Microblog Retrieval Using Topical Features and Query Expansion.

Abstract

Retrieving information from Twitter is always challenging given its volume, inconsistent writing and noise. Existing systems focus on term-based approach, but important topical features such as person, proper noun and events are often neglected, leading to less satisfactory results while searching information from tweets. This paper propose a novelty feature extraction algorithm which targets the above problems, and present the experiment results using TREC11 dataset. The proposed approach considers both term-based and patternbased features and distribute weights accordingly. We experiment four different setting to evaluate different combinations and results show that our approach outperformed traditional method of using term-based or pattern only methods and signify the importance of topical features in microblog retrieval. 1.