the defined data format. For example, different data sources define different gender data formats. Some data sources define gender data formats as male or female, but some data sources use corresponding symbols (such as 0 and 1, or gender symbols) to define male or female. Another example is that time data is not recorded strictly in the standard data format of year, month, and day, which results in the unavailability of gender data and time data.

[0043] With respect to the above four types of data errors, during data cleaning, if it is detected that the training process data has at least one type of data error among the above four types, the training process data with the four types of data errors will be cleaned.

[0044] Step 103: Aggregate the unstructured data in the target training data to obtain the collected training process data.

[0045] For the target training data obtained after data cleaning, it is considered that there are still some unstructured data in the target training data. For example, in the virtual training system, some text training data of text type will also be recorded, which is called unstructured data, such as the training situation in the virtual training system, the conclusions of the expert consultation and review of the training process, and other text type data. The semantics of this type of text training data can generally show the quality status of the training process, which is helpful for the subsequent evaluation of the effect of virtual training. Therefore, after performing the data cleaning in step 102, this type of text training data also needs to be accurately segmented, effectively vectorized, and deeply semantically understood through the pre-trained language model, and finally relevant features are extracted for modeling analysis, so as to quantitatively analyze the training status of personnel. These processes all belong to the aggregation process of training process data. Therefore, the present invention performs aggregation processing on the unstructured data in the target training data to obtain the collected training process data, thereby completing the collection of training process data, which is described in detail below.

[0046] First, unstructured data, that is, text training data of text type, is extracted from the target training data obtained in step 102, and then the text segment corresponding to the text training data is segmented into words to obtain multiple words of the text segment to achieve accurate word segmentation. Next is effective vectorization, by calling the word embedding model to embed the words and obtain the embedded numerical vector corresponding to the word. Next, the pre-trained language model can be called to perform semantic analysis on the embedded numerical vector corresponding to the word to obtain the converged target training data.

[0047] The process of semantic analysis is to calculate the distance between embedded numerical vectors to reflect semantic similarity. Semantic similarity enables the pre-trained language model to understand the semantics of unstructured data, such as judging the learning situation in the virtual training system, predicting the corresponding training topic, or predicting the emotional polarity from the conclusion of the expert consultation and review of the training process, thus representing the quality of the training process.

[0048] After the aggregation process is completed, relevant features can be extracted from the processed data according to the actual application needs. Therefore, entity features are extracted from the aggregated target training data to obtain corresponding feature items. This feature item is expressed as structured data for subsequent virtual training effect evaluation. Entity feature extraction generally involves extracting sentiment polarity features from the aggregated target training data, or extracting training theme features from the aggregated target training data, and using sentiment polarity features and training theme features as corresponding feature items. Here, training theme features can be extracted based on the training situation in the virtual training system, such as determining the purpose of training, training goals, training effects, etc., or sentiment polarity features can be extracted based on the conclusions of the expert consultation and review of the training process, such as determining training effect feedback, training quality evaluation, etc.

[0049] Emotional polarity features and training topic features are entity features, which play a great role in the subsequent evaluation of virtual training effects. In addition, these entity feature items belong to structured data, so the aggregation of data also realizes the transformation from unstructured data to structured data. Finally, the feature items and target training data are used as the training process data after collection, for example, as additional feature items in the training process data for supplementary recording.

[0050] In some embodiments, entity feature extraction can adopt a variety of methods, which can be based on statistics, natural language processing, and machine learning. These methods extract different text features for text data, so the effect of extracting features is also the same. In actual application environments, multiple methods can be used for feature extraction based on the text corpus of unstructured data, and the advantages can be complemented by comparing the extraction effects to improve the success rate of extraction. At the same time, when extracting entity features, external knowledge bases can be used to enrich the content of the extracted text data, reveal the subject features of the text data (for example, adding text tags), and improve the extraction effect.

[0051] As for machine learning methods, more text data corresponding to unstructured features can be collected to extract corresponding entity features, and then these entity features can be used to train the machine learning model and make predictions, so as to improve the success rate of the machine learning model in extracting entity features.

[0052] Step 104: compile the collected training process data in the form of data cubes, wherein the compiled training process data is used for evaluating the effect of virtual training.

[0053] After obtaining the collected training process data through the above step 103, in order to make the collected training process data be used as the input of the evaluation system in a more intuitive form, the virtual training effect evaluation is realized. In the embodiment of the present invention, the collected training process data is compiled in the form of a data cube. A data cube is a multidimensional data model, and the most common mode is a star mode, which includes a large central table (fact table) in which a large number of non-redundant data are recorded, and also includes a group of small subsidiary tables. An subsidiary table represents a data dimension, and the attributes in the subsidiary table will form a data hierarchy. This can meet the needs of users for data query and analysis from multiple dimensions and levels. When data query and analysis are required, data query processing can be performed from the data cube in an online analytical processing manner, thereby optimizing the data query method and improving the flexibility of data analysis.

[0054] The compiled training process data can directly go through the subsequent process, such as being input into the evaluation system for virtual training effect evaluation. Since the training process data has gone through a series of processes such as data cleaning, aggregation, and compilation, it can be directly used in the evaluation system for virtual training effect evaluation, which can meet multi-dimensional evaluation needs. The training process data transmitted and collected in real time from the virtual training system can also be cleaned, aggregated, and compiled in real time, so it can also be used for real-time virtual training effect evaluation, ensuring the timeliness of virtual training effect evaluation.

[0055] The training data collection method for virtual training effect evaluation provided by the present invention, when completing the data transmission of the training process data of the virtual training system, adopts two methods of transmission and caching to complete the collection of the training process data. Then, the transmitted training process data is cleaned, and the unstructured data therein is aggregated to obtain the collected training process data, and finally the data is reorganized in the form of data cubes for virtual training effect evaluation. Thus, by providing an efficient data collection means, it supports receiving the training process data from the virtual training system and performing real-time aggregation and efficient storage, thereby meeting the real-time training evaluation needs.

[0056] The training data collection device for virtual training effect evaluation provided by the present invention is described below. The training data collection device for virtual training effect evaluation described below and the training data collection method for virtual training effect evaluation described above can refer to each other.

[0057] See also Figure 2 , Figure 2 is a structural schematic diagram of a training data acquisition device for virtual training effect evaluation provided by the present invention, such as Figure 2 As shown, the training data acquisition device for virtual training effect evaluation includes: a transmission module 201, a cleaning module 202, a convergence module 203, and an editing module 204. Among them, the transmission module 201 is used to cache the training process data of the virtual training system when the training process data is transmitted, and delete the cached training process data when the data transmission is completed; the cleaning module 202 is used to perform data cleaning on the transmitted training process data to obtain target training data; the convergence module 203 is used to perform convergence processing on the unstructured data in the target training data to obtain the collected training process data; the editing module 204 is used to use the data form of data cube to perform data editing on the collected training process data, wherein the edited training process data is used for virtual training effect evaluation.

[0058] It should be noted that the beneficial effects of the training data acquisition device used for virtual training effect evaluation here and the training data acquisition method used for virtual training effect evaluation mentioned above can correspond to each other, so the beneficial effects of the training data acquisition device used for virtual training effect evaluation will not be repeated here.

[0059] Figure 3 is a schematic diagram of the physical structure of an electronic device provided by the present invention, such as Figure 3 As shown, the electronic device may include: a processor 310, a communication interface 320, a memory 330 and a communication bus 340, wherein the processor 310, the communication interface 320 and the memory 330 communicate with each other through the communication bus 340. The processor 310 may call the logic instructions in the memory 330 to execute the training data acquisition method for virtual training effect evaluation, the method comprising: when data transmission is performed on the training process data of the virtual training system, the training process data is cached, and when the data transmission is completed, the cached training process data is deleted; data cleaning is performed on the transmitted training process data to obtain target training data; aggregation processing is performed on the unstructured data in the target training data to obtain the collected training process data; data cube data is used to organize the collected training process data, wherein the organized training process data is used for virtual training effect evaluation.

[0060] In addition, the logic instructions in the above-mentioned memory 330 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0061] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the training data collection method for virtual training effect evaluation provided by the above-mentioned methods, the method including: when data is transmitted for the training process data of the virtual training system, the training process data is cached, and when the data transmission is completed, the cached training process data is deleted; data cleaning is performed on the transmitted training process data to obtain target training data; aggregation processing is performed on the unstructured data in the target training data to obtain the collected training process data; data compilation is performed on the collected training process data in the form of data cubes, wherein the compiled training process data is used for virtual training effect evaluation.

[0062] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the training data collection method for virtual training effect evaluation provided by the above-mentioned methods, the method comprising: when data is transmitted for the training process data of the virtual training system, caching the training process data, and deleting the cached training process data when the data transmission is completed; performing data cleaning on the transmitted training process data to obtain target training data; performing aggregation processing on the unstructured data in the target training data to obtain the collected training process data; and performing data compilation on the collected training process data in the form of data cubes, wherein the compiled training process data is used for virtual training effect evaluation.

[0063] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0064] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A training data collection method for virtual training effect evaluation, characterized in that: The method comprises: When data transmission is performed on the training process data of the virtual training system, the training process data is cached, and when the data transmission is completed, the cached training process data is deleted; Perform data cleaning on the transmitted training process data to obtain target training data; Aggregating the unstructured data in the target training data to obtain the collected training process data; The collected training process data is reorganized in the form of data cubes, wherein the reorganized training process data is used for evaluating the effect of virtual training.

2. The training data collection method for virtual training effect evaluation according to claim 1, characterized in that: The data transmission of the training process data is provided with transmission conditions, wherein the transmission conditions include constrained input conditions and standardized output conditions; The constraint input conditions include: applying the data input format defined in the data normalization model; The standard output conditions include: outputting data in a format defined in a formulated data dictionary.

3. The training data collection method for virtual training effect evaluation according to claim 1 is characterized in that: Perform data cleaning on the transmitted training process data, including: For the transmitted training process data, perform at least one of the following: Clean up the training process data with missing data; Clean the training process data that contains duplicate data; Clean up the training process data with data errors; Clean out the training process data where there is unavailable data.

4. The training data collection method for virtual training effect evaluation according to claim 1 is characterized in that: The unstructured data includes text training data of text type, and the unstructured data in the target training data is aggregated to obtain the collected training process data, including: Performing word segmentation on a text segment corresponding to the text training data to obtain a plurality of words in the text segment; Perform word embedding processing on the word to obtain an embedding numerical vector corresponding to the word; Calling a pre-trained language model to perform semantic analysis on the embedded numerical vector corresponding to the word to obtain aggregated target training data; Entity features are extracted from the aggregated target training data to obtain corresponding feature items, and the feature items and the target training data are used as the collected training process data.

5. The training data collection method for virtual training effect evaluation according to claim 4 is characterized in that: The extracting entity features from the aggregated target training data to obtain corresponding feature items includes: Extracting sentiment polarity features from the aggregated target training data; or extracting training topic features from the aggregated target training data; The sentiment polarity feature and the training topic feature are used as corresponding feature items.

6. The training data collection method for virtual training effect evaluation according to claim 1, characterized in that: The training process data of the virtual training system includes state data without data change and stateless data with data change, and the method further includes: When data transmission is performed, the transmitted training process data is stored through the training database; The storage process of the transmitted training process data includes: For state data that does not have data changes, a full copy method is used for storage; For stateless data with data changes, a differentiated approach is used for storage.

7. A training data acquisition device for virtual training effect evaluation, characterized in that: The device comprises: A transmission module, used for caching the training process data of the virtual training system when the training process data is transmitted, and deleting the cached training process data when the data transmission is completed; A cleaning module is used to clean the transmitted training process data to obtain target training data; An aggregation module is used to aggregate the unstructured data in the target training data to obtain the collected training process data; The compilation module is used to compile the collected training process data in the form of data cubes, wherein the compiled training process data is used to evaluate the effect of virtual training.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the training data collection method for virtual training effect evaluation as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the training data collection method for virtual training effect evaluation as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the training data collection method for virtual training effect evaluation as described in any one of claims 1 to 6 is implemented.

Images