DEIM Forum 15 A5-1 35 855 1- E-mail: {yamahei,satoh}@ce.slis.tsukuba.ac.jp Twitter Twitter 1. Twitter 1 Twitter Twitter Twitter [15] 1 1 5 1 3 7 3 1https://twitter.com/ https://twitter.com/who to follow/interests 3http://naojcamp.mtk.nao.ac.jp/phenomena/151/ Number of tweets May 1, 1 weekday Annular Solar Eclipse : 3: 6: 9: 1: 15: 18: 1: : time 1 1 5 1 [1]
Twitter 3 5 6. Twitter. 1 Twitter Weng [1] Twitter PageRank TwitterRank Cha [] Twitter 1 Pal [7] authority non-authority authority Sae-Trumper [8]. Twitter [1] [13] Yamaguchi [11] Twitter 3. 3. 1 i s n i,s S N i = {n i,1, n i,,, n i,s} n i,s µ i σ i N i µ i σ i µ i = 1 S n i,s, σ i = 1 S (n i,s µ i ). (1) 1 i,s i,s = n i,s µ i σ i. () i,s s i s b i,s 1 if i,s > 3σ i, b i,s = otherwise, i B i = {b i,1, b i,,, b i,s} 3. (3) 1 [1] 3.. 1 [1]
3.. [3] 1 i,s i,s 3. 3 3. i s b i,s s k p s,k i k S b (i, k) S b (i, k) = b i,s p s,k. () i k s S b (i, k) 1 i s i,s S (i, k) = i,s p s,k. (5) i s i,s k s k Mean 15 1 5 5 1 15 Standard deviation µ i σ i.. 1 1 5 1 13 3 1 Twitter Search API s 1 µ i σ i 1 1. 51,889 1 µ i σ i 1 9 61,6 1 51,89. 3 1 5 1 1 6 6 AKB8 1 1 31 13 1 1 13 3 8 http://search.twitter.com/search.json
Number of bursts 5 5 35 3 5 15 1 5 1-5 Annular solar eclipse 3 AKB8 general election New year WBC 1-6 1-7 1-8 1-9 1-1 1-11 1-1 13-1 13-13-3 13-13-5 WBC. 3 i,s α β [] α = 5 β =.1 K 1 [9] 1%.58 6 s p s,k 1 5 1 (a)(b)(e)(f) (c)(d) (f) 1 9 5 (a)(b)(e)(f) (c)(d) 1 3.. 3 6 S b S i m(i) m(i) = S b i,s rel(s, i) S bi,s (6) rel(s, i) i s 1 1 6 i m(i) = 1 m(i) = ndcg [5] ndcg ndcg l = DCG l (7) IDCG l l m(ir) 1 DCG l = log (1 + r) r=1 i r r IDCG l l DCG l l 1 < = l < =. 5 ndcg ndcg.71 (a) ndcg.9 (d) (b)(c)(d)(e)(f) S b (b)(c)(f) (a)(d)(e) 1 Twitter (a) @eva information (d) @Nintendo i,s 5 6 i,s i,s i,s (a) i,s 1 5 1 11 1 1 11
.9.8.7.6.5..3..1 1-5.9.8.7.6.5..3..1 1-5.9.8.7.6.5..3..1 1-5 1-6 1-7 1-8 1-9 1-1 1-6 1-7 1-8 1-9 1-1 1-6 1-7 1-8 1-9 1-1 1-11 (a) 1-11 (c) 1-11 (e) Spearman s ρ =.667 1-1 13-1 13-13-3 13-13-5 Spearman s ρ =.68 1-1 13-1 13-13-3 13-13-5 Spearman s ρ =.81 1-1 13-1 13-13-3 13-13-5.8.7.6.5..3..1 1-5.9.8.7.6.5..3..1 1-5.7.6.5..3..1 1-6 1-7 1-8 1-9 1-1 1-6 1-7 1-8 1-9 1-1 1-5 1-6 1-7 1-8 1-9 1-1 1-11 (b) 1-11 (d) 1-11 (f) Spearman s ρ =.685 1-1 13-1 13-13-3 13-13-5 Spearman s ρ =.885 1-1 13-1 13-13-3 13-13-5 Spearman s ρ =.99 1-1 13-1 13-13-3 13-13-5 1 (a) (b) (c) (d) (e) (f) 1 #eva #FNS #chukoi #NintendoDirectJP # #AKB8 #ntv #fujitv # #nyaruko # #AKB 3 #join eva #agqr #tx kodokugurume #agqr #ntv #AKBvote1 # #TFBJP #tokyomx #followback # #akbsenkyo 5 #evangellion # #tvtokyo #E3 # #akb8 ndcg@ (a) (b) (c) (d) (e) (f) S b.731.99.91.938.799.958.69.913.919.95.86.87 S.71.87.813.939.789.958.738.87.86.97.88.951.71.89.865.9.86.98 (d) i,s i,s 5. 5. 1 (a) ndcg@.8 6 5 S b i,s S ndcg (c) S b S.1 ndcg (c) S (c) 1 9 13 1 [11] i,s i,s
Cosine similarity.9.8.7.6.5..3..1 1-5.9.8.7.6.5..3..1 1-5.9.8.7.6.5..3..1 1-6 1-7 1-8 1-9 1-1 1-11 1-1 13-1 13-13-3 13-13-5 (a) @eva information i,s 1-6 1-7 1-8 5 1-9 1-1 1-11 1-1 5.5 3.5 3.5 1.5 1.5 -.5 value.9.8.7.6.5..3..1 1-5 1-6 1-7 1-8 1-9 1-1 1-11 1-1 13-1 13-13-3 13-13-5 (a) i,s (a) S () (d) @Nintendo i,s 6 13-1 13-13-3 13-13-5 7 6 5 3 1-1 value.9.8.7.6.5..3..1 1-5 1-6 1-7 1-8 1-9 1-1 1-11 1-1 13-1 13-13-3 13-13-5 (d) i,s (d) S b () Pearson s r =.8 Cosine similarity.7.6.5..3. 18 16 1 1 1 8 6-1 1 8 6 - Pearson s r =.3 value value.5 1 1.5.5 3 3.5 Proposal score (S :).1..3..5.6.7.8.9 Proposal score (S b :) (a) 7 (d) (f) ndcg (f) 1 9 (f) 1 AKB8 1 9 18 AKB8 9th 5 AKB8 (f) ndcg (d) 13 6 @Nintendo 13 17 3 Nintendo Direct Luigi special 6 5. 56 5http://ja.wikipedia.org/wiki/AKB8 9th 6http://ja.wikipedia.org/wiki/Nintendo Direct
MeCab [6] 1 1 51 5 7 (a)..8 (d) 6. JSPS 5811 [1] David M. Blei, Andrew Y. Ng, and Michael I. Jordan. Latent dirichlet allocation. The Journal of Machine Learning Research, Vol. 3, pp. 993 1, 3. [] Meeyoung Cha, Hamed Haddadi, Fabrcio Benevenuto, and Krishna P. Gummadi. Measuring user influence in twitter: The million follower fallacy. In Proceedings of international AAAI Conference on Weblogs and Social, ICWSM 1, pp. 1 17, 1. [3] A Cichocki, R Zdunek, A-H Phan, and S Amari. Nonnegative Matrix and Tensor Factoriations: Applications to Exploratory Multi-way Data Analysis and Blind Source Separation. John Wiley & Sons, Ltd, 9. [] Thomas L. Griffiths and Mark Steyvers. Finding scientific topics. The National Academy of Science, Vol. 11, pp. 58 535,. [5] Kalervo Järvelin and Jaana Kekäläinen. Ir evaluation methods for retrieving highly relevant documents. In Proceedings of the 3rd Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, SI- GIR, pp. 1 8, New York, NY, USA,. ACM. [6] Taku Kudo. Yet another part-of-speech and morphological analyer. http://mecab.sourceforge.net/, 5. [7] Aditya Pal and Scott Counts. Identifying topical authorities in microblogs. In Proceedings of the Fourth ACM International Conference on Web Search and Data Mining, WSDM 11, pp. 5 5, New York, NY, USA, 11. ACM. [8] Diego Sae-Trumper, Giovanni Comarela, Virgílio Almeida, Ricardo Baea-Yates, and Fabrício Benevenuto. Finding trendsetters in information networks. In Proceedings of the 18th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD 1, pp. 11 1, New York, NY, USA, 1. ACM. [9] Charles Spearman. The proof and measurement of association between two things. The American Journal of Psychology, Vol. 15, No. 1, pp. 7 11, 19. [1] Jianshu Weng, Ee-Peng Lim, Jing Jiang, and Qi He. Twitterrank: finding topic-sensitive influential twitterers. In Proceedings of the 3rd International Conference on Web Search and Data Mining, WSDM 1, pp. 61 7, 1. [11] Yutaro Yamaguchi, Yamamoto Shuhei, and Satoh Tetsuji. Behavior analysis of microblog users based on transitions in posting activities. In Proceedings of International Conference on Information Integration and Web-based Applications & Services, iiwas 13, pp. 63 67, New York, NY, USA, 13. ACM. [1],,.. 7 Web WebDB Forum1, A-5, 1. [13],,,.. TOD, Vol. 6, No. 5, pp. 71 8, dec 13. [1],,,,..., Vol. 6, No. 3, pp. 73 89, jun 13. [15],,,,,. Twitter., Vol. 7, No., pp. 1 5, 13.