DEIM Forum 2011 F5-4 contexthashtag Twitter 525 8577 1 1 1 525 8577 1 1 1 E-mail: kaieda@coms.ics.ritsumei.ac.jp, huang@fc.ritsumei.ac.jp, kawagoe@is.ritsumei.ac.jp contexthashtag Twitter Twitter Twitter Hashtag Hashtag Hashtag Hashtag contexthashtag Hashtag contexthashtag contexthashtag Hashtag Hashtag Twitter Hashtag Abstract An Event Recommendation System for Twitter Users Using the contexthashtag Takahiro KAIEDA, Hung-Hsuan HUANG, and Kyoji KAWAGOE Graduate School of Science and Engineering, Ritsumeikan University Nojihigashi 1 1 1, Kusatsu, Shiga, 525 8577 Japan Colledge of Information Science and Engineering, Ritsumeikan University Nojihigashi 1 1 1, Kusatsu, Shiga, 525 8577 Japan E-mail: kaieda@coms.ics.ritsumei.ac.jp, huang@fc.ritsumei.ac.jp, kawagoe@is.ritsumei.ac.jp In this paper, we propose an event recommendation system for Twitter users using contexthashtag. Twitter users can get rich event information by Hashtags benefitting from the widespread of Twitter. However, event retrieval merely using Hashtags may not meet the user s expectation due to Hashtags property where they can be defined independently by different users and lack coincidence. In order to solve this problem, we propose contexthashtag that forms a structured event-hashtag space where an existing event Hashtag can be located as a region. The system then recommends the user new events based on similarity comparison of contexthashtags between new events and ones of the events which the user ever participated. Key words Twitter Hashtag Text Classification Recommendation System 1. eventpage.jp 1 Twitter 2 [1] [2] [3] [4] 1 http://eventpage.jp/ 2 http://twitter.com/ Web Twitter Hashtag Hashtag Twitter hanabi ) Hashtag Hashtag Hashtag Web hashtagsjp 20000 2010 12 26 Hashtag
Web Hashtag Hashtag Twitter Hashtag Twitter Hashtag Hashtag Hashtag Twitter Hashtag Twitter Hashtag contexthashtag contexthashtag Hashtag contexthashtag Hashtag Hashtag Hashtag Twitter 2. Hashtag contexthashtag 2. 1 Hashtag Twitter Hashtag Hashtag 3 Hashtag Twitter Twitter ricebowl Twitter Hashtag ricebowl Hashtag Hashtag Hashtag Hashtag Hashtag Hashtag Hashtag Hashtag Hashtag 2010 12 31 NHK Hashtag nhk kouhaku61 kouhaku Twitter Hashtag Hashtag Hashtag Hashtag Hashtag Hashtag 歌 手 倉 木 麻 衣 #Mai #Mai&Ayumi 浜 崎 あゆみ #Ayumi 20062007200820092010 開 催 日 時 年 ) 倉 木 麻 衣 #Mai 1 contexthashtag #Mai&Ayumi[2009, 1, 1, ]t 020062007200820092010 開 催 日 時 年 ) 1 浜 崎 あゆみ #Ayumi[2010, 0, 1, ]t 1 [2007, 1, 0, ]t 2 contexthashtag Hashtag Twitter 2. 2 contexthashtag contexthashtag Hashtag Hashtag Hashtag Hashtag contexthashtag Hashtag contexthashtag Wikipedia 4 Hashtag Hashtag Hashtag contexthashtag contexthashtag 1 1 contexthashtag 1 1 contexthashtag Hashtag Ayumi 2010 Mai 2007 Mai&Ayumi 2009 AKB AKB48 2010 3 http://hashtagsjp.appspot.com/ 4 http://ja.wikipedia.org/wiki/
Hashtag 1 contexthashtag ) 1 1 Hashtag Ayumi ) =, 2010) Ayumi Mai& Ayumi Mai 2 contexthashtag Hashtag Hashtag Mai&Ayumi contexthashtag AKB AKB48 Hashtag AKB 2011 Hashtag Mai 2011 Hashtag contexthashtag Hashtag Hashtag contexthashtag Hashtag Hashtag Hashtag 3. contexthashtag Twitter contexthashtag Hashtag Hashtag Hashtag contexthashtag Hashtag 3 1 20 1 URL 3 Hashtag Hashtag 5 hashtagsjp Hashtag Hashtag Hashtag Hashtag Hashtag Hashtag Hashtag 2 Hashtag contexthashtag Hashtag Hashtag contexthashtag Hashtag 3. 1 Hashtag Hashtag Hashtag contexthashtag Hashtag Hashtag Hashtag Hashtag Hashtag Hashtag Hashtag Hashtag 3. 2 contexthashtag contexthashtag [5] contexthashtag Hashtag Hashtag contexthashtag contexthashtag Hashtag Hashtag h i N Hashtag H H = {h 1, h 2,, h N contexthashtag g k M contexthashtag G G = {g 1, g 2,, g M contexthashtag g k n k E k = {E k1,, Eknk E k j Hashtag H contexthashtag H = H 1 H 2 H N g k H k h k j Hk h k j {e = k 1,, ek n k ) e k i Ẽi k Ei ki = 1,, n k) 1) h k j gk 2) h k j contexthashtag gk Hashtag Ẽ k i E k i contexthashtag g k Ig k 5 http://hashtagcloud.net/
Ig k Hashtag h k j ek i contexthashtag g k Ei k 3. 3 Hashtag contexthashtag Hashtag 1) E k 3 contexthashtag contexthashtag g k Hashtag h k j p j e j f j h k j = p t j e j f j t 3) t 3. 3. 1 Geocoding API 6 Geocoding API XML 35.661913 139.700943 contexthashtag g k Hashtag p j p jx, p jy ) p j = p jx, p jy 4) 3. 3. 2 contexthashtag g k Hashtag f j f ji i = 1,..., n k ) f j = f j1,..., f jnk 5) 3. 4 Hashtag Hashtag Hashtag Hashtag Hashtag Hashtag Hashtag contexthashtag Hashtag h i Hashtag w j 2 h i w j S imh i, w j ) 6) 0 < = S imh i, w j ) < = 1 Hashtag h i p i e i f i Hashtag w j p j e j f j S imh i, w j ) = S im p p i, p j ) S im f f i, f j ) S im e e i, e j ) 6) S im p p i, p j ) 7) S im f f i, f j ) 8) S im p p i, p j ) 1 S im f f i, f j ) S im p p i, p j ) = S im f f i, f j ) = 1 1 + dp i, p j ) f i f j f i f j dp i, p j ) 7 100km ) S im e e i, e j ) 9) α F x e i, e j ) x contexthashtag ) e i e j S im e e i, e j ) F x e i, e j ) 1 7) 8) f ji Hashtag i 1 Hashtag i ) f ji = 0 ) contexthashtag $concert Hashtag UH WL $concert f j =, AKB48, UH WL 2010 12 8 Hashtag 35.512228, 139.620165, 2010/12/8, 0, 0, 1 ) 1 1 α S im e e i, e j ) = 9) F x e i, e j ) + 1 α 1 α = 4 4. contexthashtag Twitter 4. 1 6 http://www.geocoding.jp/ 7 http://yamadarake.web.fc2.com/trdi/2009/report000001.html
1. 前 処 未 理 来 イベントHashtagを イベント 推 薦 システム 取 得 Hashtag 検 索 サイト メッセージを 利 STEP1 3.contextHashtag 2. contexthashtagに 基 本 データをもとに 空 分 間 対 類 内 応 に する 位 置 付 け 基 URL 本 イベントHashtag データ,メッセージ ログイン 用 者 の 過 去 取 イベントHashtag 得 付 き contexthashtag DB 利 用 者 Hashtagと 利 STEP2 類 似 用 度 者 を が 使 未 用 来 した イベントHashtag 全 ての 過 去 イベント 間 の ユーザDB イベント とメッセージ 情 報 絞 STEP2で STEP3 り 込 み算 出 した 値 を 閾 値 により Twitter 3 3 Twitter Hashtag Twitter contexthashtag Hashtag contexthashtag contexthashtag Hashtag Twitter Hashtag Hashtag Hashtag URL Hashtag contexthashtag Hashtag contexthashtag STEP1 Hashtag STEP2 STEP1 contexthashtag Hashtag Hashtag STEP3 STEP2 Hashtag 4. 2 contexthashtag 4. 2. 1 3 contexthashtag 4 2 3 Hashtag Hashtag 100 Hashtag 3 Hashtag contexthashtag 9 9 3 1990 2010 緯 度 経 度 A1 B 1 C D E F G H I 開 催 日 2006 1999 2010 時 1 2003 1998 1992 2010 1996 1 1 1998 35.689506 1 1991 139.691701 2000 1 1 1995 2005 35.443708 139.638026 0 0 0 0 0 0 10 10 01 1990 1998 4 ) 4. 2. 2 4 5 6 4 Hashtag 1996MaiAyu Hashtag 3 5 Hashtag 1992S azan Hashtag 3 6 Hashtag 1991S map Hashtag 3 4 5 6 SMAP Hashtag 4 5 6 Hashtag 5 2 Hashtag Hashtag ) 1996MaiAyu 1996 1992S azan 1992 1991S map SMAP 1991 3, 35.689506,139.691701 35.443708,139.638026 35.011636,135.768029 34.693738,135.502165 43.062096,141.354376 38.268215,140.869356 35.181446,136.906398 33.590355,130.401716
4 1 ) EXILE 2000 0.546 2006 0.388 2010 0.359 5 2 ) 1998 0.615 1997 0.468 1993 0.286 6 3 ) SMAP 2001 0.549 SMAP 1993 0.439 SMAP 1999 0.408 4 5 6 contexthashtag 5. 5. 1 contexthashtag Java Twitter API Java Twitter4J contexthashtag Hashtag 2010 5 2010 8 Hashtag Hashtag contexthashtag 680 150 contexthashtag Hashtag 20 2010 8 Hashtag25 Hashtag Hashtag Hashtag 0.05 contexthashtag halcali contexthashtag f j =,, halcali, Per f ume,,, ) halcali HALCARI 2011 2 13 Hashtag 35.655289, 139.704536, 1, 0, 0,, 2011 6 Hashtag Hashtag F contexthashtag Hashtag Hashtag 3 1 ) 4 ) 1 2 3 5. 2 5 7 8 7 F F Hashtag 8 Hashtag 7 0.269 0.247 0.431 0.611 F 0.309 0.351 8 2.67 3.5 2.67 0.47 0.5 0.47
5 1 contexthashtag 2 6. contexthashtag Twitter contexthashtag Hashtag Hashtag Hashtag Twitter contexthashtag Hashtag Twitter hashtagsjp [1] Akshay Java, Xiaodan Song, Tim Finin, Belle Tseng Why we twitter: understanding microblogging usage and communities WebKDD/SNA- KDD 07 Proceedings of the 9th WebKDD and 1st SNA-KDD 2007 workshop on Web mining and social network analysis, pp. 56-65, 2007. [2] Owen Phelan, Kevin McCarthy, Barry Smyth Using Twitter to Recommend Real-Time Topical News RecSys 09 Proceedings of the third ACM conference on Recommender systems, pp. 385-388, 2009. [3] Miles Efron Hashtag Retrieval in a Microblogging Environment SIGIR 10 Proceeding of the 33rd international ACM SIGIR conference on Research and development in information retrieval, pp. 787-788, 2010. [4], Twitter Tweet, Web, 2010, 7, pp. 31-32, 2010. [5], Vol.19, No.3, pp.365 372, 2004.