DeepLink: Deep-Learning Word Sementics to Link Software Artifacts

MPhil Thesis Defence


Title: "DeepLink: Deep-Learning Word Sementics to Link Software Artifacts"

By

Mr. Hongjoo LEE


Abstract

It is widely believed that developers' language and bug reporters' 
language are different, and the difference limits traceability between bug 
reports and commit changes. However, few studies have revealed the 
difference and tried to overcome the challenges. This paper deals with 
these issues. First, we clarify the textual difference and lexical 
relations between bug reports and commit changes by projecting words into 
context space with a deep learning technique. We also clarify some 
limitations of conventional textual similarity measure between bugs and 
changes on VSM due to the textual difference. Second, we propose a novel 
approach, DeepLink, that automatically analyzes the textual information 
and precisely recovers traceability between commit changes and 
corresponding bug reports. Lastly, we evaluate the performance of DeepLink 
on 10 large opensource projects. Our experimental results show that 
DeepLink outperforms conventional technique up to 28% in F-score.


Date:			Monday, 24 August 2015

Time:			4:00pm - 6:00pm

Venue:			Room 3494
 			Lifts 25/26

Committee Members:	Dr. Sunghun Kim (Supervisor)
 			Prof. Shing-Chi Cheung (Chairperson)
 			Dr. Qiong Luo


**** ALL are Welcome ****