Twitter ( ), ( ). i

Size: px

Start display at page:

Download "Twitter ( ), ( ). i"

けいしょうちゅうか
5 years ago
Views:

1 ( : A9TB2251)

2 Twitter ( ), ( ). i

3 A 25 ii

4 1 SNS( ) SNS SNS Twitter( ) SNS Twitter 140 SNS 1.4 ( ) Twitter 140 Twitter, Twitter Twitter [9, 10] HP 2 [1, 2] [3, 4] [6], 1

5 [7] Twitter [5] ( )

6 5 6 3

7 2 2.1 [1] Twitter [4] Twitter : RT RT URL web URL # # URL URL 4

8 URL URL, 2.2 [2] [1], [7] Twitter Twitter 5 Score = l(n ) n( ) (2.1) 8 5

9 3 2.3 [2] SVM URL Carlos [8] Carlos Twitter URL URL 6

10 3 3.1 (2011/3/11 9:00 201/3/18 9:00) ( 1 8 ) 3.2 [5].., 2.1 ( ) 7

11 3.1: ( ) ONEPEACE w 2.1 D P (w D w) = P (w D) P (w) = w w (3.1). 2.2 Score(s, t) = P MI(t, w) hist(len s t) (3.2) w C s w t PMI( ) s t Cs s PMI 2 lens s hist(l, t) s t l 14 ( 3.1) ( ) ( AND ) 8

12 1 3.3 ( ) ( ) ( ) Web URL ( ) ( ) ( 3.1) x x 9

13 [1, 2, 4, 5, 6, 7] 10

14 4, (TTC: Time To Correction) (TTS: Time To Suppress) (TTE: Time To End)

15 4.1: TTC( ) TTS( ) TTE( ) ONEPEACE ( )

16 4.1: 4 4.2: 4 ( ) 100 ONEPEACE

17 , 14

18 [5]. 5.1 [2, 5, 6, 7] (T): 121 ( A) Bag of words (B): ( ) URL (U): web URL URL 15

19 (R): (TW): 5 (D): ( ) ( ) 2.1 ( ) (SU SB): ( SU SB) ( 3.1) True Positive True Negative, False Positive False Negative ( 5.1) Precision, Recall, F1 5.1, 5.2, 5.3 P recision class = T ruep ositive class T ruep ositive class + F alsep ositive class (5.1) Recall class = T ruep ositive class T ruep ositive class + F alsenegative class (5.2) F 1 class = 2 P recision class Recall class P recision class + Recall class (5.3) 16

20 5.1: True Positive False Positive False Negative True Negative Accuracy ( 5.4) Precision Recall F1 ( 5.5, 5.6, 5.7) Accuracy = (T P + T N ) + (T P + T N ) + (T P + T N ) 3 AllT weet (5.4) P recision = P recision + P recision + P recision 3 (5.5) Recall = Recall + Recall + Recall 3 (5.6) F 1 = F 1 + F 1 + F 1 3 (5.7) (T) F1 ( 8 ) F ( ) Bag of words (7 ) F Bag of words URL 0.81, F URL (U) Bag of words 17

21 5.2: Accuracy Precision Recall F1 T( ) B(Bag of words) SB( ) SU( ) D( ) TW( ) (B ) T( ) R(RT ) U(URL ) Bag of Words (TW) URL, Web F1 F1 18

22 5.3: (Bag of Words 7 ) 2.73 (SU) 1.25 (T=True) 2.05 (SB) 0.79 (TW[0]= ) 0.82 (T=False) 0.69 URL (U=True) 0.71 (R=Ture) (R=False) (T=False) (T=True) (TW[1]= ) (TW[2]= ) (SU) 5.4: Precision Recall F1 T( ) ( Bag of Words) T( ) ( Bag of Words) T( ) ( Bag of Words) ( B(Bag of words))

23 5.1: 5.3 URL cc RT

24 URL RT ( ) ( ) ( ) 21

25 6 Bag of words URL URL 22

26 Twitter Japan 23

27 [1],,,,, Vol54, [2],,,, 18, [3],,,,, 10, [4],,,, Twitter RT,, [5],,,,, 19, [6],,,,,, Twitter, 26, [7],,,, Vol3, [8] Carlos Castillo, Marcelo Mendoza, Bardara Poblete, Insormation Crebidility on Twitter, International World Wide Web Conference, 2011 [9],,, [10] (ispp), 3.11,,

28 A A.1: 25

Twitter‡Ì”À‰µ…c…C†[…g‡ðŠŸŠp‡µ‡½…^…C…•…›…C…ﬁ‘ã‡Ì…l…^…o…„‘îŁñ„�™m

Twitter‡Ì”À‰µ…c…C†[…g‡ðŠŸŠp‡µ‡½…^…C…•…›…C…ﬁ‘ã‡Ì…l…^…o…„‘îŁñ„�™m 27 Twitter 1431050 2016 3 14 1 Twitter,,.,.,., Twitter,.,,.,,. URL,,,. BoW(Bag of Words), LSI(Latent Semantic Indexing)., URL,,,,., Accuracy, AUC(Area Under the Curve), Precision, Recall, F,. URL,,,.,