ARG WI2 No.6, 2015 a b b 565-0871 2-1 a) yoshitake@nanase.comm.eng.osaka-u.ac.jp b) {naoko, babaguchi}@comm.eng.osaka-u.ac.jp 1 Citizen Sensor [1] Twitter 140 Twitter Sakaki [2] [3] Massoudi [4] [5] Copyright is held by the author(s). The article has been published without reviewing. 2 2 q Twitter q
Web 1 q q 2 1 2 Step1) Twitter Step2) (w i, w j ) S(w i, w j ) Step3) q 2 2 2.1 I Twitter MeCab[6] URL http:// @ 2.2 (w i, w j ) S(w i, w j ) I w i w j 1 2 w i w j B(w i, w j ) B(w i, w j ) w i w j B(w i, w j ) β (w i, w j )
Proceedings of ARG WI2 3 4 S(w i, w j ) (w i, w j ) (w i, w j ) S(w i, w j ) S(w i, w j ) = 1 (w i, w j ) S x (w i, w j ) S x+1 (w i, w j ) S x+1 (w i, w j ) = c 1 S x (w i, w j ) d1 (1) d 1 < 1 (w i, w j ) S x+1 (w i, w j ) S x+1 (w i, w j ) = c 2 S x (w i, w j ) d2 (2) d 2 > 1 (1) (2) 3 d 1 < 1 (1) S max d 2 > 1 (2) 0 (2) (1) c 1 c 2 S(w i, w j ) S max c 1 = S (1 d 1) max (3) c 2 = S (1 d2) max (4) S(w i, w j ) < 1 (w i, w j ) 2.3 q q q q q w q S(q, w) 1 w a) q q q b) q q q c) 4
Web MeCab SVM 3 2013/10/25 2013/11/27 32 21,134,159 I 24 S max = 10 β = 2.0 d 1 = 0.4 d 2 = 1.5 2013/11/27 5 227 354 11/7 11/16 q 10 30 50 P N AP N P N = R N (5) AP N = 1 N R k=1 (P k rel (k)) (6) R N rel(k) 1 q P 10 P 30 P 50 AP 10 AP 30 AP 50 11/7 1.00 0.87 0.86 1.00 0.98 0.94 11/16 1.00 0.93 0.84 1.00 0.97 0.94 11/7 0.60 P 24 = 0.63 0.87 AP 24 = 0.75 11/16 0.00 0.03 0.02 0.00 0.04 0.04 11/7 0.80 0.57 0.36 0.86 0.78 0.76 11/16 0.60 0.77 0.78 0.50 0.67 0.70 11/7 0.30 0.13 0.22 0.32 0.30 0.22 11/16 0.70 0.83 0.78 0.79 0.81 0.81 11/7 0.20 0.27 0.38 0.18 0.25 0.31 11/16 0.30 0.50 0.56 0.30 0.43 0.48 k 1 0 1 q 11/7 24 10 24 q 11/7 q 2 S(q, w) 1.0 w 1-1 1-4 q 1-2 1-3 q q q 1-3 1-4 1-5 1-6 q S(q, w) 1.0 w 11/7 11/16 q 3 4 q 5 6 q 7 8 q 9 10 q 11/7 2-1
Proceedings of ARG WI2 2 11/7 q 1-1 1 unko kanto JR ( ) (11/07 09:30) # # 1-2 21 11/07 17:15 # #Kanto 16:43 (17:09) Y378 #TrainDelay 1-3 36 1-4 37 1-5 1-6 2355 3 11/7 q 2-1 4 2-2 6 @(ID) ( ˆ ˆ;) 2-3 19 - NHK (URL) 4 11/16 q 3-1 1 (URL) 3-2 5 (URL) 3-3 26 16 06 5 11/7 q 4-1 1 [22:18 ] N37.1 E140.7 10km M3.9 (URL) #earthquake 4-2 22 @(ID) 6 11/16 q 5-1 5 FNN 16 3 58 3 (URL) 5-2 21 5.8 4 17 19 (URL) 5-3 37 M5.5 4 7 11/7 q 6-1 3 #AmazonJP # (URL) 6-2 6 6-3 8 #8: 10 KM-012 (URL) # #amazon 8 11/16 q 7-1 4 1 35 4 16 1 35 4 ( ;) 7-2 32 16 20 44 3 7-3 38 - TBS News (URL) # 2-2 2-3 11/16 3-3 3-1 3-2 3-1 3-2 3-1
Web 9 11/7 q 8-1 4 by 8-2 6 (URL) #FNN 8-3 19 @(ID) 10 11/16 q 9-1 15 2-2 2013/11/16 # (URL) 9-2 18 92 1-1 9-3 26 3-8 (URL) q 4-1 4-2 5-1 5-2 5-3 1 11/7 11/16 11/16 5-3 q 11/7 6-1 6-3 6-2 11/16 7-1 7-2 7-3 q 11/7 8-2 8-1 8-3 11/16 9-1 9-2 9-3 11/7 4 Twitter 2013 32 Twitter [1] Sheth, A.: Citizen Sensing, Social Signals, and Enriching Human Experience, IEEE Internet Computing, Vol. 13, No. 4, pp. 87-92, 2009. [2] Sakaki, T., Okazaki, M. and Matsuo, Y.: Earthquake Shakes Twitter Users: Real-time Event Detection by Social Sensors, Proc. WWW, pp. 851-860, 2010. [3] IFAT Vol. 111, No. 31, pp. 1-6, 2013. [4] Massoudi, K., Tsagkias, M., Dijke, M. D., et al.: Incorporating Query Expansion and Quality Indicators in Searching Microblog Posts, Proc. ECIR, pp. 362-367, 2011. [5] Twitter DEIM forum C9-5, 2013. [6] MeCab Japanese morphological analyzer, https://code.google.com/p/mecab