A decision tree based article recommanding system

In this study, an article recommendation system for English reading comprehension improvement is proposed. The goal of this study is to find out the most important attributes that affect the difficulty of an article according to the levels granted by the General English Proficiency Test (GEPT). Using the determined attributes to classify the articles gathered by the crawler from the Internet everyday and recommending the proper ones to the user, the proposed system is designed to keep the users from being recommended the articles those are too hard or too simple and retain their learning enthusiasm. To determine the attributes that affect the difficulty of an article, the classification algorithms of decision tree are used to construct the classification rules. The experimental result shows that to classify article into the 3 levels defined as elementary, intermediate, and high-intermediate according to GEPT, require 5 attributes to achieve above 70% above accuracy; while to classify articles into just elementary and high-intermediate level, only 2 attributes are required for 80% above accuracy.