Publication Type

Conference Proceeding Article

Version

acceptedVersion

Publication Date

11-2011

Abstract

This paper presents our work for the Realtime Adhoc Task of TREC 2011 Microblog Track. Microblog texts like tweets are generally characterized by the inclusion of a large proportion of irregular expressions, such as ill-formed words, which can lead to significant mismatch between query terms and tweets. In addition, Twitter queries are distinguished from Web queries with many unique characteristics, one of which reflects the clearly distinct temporal aspects of Twitter search behavior. In this study, we deal with the first problem by normalizing tweet texts and the second by capturing the temporal characteristics of topic. We divided topics into two categories: time-sensitive and time-insensitive. For the time-sensitive ones, we introduce a decay factor to adjust the relevance score of results according to the expected date of the topical event to happen, and then re-rank the search results. Experiments demonstrate that our methods are significantly better than baseline and outperform the medium of all runs.

Discipline

Social Media | Theory and Algorithms

Research Areas

Data Science and Engineering

Publication

Proceedings of the 20th Text REtrieval Conference (TREC 2011)

Publisher

NIST Special Publication: SP 500-296

City or Country

Geithersburg, MD, USA.

Share

COinS