Indirect Tweet Indirect Tweet 2. 2 Tweet Simple Tweet Reply Mention Indirect Tweet Tweet Tweet Indirect Tweet Tweet Tweet Indirect Tweet Tweet Tweet 2

Size: px

Start display at page:

Download "Indirect Tweet Indirect Tweet 2. 2 Tweet Simple Tweet Reply Mention Indirect Tweet Tweet Tweet Indirect Tweet Tweet Tweet Indirect Tweet Tweet Tweet 2"

せぴあやがい
5 years ago
Views:

1 DEIM Forum 207 D6-4 Twitter Indirect Tweet Twitter Indirect Tweet Indirect Tweet TL Indirect Tweet Indirect Tweet Indirect Tweet. SNS Twitter Indirect Tweet Twitter Twitter Tweet Tweet Tweet Twitter Tweet Mention Reply Tweet Retweet Quote Tweet Indirect Tweet( ) Indirect Tweet Tweet Mention Reply Tweet Tweet Twitter Tweet Indirect Tweet Tweet Indirect Tweet Indirect Tweet Tweet Indirect Tweet 2 HomeTimeline Reply Retweet Tweet 50% Indirect Tweet Tweet Tweet Tweet Tweet Tweet Tweet Tweet Indirect Tweet Tweet Indirect Tweet Tweet Tweet Tweet Indirect Tweet Tweet Indirect Tweet Tweet 2. Twitter Indirect Tweet Tweet 2. Indirect Tweet Simple Tweet Simple Tweet Tweet Tweet Indirect Tweet Reply Mention Tweet Tweet Indirect Tweet Tweet Indirect Tweet Tweet Indirect Tweet Indirect Tweet Indirect Tweet Tweet HomeTimeline Tweet Reply Mention

2 Indirect Tweet Indirect Tweet 2. 2 Tweet Simple Tweet Reply Mention Indirect Tweet Tweet Tweet Indirect Tweet Tweet Tweet Indirect Tweet Tweet Tweet 2 Reply Indirect Tweet Indirect Tweet Tweet Tweet 2. 3 u SNS u U Tweet p = (user, time, text, class, number) SNS Tweet p user Tweet p i ID time Tweet p i text Tweet p i class Tweet type simpletweet reply mention retweet quotetweet number user Tweet Tweet, 2,... Tweet P = {p, p 2,...} F u u F u ID i HomeTimeline Tweet P i () P i = {p j Id(p j) = i Id(p j) F i} () Id(p) Tweet ID Tweet user (2) Id(p) = user (2) Type(p) Tweet Tweet class (3) Type(p) = class (3) Time(p) Tweet Time(p) = time (4) List(P, p, n) Tweet Tweet Tweet p Tweet n Tweet Indirect Tweet Tweet Indirect Tweet Tweet u, v v F u Tweet p i P Id(p i) = u Type(p i) = simpletweet Indirect Tweet Indirect Tweet Tweet Indirect Tweet I(P ) p i P Tweet p i I(P ) p i Indirect Tweet Indirect Tweet p i Tweet v F u p i I(P ) P Id(p j) = v Type(p j) = (simpletweet reply mention) Tweet Tweet Tweet p i Indirect Tweet Tweet R(p i, P ) p i, p j P p i I(P ) p j R(p i, P )[Id(p j) = v Type(p j) = (simpletweet reply mention)] Indirect Tweet Tweet ( R(p i, P ) > = ) p i I(P ) p j R(p i, P ) Time(p i) < Time(p j) p j = List(P, p i, j i) (5) 2. 4 Indirect Tweet u, v, ν U(u = v, u = ν, v = ν) u F ν v / F ν u F v v F u p i P, Id(p i ) = u p i I(P ) ν HomeTimeline P ν p i P ν Indirect Tweet p j P, Id(p j ) = v p j R(p i, P ) ν HomeTimeline P ν p j / P ν Indirect Tweet p i Tweet p j ν Indirect Tweet p i Tweet p j Tweet v F u 2 ν HomeTimeline P ν Tweet HomeTimeline Indirect Tweet p i Tweet p j P u Tweet p j v F u F u Tweet p i u % twitter.html

3 Simple Tweet Reply Mention 3. Indirect Tweet Tweet Tweet Tweet 3 u Tweet N u Noun(p) Tweet p p Noun(p) = N (8) Jaccard(N a, N b ) 2 Jaccard 2 N a, N b (9) 2 Jaccard(N a, N b ) = N a N b N a N b (9) Simpson(N a, N b ) 2 Simpson (0) N a N b 0 0 Simpson(N a, N b ) = N a N b min( N a, N b ) (0) 3. E u u, v U v E u v u u v M u u F u E u Text(p) Tweet Tweet text (6) Text(p) = text (6) Tweet T u u Tweet p Tweet T u = {p i Id(p i ) = u} UserTimeline Tweet Number(u, p) Number(u, p) u Tweet p Tweet number (7) Number(u, p) = number (7) N u Tweet Tweet TimeSim(c a, c b ) 2 c a, c b c a c b 3. 2 Tweet Tweet Tweet u U M u v M u v Tweet () T v = {p i Id(p i) = v, v M u} () Tweet T v Tweet v N v = {Noun(p i ) p i T v, Number(v, p i ) < = k} (2) k Tweet u (),(2) u N u Tweet u M u v M u j v = Jaccard(N u, N v) (3) j v u v 3. 3 Tweet Tweet Indirect Tweet Tweet

4 p i P Tweet Tweet {p j p j = List(P, p i, l), Id(p j) M u} (4) l =, 2,..., n(n ) Tweet Tweet Tweet N i = Noun(p i) (5) N j = Noun(p j) (6) (i < j) Tweet p i N i Tweet Tweet p j N j Simpson Tweet s j = Simpson(N i, N j ) (7) s j p i p j Tweet 3. 4 Tweet Indirect Tweet Tweet 3. 3 Tweet Tweet p i {p j } Tweet c i = Time(p i ) (8) c j = Time(p j ) (9) TimeSim() Tweet t j = TimeSim(c i, c j) (20) t j Tweet p i Tweet Tweet p j Tweet Timesim() 3. 5 Tweet Reply Tweet Indirect Tweet Indirect Tweet Tweet Indirect Tweet Tweet 2 f(x) = λz B ( z) C (2) B, C µ r = B r B r + C r (22) B C (23) m A r =, 2,...h(h ) B r = + m(r) (23) C r = + A m(r) (24) µ r 3. 6 Tweet Tweet Tweet Sim(v, p j) = j v + s j + t j (25) Sim() Tweet Tweet p j Tweet v M u Tweet Tweet j v (3) Tweet u v s j (7) Tweet p j p i Tweet t j (22) p j p i Tweet 3. 7 Tweet Tweet {p j p j = List(P, p i, l), Id(p j ) M u } Tweet 3. 6 n 4. Indirect Tweet Indirect Tweet Indirect Mention URL Tweet Tweet Indirect Tweet [2] [3] [4] [5] Belkaroui [6] Reply Mention Tweet Tweet URL Tweet Indirect Tweet Tweet Twitter [7] [] LSA(Latent Semantic Analysis : ) [2] LDA(Latent Dirichlet Allocation : ) [3]

5 Tweet Zhao [4] Twitter-LDA Steyvers [5] Author-topic LDA [6] Twitter-LDA Twitter-TTM Tweet Tweet Indirect Tweet Tweet Indirect Tweet [7] [8] [9] Reply Indirect Tweet Reply [] Tweet 5. 3 Indirect Tweet Tweet Indirect Tweet Twitter Tweet Indirect Tweet Mention Tweet 2 Indirect Tweet Tweet Tweet 5. Indirect Tweet Tweet Tweet 5. u hop Tweet Tweet Tweet Tweet Tweet Tweet HomeTimeline u Indirect Tweet 2 Twitter Tweet Tweet k = 200 Retweet Quote Tweet Tweet Tweet A = 50 Indirect Tweet Tweet 2 3 Tweet p i I(P ) Tweet 2 R(p i, P ) = 2 R(p i, P ) = 3 Tweet Tweet Tweet MRR(Mean Reciprocal Rank ) MRR ( Tweet ) (26) u Tweet / MeCab mecab-ipadic-neologd Tweet 3 Tweet 60 Tweet Reply Mention

6 (27) a g Tweet Tweet RR g = a g (26) MRR = R(p i, P ) k RR g (27) g= Tweet 2 MRR<= 0.75 R(p i, P ) MRR MRR Tweet R(p i, P ) (27) MRR (28) Max(p i, P ) = 5. 2 R(p i, P ) k w= w 5. 3 Indirect Tweet Tweet (28) Indirect Tweet Tweet Indirect Tweet Tweet Indirect Tweet 60% Tweet 2 Tweet 2 Twitter Indirect Tweet Indirect Tweet Reply Mention 5. u Indirect Tweet Reply Mention 3 u Indirect Tweet 38 Reply Mention 個数時間差 [ 分 ] Indirect Tweet Tweet u 22% Indirect Tweet Reply Mention 8 Reply Mention 5. 5 Tweet {p j p j = List(P, p i, l), Id(p j ) M u } M u v 3 Tweet Tweet ((25) ) 2 ( Tweet ) Tweet NormSim(v, p j ) = j v + s j + t j (29) t j t j 3 (3. 2 ) Tweet (3. 3 ) CharSim(v, p j) = j v + s j (30) 4 Tweet Tweet Tweet 組数時間 [ 分 ] Indirect Tweet Tweet 3 u Indirect Tweet 38 Reply Mention % 累積度数

MRR p i I(P) (3) TimeDefSim(p j ) = Time(p i ) Time(p j ) (3) 5. 6 4 Tweet R(p i, P ) 5 MRR Tweet MRR 5 MRR 4 5. 7 4 MRR MRR Tweet MRR 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.

7 MRR p i I(P) (3) TimeDefSim(p j ) = Time(p i ) Time(p j ) (3) Tweet R(p i, P ) 5 MRR Tweet MRR 5 MRR MRR MRR Tweet MRR 課題番号総合関連度正規化総合関連度文字情報関連度時系列関連度 4 ( =55) 4 MRR R(p i, P ) Tweet R(p i, P ) Tweet Tweet Tweet Tweet 4 R(p i, P ) Tweet MRR 5 Tweet MRR Tweet 2 Tweet 7,8,2 Tweet 7 Tweet Tweet Tweet Tweet ( ) Tweet Tweet Tweet 言及先投稿数総合関連度正規化総合関連度文字情報関連度時系列関連度 5 ( Tweet )

8 0 Tweet 30 4 Tweet Tweet Indirect Tweet Tweet Indirect Tweet Tweet Tweet Tweet Tweet Tweet Tweet Tweet Tweet Tweet Tweet Tweet Tweet Twitter [20] 6. Indirect Tweet Tweet Tweet Tweet 2 Tweet 3 Tweet Indirect Tweet Tweet Indirect Tweet Indirect Tweet Indirect Tweet Tweet Tweet Indirect Tweet [],,,,.. TOD. 203, vol. 6, pp [2] IMRAN Muhammad, et al. Extracting information nuggets from disaster-related messages in social media. Proc. of ISCRAM, Baden-Baden, Germany, 203. [3] TANAKA Yuko, SAKAMOTO Yasuaki, HONDA Hidehito. The impact of posting URLs in disaster-related tweets on rumor spreading behavior. System Sciences (HICSS), th Hawaii International Conference on. IEEE, 204. pp [4] SATYANARAYAN Ashwin, DAS Bk Sarthak, KRISH- NAN Divya. Analyzing Advertisements on Twitter during Valentine s Month. [5] ISBISTER Joseph. Exploring adolescents use of social networking sites and their perceptions of this can influence their peer relationships PhD Thesis. Institute of Education, University of London. [6] BELKAROUI Rami, FAIZ Rim, ELKHLIFI Aymen. Using social conversational context for detecting users interactions on microblogging sites. Revue des Nouvelles Technologies de l Information, 205. [7]. t-lda MPS 200-MPS-8(0) pp [8]. 5 pp [9]. AI 06(38) pp [0]. 20-NL-204(6) pp []. JSA03(0) pp [2] Scott Deerwester, Susan T.Dumais, George W.Furnas, Thomas K. Landauer, Richard Harshman, Indexing by latent semantic analysis, Journal of the American Society of information Science, 4(6), pp , 990. [3] Blei, D.M., Ng, A.Y. and Jordan, M, Latent Dirichlet allocation, The Journal of Machine Learning Research, Vol.3, pp , [4] W.X. Zhao, J.jiang, J.He, Y.Song, P.Achananuparp, E.-P. Lim, and X.Li, Topical keyphrase extraction from twitter, The Annual Meeting of the Association for Computational Linguistics 20, pp , 20. [5] M.Steyvers, P.Smyth, M.Rosen-Zvi, and T.Griffiths, Probabilistic author-topic models for information discovery, SIGKDD2004, [6]. Twitter 7() pp [7],,. Tweet.. DE,. 203, vol. 3, pp [8],. Twitter. IFAT. DEIM Forum 204. [9],,,. Twitter. IFAT. 204, vol. 204, pp. -6. [20]. 55(5) pp

main.dvi

main.dvi DEIM Forum 2018 J7-3 305-8573 1-1-1 305-8573 1-1-1 305-8573 1-1-1 () 151-0053 1-3-15 6F URL SVM Identifying Know-How Sites basedonatopicmodelandclassifierlearning Jiaqi LI,ChenZHAO, Youchao LIN, Ding YI,ShutoKAWABATA,