Software Feature Location Based on Topic Models

Abstract

Feature location is the activity of identifying an initial location in the source code that implements special functionality in a software system. Existing techniques for feature location broadly fall into three categories, based on the type of information they use: text, static, and dynamic. The techniques based on dynamic may generate large amount of data and is difficult to utilize. This paper presents a method combing the text and static techniques. A feature location technique based on topic modeling is introduced, and the topic cohesion and coupling is computed by software dependency network to improve the effects of feature location. When the topic cohesion degree is low and topic coupling degree is high, software dependency network is used to find additional candidate program elements. This method is empirically evaluated through several experiments. Experimental results show that the topic modeling based feature location improves the effectiveness of feature location when compared with other techniques.