Construction And Automatic Classification Of Classification Dataset For CS1 Test Items Based On Bloom’s Taxonomy

Curriculum evaluation is a key link of teaching reform,which involves teaching cases,test items and classroom teaching.Aiming at the evaluation of test items in computing courses,this paper uses Bloom’s Taxonomy and takes the test items of "Introduction to Computer Science"(CS1)course as the corpus to construct classification dataset for CS1 test items.Using machine learning technology,analyze the difference between machine classification and manual labeling,discuss and improve the test items of computing courses,and improve the teaching quality of computing courses.The main work is as follows:(1)Aiming at the lack of data sets of computing course test items classification,this paper takes 2611 test items of CS1 from Princeton University and Guilin University of Electronic Science and Technology as the corpus,based on Bloom’s Taxonomy,gives the corresponding verb and noun seed bank for CS1 cognitive process dimension and knowledge dimension,and marks the positions of the two-dimensional matrix of Bloom’s Taxonomy that the test items can reach.Classification dataset for CS1 test items is constructed to provide high quality test items classification training set and test set for machine automatic classification of computing course test items.(2)Aiming at the automatic classification of computing course test items based on Bloom’s Taxonomy,this paper proposes an automatic classification model of CS1 test test items named TFERNIE-LR.The model consists of three parts: CSTFPOS-IDF algorithm,ERNIE model and Logistic Regression classifier.On the basis of the TFPOS-IDF algorithm,the CSTFPOS-IDF algorithm improves the model’s attention to the computing course keywords through the weight factor of the computing course keywords,and generate the word weight.At the same time,the pre-training model ERNIE augmented with entity knowledge was used to embed the test items word level vector,and the word weight and word level vector were combined to generate the test items text vector for automatic classification.Finally,Logistic Regression classifier is used to classify the CS1 test items automatically,and the F1 value in the cognitive process dimension and knowledge dimension of Bloom’s Taxonomy reaches 83.1% and 96.0% respectively.Compared with the experimental results of TFERNIE-LR model,ERNIE+TF-IDF and ERNIE+TFPOS-IDF,the F1 value in the cognitive process dimension is increased by 13.9% and 1% respectively,and the F1 value in the knowledge dimension is increased by 2.1% and 0.6% respectively.In addition,the K-nearest neighbor and support vector machine classifiers are used for experiments and comparison,and the TFERNIE-LR model has great improvement in effect.(3)Aiming at the calculation of the total value and gold content of computing course papers based on Bloom’s Taxonomy,this paper gives a two-dimensional matrix of test items value weight based on Bloom’s Taxonomy,and introduces three weight constraints: concept complexity,language complexity and difficulty.The calculation algorithm of the total value and gold content of the computing course papers is given,and the calculation is carried out on three sets of CS1 course papers of Guilin University of Electronic Science and Technology from 2019 to 2021.