wki_shuronn.pdf

Size: px

Start display at page:

Download "wki_shuronn.pdf"

さなえながおか
7 years ago
Views:

1 No. 161 Sentiment Extraction from Live Tweets

2 Twitter Twitter Summary Recently, microblogs such as Twitter become popular, and we can tweet about our own daily life easily. The user who tweets during watching sports programs. dramas, and movies on TV becomes increase. In this way, users can share their sentiment with other people who watch the same program on TV by tweeting some scene of the program in real time. Furthermore, user has different sentiment on the timeline based on a scene of a program. In this paper, we propose extracting users sentiment from live tweets. In this time we use the tweets which tweet about movie programs on the TV. Tweets have emoticon and special words. Users often use them to present their sentiment on the tweet. Then we also extract sentiment of the emoticon and special words and we include these sentiment to the tweet sentiment.

3 Web

4 Juman

6 1 Twitter 1 Twitter 140 Twitter 1 2 1,000 Facebook 19.2 mixi 19.5 Twitter 37.2 Twitter 1: 1 Twitter

7 1 (ˆoˆ) (ˆoˆ) 3 (ˆoˆ) (ˆoˆ) (ˆoˆ)

8 2 Plutchik[1] 8 4 [2] [3] [4] [2] ( ) [5] ( ) neutral 11 [6] [7] Twitter [8] Twitter Ekman [9] Twitter ( ) 4 [10] [11] 1 1 [20] [21] [22] [24] [28] [29] 3

9 1: + [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [20] [28] [29] [30] [31] [12] [13] [14] [16] [30] [12] 4 [13] 3 [30] 14 5 [14] + [32] ( ) 3 4

10 [3][33] : 5

11 3.1.1 Web Twitter 100 Web Web URL [3] 3 PHP Twitter Search API 3 3: (RT). 2. # URL 6

12 2: 7

13 3. 4. (3) 5. (1) (4) Twitter (ˆoˆ) orz Twitter

14 3: 9

15 4: 5: 6: 10

16 4:

17 5: 12

18 (T-T) (;o;) ( ) 0.4 ( ) (ˆoˆ) 10 (ˆoˆ) (ˆoˆ) (ˆoˆ) [34] 3 ( )

19 7: 8: 7 9: 7 14

10: 3 11 11 (*ˆˆ*) ((((; )))))) ( ˆ ˆ ) ( ˆ ˆ

20 10: (*ˆˆ*) ((((; )))))) ( ˆ ˆ ) ( ˆ ˆ ) (* ` *) ( ) ( ` ) (T T) ( ) 4 2 (ˆoˆ)!! ( ) 15

21 ( `) TIM e i = DIMi α (1) TIM s j = DIMj β (2) TIM r k = DIMk γ (3) i j k TIM e i i TIM s j j TIM r k k DIMi i DIMj j DIMk k α, β, γ α>γ>β F 7 α =2.0 β =0.2 γ =0.3 F 33% 35% F F 16

22 11: 3 6: (ˆOˆ)/ = (* `*) + ( ` ) (; ;) +( ) ( ) + ˆˆ 17

23 12: 18

24 7: 33% 46% 40% 38% 11% 25% 32% 47% 20% 43% 34% 53% 53% 42% F 39% 28% 42% 36% 18% 34% 33% 38% 64% 44% 36% 13% 25% 37% 63% 22% 58% 21% 56% 47% 45% F 48% 33% 50% 27% 21% 32% 35% [3] Twitter 3 [2] Yahoo! 3 74, Plutchik [1] Juman Juman 13 / 3 Yahoo! 19

25 8: 13: Juman 20

26 9:

27 5.2 [32] (1) (1) 5.4 TW i,j = 1 h (R NR) (4) h 10 i=1 TW i,j i j h R j i 22

28 10: ( ) 67% 2% 31% 35% 10% 55% 27% 2% 71% 29% 10% 61% 80% 2% 18% 33% 14% 53% 100% 0% 0% 84% 4% 12% 8% 41% 51% 14% 0% 86% NR j i TW i,j j TW i,j j Twitter 10 TW i,j % % 23

29 11: % 14.3% 12.9% 45.5% 27.2% 27.3% 23.5% 20.6% 55.9% 27.0% 29.7% 43.2% 49.4% 31.6% 19.0% 20.8% 37.5% 41.2% 71.4% 24.5% 4.1% 20.5% 31.3% 48.2% 4.3% 26.1% 69.6% 0% 11.1% 88.9% i TW i TW i =(SS i SW i )/ max(ss i,sw i ) (5) (5) SS i i ( ) SW i i ( ) TW i

30 14: 25

31 , : % 17 0% % 2 0% 4 0% 16 0% 6 0% 14 0% % %

32 2. 3. F : F : F F 27

33 F F F F F 6 ONE PIECE FILM Z

34 15:

35 16:

36 Yahoo! Twitter 3 Twitter Twitter

37 (DEIM2013), ARG Web (ARG SIG-WI2) (DEIM2014), 2014(to appear) (to appear). 32

38 5 (DEIM2013), (DEIM2014), 2014(to appear). [1] R.Plutchik The nature of emotions American Scentist Vol.89 pp [2] 1993 [3] Vol. J94-D No.3 pp [4] Web (WebDB2011) 9 pages [5] Web, Vol.50 pp [6] (TOD) Vol.46 No.SIG10(TOD28) pp [7] 19 (WISS2011) 2011 [8] Twitter 2012 [9] FIT 2011 pp [10] Twitter TV 26 IC-R [11] Twitter 3 (DEIM Forum 2011) F

39 [12] 37, pp , 2003 [13],, : 22(2), pp 35 43, 2006 [14],. DD, [ ] 2012-DD-85(1), 7 pages, 2012 [15], (4), pp , 2004 [16],,, 13, pp [17],, 51(3), pp , 2010 [18],,. 2007(6), pp 13 18, 2007 [19],. DE, 106(150), pp , 2006 [20],,,. 2003(23), pp , 2003 [21] M.Ptaszynski (3), pp , 2012 [22] CHO HEERYON,,,,. ICS,[ ] 2006(110), pp 1 8, 2006 [23],,,,, II. GN, [ ] 2008(31), pp , 2008 [24] 7, pp 67 82, 2008 [25],,,, Ho,Tu Bao, Dam,Hieu Chi 18 (NLP2012),pp ,

40 [26], 2011(1), pp , 2011 [27] : KARP 6, pp 29 46, 2004 [28],. GN, [ ] 2008(31), pp , 2008 [29],,, : 22(4), pp 31 39, 2007 [30], 1, pp , 2010 [31],,. [ ] 2008(69), pp 51 56, 2008 [32] 5 (DEIM Forum 2013) A [33] No.3 pp [34] 17 pp

Vol. 28 No. 2 Apr. 2011 173 1. 1 Web Twitter/Facebook UI 4 1. 2. 3. 4. Twitter Web Twitter/Facebook e.g., Web Web UI 1 2 SNS 1, 2 2

Vol. 28 No. 2 Apr. 2011 173 1. 1 Web Twitter/Facebook UI 4 1. 2. 3. 4. Twitter Web Twitter/Facebook e.g., Web Web UI 1 2 SNS 1, 2 2 172 SNS Web Web As social web sites such as blog and SNS(Social Network System) became popular, many people have communicated with their friends on the Web. Meanwhile, several problems of social web sites