1. Twitter 1 2 3 3 3 Twitter Twitter ( ) Twitter (trendspotter) Twitter 5277 24 trendspotter TRENDSPOTTER DETECTION SYSTEM FOR TWITTER Wataru Shirakihara, 1 Tetsuya Oishi, 2 Ryuzo Hasegawa, 3 Hiroshi Hujita 3 and Miyuki Koshimura 3 It is too difficult for us to find out trends with search engines. Twitter, a popular microblogging tool, has seen a lot of growth since it launched in October, 2006. Information about the trends are posted by many twitterers. If we find out trendspotters from twitterers, and follow them, we can get it more easily. Our system uses the burst detection algorithm, and we verified its effectiveness for Twitter s posts. Finaly, we succeeded in detecting the 24 trendspotters by 5277 users. SNS Twitter Twitter 2010 1 473 (2010 2 ) 3) http://twitter.com Twitter Twitter SNS Twitter Twitter Twitter Twitter Twitter 1 Graduate School of Information Science and Electrical Engineering, Kyushu University 2 Research Institute for Information Technology, Kyushu University 3 Faculty of Information Science and Electrical Engineering, Kyushu University 1 c 2010 Information Processing Society of Japan
2. Twitter Twitter 2.1 Twitter Twitter( ) 140 2.2 Twitter Twitter ( 1 ) RT ReTweet RT ReTweet RT ( 2 ) URL Twitter Twitter 140 URL URL URL 140 URL URL ( 3 ) bot (bot) JR TV ( 4 ) API 2006 Twitter Twitter API Twitter Twitter buzztter 7) buzztter Twitter buzztter ( 5 ) (#) Twitter Twitter #traindeley #traindeley # 2.3 Twitter ( SNS) Twitter 2.3.1 Twitter Twitter Twitter Twitter 140 Twitter 2.3.2 SNS Twitter SNS 2 c 2010 Information Processing Society of Japan
mixi 8) SNS RT Twitter SNS SNS Twitter SNS Twitter SNS Twitter Twitter 3. 4) Kleinberg 5) document stream document stream document ( ) Twitter 4. ( 1 ) buzztter s buzztter ( ) ( 2 ) Twitter API s ( 1 Fig. 1 Overview of the system ) ( 3 ) s ( ) 3 ( 4 ) ( t )20 20 s trendspotter ( 5 ) 1 4 trendspotter 5. trendspotter ( 1 ) (5.1 ) trendspotter ( 2 ) trendspotter (5.2 ) trendspotter ( 3 ) trendspotter (5.3 ) 3 c 2010 Information Processing Society of Japan
2 Fig. 2 A number of posts including the buzz word 3 Fig. 3 A number of posts including the buzz word 5.2 trendspotter 5.1 buzztter ( ) twitter 5.1.1 2010 1 27 WILLCOM twitter 2010 1 26 22 00 27 6 00 4.1 (20 ) 20 1 27 2 5.1.2 twitter 2010 1 26 7 00 26 6 00 4.2 (20 ) 20 1 26 9 00 10 00 1 26 18 00 5.1.3 twitter 2010 1 28 16 4 Fig. 4 A number of posts including the buzz word 00 2 1 10 00 4.3 (20 ) 20 1 31 21 20 2 1 1 40 2 1 7 00 5.1.4 5 (a) 4 c 2010 Information Processing Society of Japan
5 Fig. 5 Consideration1: (a) (a) Tue, 26 Jan 2010 23:04:31 +0900 mubot http://twitter.com/mubot 2 19 ( WILLCOM ) ( ) (b) (c) (c) (c) Wed, 27 Jan 2010 04:01:40 +0900 hagexx http://twitter.com/hagexx Wed, 27 Jan 2010 04:06:41 +0900 ysbee http://twitter.com/ysbee = RT @sarustar RT @Hagexx: - http://ow.ly/10fs9 Wed, 27 Jan 2010 04:08:47 +0900 awazeno999 http://twitter.com/awazeno999 Wed, 27 Jan 2010 04:23:17 +0900 takeori http://twitter.com/takeori http://bit.ly/bw56yi Wed, 27 Jan 2010 04:24:12 +0900 tdaiki http://twitter.com/tdaiki RT @takeori: http://bit.ly/bw56yi Wed, 27 Jan 2010 04:26:13 +0900 tabloid http://twitter.com/tabloid RT http://bit.ly/bw56yi (via @takeori) 04 08 04 23 20 1 ( A B ) 5 c 2010 Information Processing Society of Japan
6 Fig. 6 Consideration2: 7 Fig. 7 Consideration3: 5.1.5 ( 6) (d) (e) (d) Tue, 26 Jan 2010 09:10:44 +0900 (e) togamim http://twitter.com/togamim @ Tue, 26 Jan 2010 18:01:57 +0900 MAKEPURA http://twitter.com/makepura NHK (d) (e) 5.1.6 7 2010 1 28 16 00 2 1 10 00 5.1.7 Twitter ( ) 5.2 : trendspotter trendspotter Buzztter 200 1500 2010 1 2 trendspotter 20 trendspotter 6 c 2010 Information Processing Society of Japan
5.2.1 200 5277 trendspotter 200 20 =4000 N trendspotter (5277 ) 1 1 N trendspotter trendspotter 24 5.3.1 110 34381 N (34381 ) 2 Table 1 A number of trendspotters for N buzz words and percentage of total (N) (%) 1 4734 89.71 2 463 8.77 3 56 1.06 4 12 0.23 5 9 0.17 6 3 0.06 2 N Table 2 A number of users posted N buzz words in bursts and percentage of total (N) (%) 2 31174 90.67 3 5 2916 8.48 6 8 227 0.66 9 11 46 0.13 12 18 0.05 5.2.2 1 trendspotter 1.52% 5277 80 5.3 : trendspotter trendspotter Buzztter 110 1500 2010 2 5 2 7 ( ) trendspotter 24 N (24 ) 3 3 trendspotter 24 N Table 3 A number of the trendspotters posted N buzz words in bursts and percentage of total 5.3.2 (N) (%) 2 8 33.3 3 5 8 33.3 6 8 4 16.7 9 11 3 12.5 12 1 4.2 2 3 trendspotter 7 c 2010 Information Processing Society of Japan
100 6 1% trendspotter 33.4% 1 1.43 trendspotter 1 4.54 6. Twitter (trendspotter) ( ) Twitter ( ) trendspotter 7. Twitter ( ) trendspotter ( ) Twitter ( IT ) ( ).. Twitter /. ( ) ( ). Hadoop. Hadoop HBase, Cassandra. 21500102 1) Twitter 2009 2) 140 2009 3) http://www.netratings.co.jp/new news/news02242010.htm 4) document stream burst IPSJ SIG Notes 2004(23) pp.85-92 20040304 5) Jon Kleinberg Bursty and hierarchical structure in streams the 8th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining 2002 6) 15 pp.568-576 1996 7) http://buzztter.com 8) http://mixi.jp 8 c 2010 Information Processing Society of Japan