SQL-DP

Introduction

This repository contains the implementations for difficulty prediction task for Structured Query Language (SQL) programming problems.

The main objective of the project is to predict the difficulty of each given SQL programming problem based on its stem and answer (i.e., SQL code).

SQL-DP Framework

In SQL-DP, semantic features of problem stems and structure features of problem answers in the form of SQL codes are both computed at first, using the NLP and the neural network techniques. Then, these features are used as the input to train a difficulty prediction model with the statistic error rates in tests as the training labels, where the whole modeling do not introduce any experts, some as knowledge labeling. Finally, with the trained model, we can automatic predict the difficulty of each SQL programming problem.

As shown in the picture above, SQL-DP framwork includes two modules: Feature Extraction (FE) module and Difficulty Prediction (DP) module. Besides, the Feature Extraction Module includes SQL Text Semantic Fearure Extraction (SQL-TSFE) module and SQL Code Struture Feature Extraction (SQL-CSFE) module.

The SQL-TSFE modul uses NLP technologies (e.g., Word2vec, TF-IDF) to extract text semantic feature from the stem of SQL programming problem.

And The SQL-CSFE module uses TBCNN model to extract the code structure feature from the answer of SQL programming problem.

In the DP module, some machine learning algorithms use the above features to predict the difficulty of SQL programming problems. It is worth noting that we need to convert the SQL code into AST, and then serialize the AST before entering the TBCNN model. So we use the SQL parser pglast to get the AST of SQL.

Requirements

SQL-TSFE module

• python 3.7
• gensim 4.1.0
• jieba 0.42.1
• numpy 1.21.2
• pandas 0.3.1
• scikit-learn 0.24.2

SQL Parse

• python 3.7
• pglast 3.7
• xlrd 1.2.0

Other

• python 2.7
• tensorflow 1.0.0
• tensorflow-fold 0.0.1
• pytest 3.0.7
• ddt 1.1.1
• future 0.16.0
• six 1.16.0
• numpy 0.16.6
• scikit-learn 0.20.4
• pytest 3.0.7
• ipdb 0.10.2
• ipython 5.3.0
• flake8 3.3.0

Data

Due to the problem of data privacy, we only give examples of data. For data samples, please check the data folder in the project.

Besides, the original code of the TBCNN model we use is here.