Automatic parallel test paper generating method and system based on machine learning

Abstract

The invention provides an automatic parallel test paper generating method and system based on machine learning. The method comprises the steps: obtaining an original test paper; according to the questions in the original test paper, obtaining questions similar to each question in a candidate question bank through a similarity algorithm, and forming a first question list; filtering the first question list to obtain a second question list; establishing a prediction model according to historical processing behaviors of the user on the questions and the test papers containing the questions, and predicting the probability of the processing behaviors; calculating question scores according to the probability of the processing behaviors, sorting according to the question scores, and selecting a certain number of questions ranked in the front; ranking the probability of each processing behavior, converting rankings into scores, fusing and ranking the scores, and then forming parallel test paper; establishing a sequence adjustment model according to historical change data of the user on the question sequence, and adjusting the question sequence in the parallel test paper; and displaying the parallel test paper, and obtaining a final test paper according to the question changing operation and the sequence adjusting operation of the user.

Description

technical field [0001] The invention relates to the technical field of machine learning, in particular to a method and system for automatic parallel paper composition based on machine learning. Background technique [0002] In educational work, assignments and exam papers (subsequently collectively referred to as paper composition, paper composition including homework composition, unit test composition, mid-term and end-of-term paper composition, etc.) need to be stored in an extremely large question bank. system (currently, the relatively mature question bank is on the order of tens of millions, and it is still increasing at a rate of several million per year). [0003] Usually, it takes about 20 seconds to 1 minute for a user to understand a question and determine whether it needs to be added to the question bank. Therefore, the workload of finding a suitable question in a large number of question banks is huge. How to enable users to efficiently search for more suitable ...

AI Technical Summary

Problems solved by technology

In this kind of practice, there are at least the following deficiencies: 1. It is necessary to classify the knowledge points and chapters of the topic very accurately. The accuracy of the knowledge point labels of the topic directly affects whether users can find the topic based on a certain knowledge point or chapter. Questions, but the general question bank does not have its own knowledge points marked when uploading, and the knowledge point system of each institution's question bank in the country is different, and it needs to be manually mapped to the local knowledge point system based on the understanding of the current topic middle

2. Users need to read a lot of questions corresponding to knowledge points and chapters, because the magnitude of the question bank of a certain knowledge point / chapter is also very large (the average magnitude of the question bank for each knowledge point is several thousand, and all screening It takes hundreds of hours to complete), a user manually finds dozens of questions out of thousands of questions, and the efficiency is also very low. Dozens of questions are already not easy

3. The quality of the test questions generated by the topic group test is limited by the teaching and research level and seriousness of the test group users themselves. Therefore, the test papers of the user group with a relatively poor level are likely to be relatively poor (such as repeated tests of the same knowledge point, difficulty The distribution of coefficients is unreasonable, some topics are out of class, etc.)

4. It is impossible to make comprehensive questions, because the questions of many subjects (such as geography, chemistry, etc.) are comprehensive in nature and will include many knowledge points, so it is impossible to search for a certain knowledge point Add it to the question bank (otherwise it will easily cause students to not do the question without learning other related knowledge points), so the effect of searching for knowledge points and screening for comprehensive questions is extremely poor of

Images