Home // ICDS 2014, The Eighth International Conference on Digital Society // View article

Detecting New Concepts in Social Media using Co-burst Pattern Mining

Authors:
Takako Hashimoto
David Shepard
Basabi Chakraborty
Tetsuji Kuboyama

Keywords: Social meda, burst pattern, unexpected words’ correlation, video service, Twitter, East Japan Great Earthquake

Abstract:
This paper proposes a method for detecting new concepts in social media using co-burst pattern mining technique. The new concepts are defined as correlations between unexpected words. The target social media are viewers’ comments attached to web videos and Twitter’s tweets related to the East Japan Great Earthquake that happened on Mar. 11 in 2011. Our proposed method first crawls viewers’ comments from web videos, and extracts words from them. Then it selects motive words candidates from words, and counts the occurrence numbers of tweets that include motive words candidates. To detect new concepts, it generates burst patterns based on occurrence numbers of motive words candidates over time and detects unexpected correlations between motive words candidates. By our method, after the earthquake, new unexpected correlations between motive words in social media are recognized as new concepts. For example, the method could extract motive words from web video comments, such as ”escape, nuclear plant” and ”Tokyo Electric Power Co., Inc.(TEPCO, that owns the nuclear plant), president.” Then it could detect the new concept ”escape (from) nuclear plant” and ”TEPCO’s president” on Twitter. In this paper, we provide the preliminary approximation results and discuss the effectiveness of our proposed method.

Pages: 176 to 181

Copyright: Copyright (c) IARIA, 2014

Publication date: March 23, 2014

Published in: conference

ISSN: 2308-3956

ISBN: 978-1-61208-324-7

Location: Barcelona, Spain

Dates: from March 23, 2014 to March 27, 2014