Query-Less Retrieval of Interesting Postings in a WebForum

Abstract

In web-based discussion forums or newsgroups much information arrives everyday, but only a small fraction of that information might be interesting to a user. Users could browse to get their required information, but browsing through large piles of links and web pages is not a very efficient process. Users could prepare search queries to find the relevant piece of information, but they may not know exactly what to look for. Therefore this research aims at developing methods that can be implemented in a system to retrieve interesting items without evaluating a user query. This work also investigates and implements some supervised learning approaches to improve the retrieval performance of such a system. The goal of this research work is:

• To build and add an intelligent component to an existing discussion forum system enhancing its functionalities and intelligence.

• To solve the problem of finding interesting items for users by applying techniques from information retrieval, collaborative filtering and machine learning.

• To build a system that is capable enough to retrieve and recommend useful bits of information to its users.

• To build a model that improves the system’s performance by learning from retrieval successes and failures