Detecting Topology of K-pop Stars on YouTube with Bigdata Analytics

Size: px

Start display at page:

Download "Detecting Topology of K-pop Stars on YouTube with Bigdata Analytics"

Jasper Norris
5 years ago
Views:

1 Detecting Topology of K-pop Stars on YouTube with Bigdata Analytics Min Song, Ph.D. Department of Library and Info. Science, Yonsei University Text & Social Media Mining Lab

2 Outline Background Research objectives Theoretical foundation Research design Methodology YouTube co-comments network with one-count YouTube co-comments network with all-count YouTube co-comments sentiment network Implication and Future work References 2

3 Background Trends in analyzing social media data as information sharing with recent big data popularity High attention to Korean cultural issue like K-pop and Korean drama The growing interest of issue detection and prediction for economical, social and cultural issue 3

4 4

5 Research objectives Who s who in K-pop on YouTube Applying the traditional LIS science methodology to social media analysis Analyzing sentiment of user comments Previous studies: using static statistical data (View count, Number of comment) Comparing the three methods Co-comment analysis with one-count Co-comment analysis with all-count Co-comment sentiment analysis 5

6 Theoretical foundation Author Co-citation Analysis Author co-citation Analysis (ACA) is a technique commonly used to identify authors in related fields. A bibliometrictool that analyses the relationships of pairs of cited authors within source documents. The underlying assumption of ACA is that the more two authors are cited together, the closer the relationship between them. Paper Title A Author, A Author, B References Author, A. (2012). Title1. Author, B. and Author, C. (2009). Title2. Author, D. (2013). Title3. Author, E., Author, F., and Author G.(2011). Title4 6

7 Paper A References Author, A. (2012). Title1. Author, B. and Author, C. (2009). Title2. Author, D. (2013). Title3. Author, E., Author, F., and Author G.(2011). Title4 Paper Title 1 Paper Title 2 Paper Title 3 Paper Title 4 Author A Paper content Author B Author C Paper content Author D Paper content Author E, F Paper content Topical links 7

8 Theoretical foundation Webometrics: Co-link analysis Web Co-link Analysis (WCA) and Author Co-citation Analysis (ACA) are sister techniques. (Onyancha, 2008) Thelwall(2004) defines co-link as an instance when two web pages have in-inks from a third page. 8

9 Theoretical foundation related work Chatzopoulou, G., Sheng, C., & Faloutsos, M. (2010, March). A first step towards understanding popularity in youtube. In INFOCOM IEEE Conference on Computer Communications Workshops, 2010 (pp. 1-6). IEEE. Identifying the popularity of Youtube video by verifying the video attribute and pattern through the time Visualizing the related videos network High correlation: viewcount, #comments, #ratings, #favorites Siersdorfer, S., Chelaru, S., Nejdl, W., & San Pedro, J. (2010, April). How useful are your comments?: analyzing and predicting youtube comments and comment ratings. In Proceedings of the 19th international conference on World wide web (pp ). ACM. Analyzing the comment score for checking the dependency between comment view and comment score Identifying the score of user comment using SentiWordnet thesaurus Predicting the community adaption of new comments through the feedback of predefined score of comment Thelwall, M., Sud, P., & Vis, F. (2012). Commenting on YouTube videos: From Guatemalan rock to el big bang. Journal of the American Society for Information Science and Technology, 63(3), Statistical analysis of comment and discussions of commenter Benevenuto, F., Rodrigues, T., Almeida, V., Almeida, J., & Ross, K. (2009). Video interactions in online video social networks. ACM Transactions on Multimedia Computing, Communications, and Applications (TOMCCAP), 5(4), 30. Identifying the information of user interaction using the video comments Detecting the user characteristics based on Youtube video attributes No existing studies deal with co-comment analysis using sentiment of comments 9

Research design Query = K-pop, Korean pop, S.M.

Comment extraction Language detection Preprocessing

Comments table Co-comment based Video Pair

Visualization Co-comment sentiment based Video

10 Research design Query = K-pop, Korean pop, S.M. Entertainment, YG Entertainment, JYP MySQL DB Comment extraction Language detection Preprocessing Comment sentiment score calculation YouTube table Comments table Co-comment based Video Pair Generation Co-comment based Video Network Visualization Co-comment sentiment based Video Network Visualization K-pop singer Dictionary K-pop singer in YouTube Topographic map 10

11 Data collection Number of YouTube Video: 3,004 (2,234) Number of Comments: 4,831,713 (avg. 2,162) Language Detection Algorithm Language Distribution Language No. of Comment Prop % English 4,035, % Korean 165, % Spanish 76, % Vietnamese 50, % 3.01% 1.86% 1.04% 0.61% 0.43% 9.69% Japanese 29, % Chinese 20, % Etc. 454, % 11

Methodology co-comment analysis V1, V2 (V1,

weight: 1 (V2, V4) weight: 1 V1, V2, V4 (V1,

12 Methodology co-comment analysis V1, V2 (V1, V2) User A V1 Pair (V1, V2) weight: 3 V1, V2, V3 (V1, V2), (V1, V3) (V2, V3) User B V2 (V1, V3) weight: 1 (V1, V4) weight: 1 (V2, V3) weight: 1 (V2, V4) weight: 1 V1, V2, V4 (V1, V2), (V1, V4) (V2, V4) User C V3 (V3, V4) weight: 0 V4 12

Methodology Sentiment analysis of lexical resource for opinion mining: SentiWordNet (http://sentiwordnet.

edu/) Allocated with a sentiment score from 0 to 1 of positive, negative or objective Strong partiality

13 Methodology Sentiment analysis of lexical resource for opinion mining: SentiWordNet ( Based on synsets from WordNet ( Allocated with a sentiment score from 0 to 1 of positive, negative or objective Strong partiality information terms are assigned with higher scores whereas less bias/subjective terms carry low scores. Positive=0.875, Negative=0, Obective=0.125 synset 13

Sentiment of YouTube Comments Maximum sentiment score: 22.

14 Sentiment of YouTube Comments Maximum sentiment score: Minimum sentiment score: Average score of comments: Proportion of Comments with sentiment score is 0: 23.92% (Neutral comments) Distribution of Sentiment score 14

15 Methodology co-comment sentiment analysis 1.6 V1 V1, V2 (V1, V2) User A V2 V1, V2, V3 (V1, V2), (V1, V3) (V2, V3) User B 0.6 V3 Pair (V1, V2) {( ) + ( )} / 2 = 2.75 (V1, V3) ( ) = 2.0 (V2, V3) ( ) = Negative Neutral Positive 15

16 Co-comment Network with One-Count (YouTube Title) Node: 237 Edges: 991 Average Degree: (Max 133: Girls generation) Graph Density:

17 Co-comment Network with All-Count (YouTube Title) Node: 428 Edges: 997 Average Degree: (Max 142: Girls generation) Graph Density:

18 Distribution of Sentiment Value (YouTube Video) Low sentiment value Min: High sentiment value Max:

19 Sentiment Co-comment Network: Positive (YouTube Title) Node: 358 Edges: 807 Average Degree: (Max 232: Girls generation) Graph Density:

20 Sentiment Co-comment Network: Negative (YouTube Title) Node: 966 Edges: 1000 Average Degree: (Max 28:Korean girl pop group 4minute steals Britney s song!) Graph Density:

21 Community Detection SCAN(A Structural Clustering Algorithm for Networks) Algorithm (Xuet al., 2007) Co-comment analysis with one-count Co-comment analysis with all-count Co-comment sentiment analysis No. Community Size No. Community Size No. Community Size PSY K-pop singer Woo Young (2PM) Teayang (Bigbang) Solo/Member IU - - Total 402 Total 376 Total 352 Boyfriend NU'EST 2PM Top Edges Giant component (Cluster#1): SM Ent., YG Ent., JYP Ent. 소속가수 Cluster #2: PSY s Music Video (Gangnam Style, Gentleman, Right Now, Korea, Gangnam Style Live) Sentiment analysis: Strong positive commenting activities of specific singers (Boyfriend, NU EST, 2PM) 21

22 Co-comment network with one-count (YouTube Title) page-rank Rank Title Page-rank View Count 1 Girls Generation ( 소녀시대 )_ Gee ,077,747 2 G-DRAGON CRAYON ( 크레용 ) ,074,994 3 BIGBANG& 2NE1 LOLIPOP ,743,315 4 EXO-K_MAMA ,164,564 5 Super Junior-M ( 슈퍼주니어-M) _ Super Girl( 슈퍼걸 ) ,284,093 6 SHINee( 샤이니 )_ LUCIFER ,128,207 7 G-DRAGON THAT XX ( 그 XX) ,990,548 8 SHINee Sherlock ,151,690 9 에프엑스 _ 피노키오 (Danger) ,460, BIGBANG- TONIGHT ,646, SHINee( 샤이니 )_ RingDingDong( 링딩동 ) ,100, NE1 - GO AWAY ,200, B.A.P. "One Shot" ,118, 에프엑스 _HOT SUMMER ,061, NE1 -CLAP YOUR HANDS ( 박수쳐 ) ,921,447 22

23 Co-comment network with all-count (YouTube Title) page-rank Rank Title Page-rank View Count 1 Girls Generation( 소녀시대 )_ Gee ,077,747 2 Girls' Generation( 소녀시대 )_ 훗 (Hoot) ,010,580 3 G-DRAGON - CRAYON ( 크레용 ) ,074,994 4 BIGBANG - TONIGHT ,646,575 5 SUPER JUNIOR 슈퍼주니어 _SPY ,220,039 6 TVXQ!( 東起 )_ 왜 (Keep Your Head Down) ,557,364 7 EXO-K_MAMA ,164,564 8 Super Junior-M( 슈퍼주니어-M) _ Super Girl( 슈퍼걸 ) ,284,093 9 Super Junior_Sexy, Free & Single ,649, BIGBANG & 2NE1 - LOLLIPOP ,564, NE1 - IT HURTS ( 아파 ) ,341, 에프엑스 _ 피노키오 (Danger) ,460, NE1 - CAN'T NOBODY ,564, SHINee( 샤이니 )_ RingDingDong( 링딩동 ) ,100, SHINee( 샤이니 )_ LUCIFER ,128,207 23

24 Sentiment co-comment network (Positive) (YouTube Title) page-rank Rank Title Page-rank View Count 1 Girls Generation( 소녀시대 )_ Gee ,077,747 2 EXO-K_MAMA ,164,564 3 BIGBANG TONIGHT ,646,575 4 G-DRAGON CRAYON ( 크레용 ) ,074,994 5 Girls Generation( 소녀시대 )_ 훗 (Hoot) ,023,935 6 BIGBANG BEAUTIFUL HANOVER ,832,746 7 SUPER JUNIOR 슈퍼주니어 _ SPY ,220,039 8 BIGBANG& 2NE1 LOLIPOP ,743,315 9 TVXQ!( 동방신기 )_ 왜 (Keep Your Head Down) ,557, NE1 CAN T NOBODY ,564, Super Junior-M( 슈퍼주니어-M) _ Super Girl( 슈퍼걸 ) ,284, INFINITE 1st Album ' 내꺼하자 (Be Mine)' ,345, NE1 -CLAP YOUR HANDS ( 박수쳐 ) ,921, NE1 - GO AWAY ,200, Super Junior_Sexy, Free & Single ,649,761 24

25 Sentiment co-comment network (Negative) (YouTube Title) page-rank Rank Title Page-rank View Count 1 Korean girl pop group 4 minute steal Britney s song! ,634 2 PSY - GANGNAM STYLE ( 강남스타일 ) Made $ 8 Million!?! ,267 3 Kpop on french TV Cavous France ,512 4 [Kpop Controversy] Block B s interview in Thailand ,125 5 Areia Remix#118 Vixx Hyde ,945 6 Woder Girls_Yubin ,486 7 The Top 7 Creepiest K-Pop Videos = ISHlist ,006 8 Too much K-Pop for Reginald ,157 9 신지훈 (Shin Jihun) [Ben]@KPOPSTAR Season ,184, Korean Pop Night Concert SHINee Ending , Bad ( 나쁘다 ) Tablo( 타블로 ) feat. Jinsil , Tiffany Hwang Throws Worst First Pitch Ever ,344, BIGBANG CONCERT - BIG SHOW SPOT , Baby VOX - Coincidence , [HD] SNSD ft SHINee Tell me your wish (genie) ,991 25

26 Correlation Co-comment analysis with one-count Co-comment analysis with all-count Co-comment sentiment analysis V-C N-C P-R V-C N-C P-R V-C N-C P-R View-count 1.64* *.46* 1.72*.51* No. of comment * * * Page-rank * <.05 Spearman correlation Attributes of YouTube video: View-count, Number of comment Network centrality: Page-rank (Edge Weight) Positive correlation between View-count and Sentiment score: r = 0.8* 26

27 Co-comment Network with One Count (Singer) Node: 61 Edges: 235 Average Degree: (Max 64: Girls generation) Graph Density: S.M. YG 27

28 Co-comment Network with All Count (Singer) Node: 99 Edges: 299 Average Degree: 6.04 (Max 64: Girls generation) Graph Density: S.M. YG 28

29 Sentiment Co-comment Network: Positive (Singer) 울림 Ent. S.M. CNC S.M. YG Node: 80 Edges: 161 Average Degree: (Max 64: Girls generation) Graph Density:

30 Sentiment Co-comment Network: Negative (Singer) Node: 241 Edges: 698 Average Degree: (Max 61: SHINee) Graph Density:

31 Page-rank comparison (singer network) Rank Co-comment analysis with one-count Co-comment analysis with all-count Co-comment sentiment analysis (Positive) Co-comment sentiment analysis (Negative) Singer Page-rank Singer Page-rank Singer Page-rank Singer Page-rank 1 Girls generation Girls generation Girls generation SHINee G-dragon NE Bigbang PSY NE Super Junior NE Girls' generation Bigbang Bigbang f(x) Bigbang SHINee Core singer group SHINee G-dragon K-pop Singer f(x) G-dragon SHINee NE Super Junior f(x) Super Junior Super Junior NE1 & Bigbang Super Junior M INFINITE K-pop Super Junior M NE1 & Bigbang EXO EXO B.A.P EXO TVXQ Cover K-pop U-KISS INFINITE Wonder Girls TVXQ EXO-K EXO-K Taeyang minute EXO TVXQ Girls' generation-tts PM Taeyang Taeyang BoA K-pop TV Show BoA Wonder Girls EXO-K Wonder Girls Low page-rank value 31

32 Implication & Future work Applying the co-citation technique to social media data with sentiment analysis Trace the pattern of information sharing behavior of user Identifying the landscape of K-pop stars on YouTube Interpreting the network structure by connecting the social phenomenon Sentiment word analysis of user comment with specific users comments (e.g. spammer) 32

33 References Benevenuto, F., Rodrigues, T., Almeida, V., Almeida, J., & Ross, K. (2009). Video interactions in online video social networks. ACM Transactions on Multimedia Computing, Communications, and Applications Chatzopoulou, G., Sheng, C., & Faloutsos, M. (2010, March). A first step towards understanding popularity in youtube. In INFOCOM IEEE Conference on Computer Communications Workshops, 2010 (pp. 1-6). IEEE. Ocholla, D. N. (2008). A co-link analysis of institutions of higher learning in eastern and southern Africa: Preliminary findings. Siersdorfer, S., Chelaru, S., Nejdl, W., & San Pedro, J. (2010, April). How useful are your comments?: analyzing and predicting youtubecomments and comment ratings. In Proceedings of the 19th international conference on World wide web (pp ). ACM. Thelwall, M. (2006). Interpreting social science link analysis research: A theoretical framework.journal of the American Society for Information Science and Technology,57(1), Thelwall, M., Sud, P., & Vis, F. (2012). Commenting on YouTube videos: From Guatemalan rock to el big bang. Journal of the American Society for Information Science and Technology, 63(3), Xu, X., Yuruk, N., Feng, Z., & Schweiger, T. A. (2007, August). SCAN: a structural clustering algorithm for networks. In Proceedings of the 13th ACM SIGKDD international conference on Knowledge discovery and data mining (pp ). ACM. 33

34 text & social media mining lab THANKS

K-Pop Idol Industry Minhyung Lee

K-Pop Idol Industry Minhyung Lee K-Pop Idol Industry 20100663 Minhyung Lee 1. K-Pop Idol History 2. Idol Industry Factor 3. Regression Analysis 4. Result & Interpretation K-Pop Idol History (1990s) Turning point of Korean Music history