Transfer Learning for Question Similarity between Stack Overflow Posts

MPhil Thesis Defence


Title: "Transfer Learning for Question Similarity between Stack Overflow Posts"

By

Mr. Victor Wing-chuen KWAN


Abstract

Developers often come to Stack Overflow to seek help about their programming 
problems. However, the technicality of the content makes the task of relevant 
question retrieval especially difficult. Questions on Stack Overflow are very 
susceptible to differences in nuance: there are varying degrees of question 
formality; there are many ways to present relevant keywords and names; and 
there are numerous degrees of specificity with which a question may be asked. 
Drawing from a wide pool of natural language processing techniques, we devise a 
model for question similarity that attempts to learn the semantic relationships 
between Stack Overflow questions using the tags and titles of posts. We 
additionally build around the idea of transferring knowledge from Quora to 
train our model to be more robust against the noisy Stack Overflow dataset. Our 
contributions include an effective model for question similarity that leverages 
transfer learning for added robustness; a study into how the model components 
contribute towards the success of the model; and a study into the differences 
between the Quora and Stack Overflow dataset through the lens of transferred 
knowledge.


Date:			Thursday, 24 May 2018

Time:			1:00pm - 3:00pm

Venue:			Room 5508
 			Lifts 25/26

Committee Members:	Dr. Sunghun Kim (Supervisor)
 			Prof. Andrew Horner (Chairperson)
 			Dr. Yangqiu Song


**** ALL are Welcome ****