AUTOMATICALLY FACILITATING DISCUSSION IN ONLINE FORUMS

Abstract

In this thesis, we investigate how NLP, and Recommendation Engines can help the users navigate through online discussion forums efficiently. Firstly, we propose building a recommendation system to improve the visibility of threads. We develop a probabilistic graphical model to consider user-interests to recommend her useful posts. Unlike traditional recommendation system settings, discussion forums also suffer from influx of new posts. We propose a deep neural network to represent a post based on the words and utilize them to identify the potentially interested users. This is addressed as an Extreme Multi-Class Multi-Labelling problem and show we that this formulation works well in practice. The open nature of the online forums attracts many participants. Consequently, often the large number of posts becomes unmanageable as many repetitive or irrelevant posts are posted. To this end, we propose a neural network based framework that can identify helpful posts in a discussion thread automatically.