dvi

Size: px

Start display at page:

Download "dvi"

しょうじとべ
5 years ago
Views:

1 Kyu S. Hahn Seulgi Jang Press Coverage of Historical and Territorial Issues between Japan and Korea: Frame Analysis Using a Topic Model Yuki Ogawa,TetsuroKobayashi,KyuS. Hahn and Seulgi Jang In order to explore differences in news frames regarding controversial historical and territorial issues between Japan and Korea, the contents of four Japanese and Korean newspapers were analyzed by using topic models. Specifically, all the articles on the first page of Yomiuri, Asahi, Donga, and Hankyoreh newspapers during the two years from 2012 to 2013 were analyzed using Latent Dirichlet Allocation (LDA). Results indicated that in Japanese newspapers, the territorial issue related to Takeshima (Dokto) was framed as a part of a broader international security topic, whereas in Korean newspapers it was more closely associated with historical issues that are unique to the relationship between Japan and Korea. The issue of comfort women was framed in Japanese newspapers as an aspect of domestic politics surrounding the statements made by Toru Hashimoto, the former mayor of Osaka city, whereas in Korean newspapers, it was more distinctively framed as a historical issue. The potential consequences of different news framing between Japan and Korea, as well as the usefulness of using LDA for content analysis are discussed. Key words: Japan-Korea comparison, Takeshima (Dokto), comfort women, content analysis, topic model, LDA (Latent Dirichlet Allocation) LDA Latent Dirichlet Allocation 1. Ritsumeikan University y-ogawa@fc.ritsumei.ac.jp City University of Hong Kong 5/F Run Run Shaw Creative Media Centre, 18 Tat Hong Avenue, Kowloon Tong, Hong Kong tkobayas@cityu.edu.hk Seoul National University Gwanak-ro 1, Gwanak-gu, Seoul 2015

2 % ) NPO % 2) 3) ) (Weaver, Graber, McCombs, & Eyal, 1981) 1) jp/index-gai.html ) NPO html ) (2014) 4) ) 2. (e.g., 2007;, 2008;, 2008; Lee & Min, 2011) (2000) 54.9% 85.1% Oh (2011) Pak (2016) Pak (2016) Semetko & Valkenburg (2000) ) 2015 NHK (NHK, 2016) % % (, 2015)

3 3 4 3 Pak (2016) Pak (2016) Pak (2016) 4 Pak (2016) Semetko & Valkenburg (2000) Pak (2016) Pak (2016) LDA Latent Dirichlet Allocation CTM Correlated Topic Models LDA Latent Dirichlet Allocation (Titov & McDonald, 2008) Twitter (Xu, Ru, Xiang, & Yang, 2011; Pennacchiotti & Pepescu, 2011) (Cao & Fei-Fei, 2007) (Iwata, Watanabe, Yamada, & Ueda, 2009) LDA Latent Dirichlet Allocation (Blei, Ng, & Jordan, 2003) LDA 1 LDA LDA θ d d z d,n β k LDA θ Dir(α)

4 LDA (, 2015) 1. LDA K D N z w θ d d β k k α θ β 0 β p(θ, z, w α, β) = p(θ α)π N n=1p(z nθ)p(w n z n,β) (1) LDA 1 θ β (Griffiths & Steyvers, 2004) LDA 2 2. LDA 2 40% 60% PC play play

5 5 LDA LDA Guo et al. (2016) LDA LDA CTM (Correlated Topic Models) CTM (Blei & Lafferty, 2007) LDA IT CTM LDA p(η, z w, β 1:K,μ,Σ) p(η μ, Σ)Π N = n=1p(z n η)p(w n z n,β 1:K) p(η μ, Σ)Π N n=1 Σ K z n=1p(z n η)p(w n z n,β 1:K)dη (2) CTM Science Blei & Lafferty, CTM CTM Σ CTM 2 η η θ CTM 4 Science (Blei & Lafferty, 2007) CTM PAM Pachinko Allocation Model (Li & McCallum, 2006) DTM Dynamic Topic Models (Blei & Lafferty, 2006)

6 CTM (Akuto, 1996) 2 2 6) 7) NAVER 8) 6) Naver Naver 3 7) 8) NAVER [ ] MeCab 9) MeCab HanDic 10) Wikipedia ) 1,453, ,369.+?.+?.+?.+?.+?.+? 9) MeCab: 10) HanDic: 11) Wikipedia Wikipedia

7 2. 3,865 542,368 9,164 2,828 500,683 8,697 2,247 410,888 7,504 1,948 312,426 5,864 2 2 3.2.2. CTM CTM LDA CTM CTM K K =100 (Blei & Lafferty, 2007) K =100 CTM β CTM R topicmodels 12) 12) CRAN topicmodels https://cran.

7 7 2. 3, ,368 9,164 2, ,683 8,697 2, ,888 7,504 1, ,426 5, CTM CTM LDA CTM CTM K K =100 (Blei & Lafferty, 2007) K =100 CTM β CTM R topicmodels 12) 12) CRAN topicmodels

8 T LDA 7 7 T19 T69 T19 T ) T19 3 T ) CRAN wordcloud

9 9 8. T19 T T16 T T16

10 T T T T67 20 T67 T16 T67 T16 T T16 T T58 T55 T17 T4 8 T69 T5 T20 T19

11 T19 T69 T80 T16 12 T69 31 T67 36 T67 36

12 44 1 86 11. 13. T69 12. T19 4.3.2. 11 CTM 11 T19 T69 12 13 T19 T69

12 T T CTM 11 T19 T T19 T69 12 T19 T39 T40 T T39 T40 T69 13 T91 T80 T

13 13 14 T16 T67 15 T16 T67 15 T16 T25 T25 15 T42 T16 T17 T T16 T

14 (Iyengar, 1991) Pak (2016) (2014) 100 (Li & McCallum, 2006)

15 15 DTM Akuto, H. (1996). Media in electoral campaigning in Japan and the United States. in S.J. Pharr and E.S. Krauss (Eds.). Media and Politics in Japan (pp ), Honolulu: University of Hawai i Press. Blei, D.M., Ng, A.Y. and Jordan, M.I. (2003). Latent Dirichlet Allocation. Journal of Machine Learning Research, 3, Blei, D.M. and Lafferty, J.D. (2006). Dynamic Topic Models. Proc. ICML 06, Blei, D.M. and Lafferty, J.D. (2007). A Correlated Topic Model of Science. The Annals of Applied Statistics, 1(1), Cao, L. and Fei, L. (2007). Spatially Coherent Latent Topic Model for Concurrent Segmentation and Classification of Objects and Scenes. Proc. ICCV 07. Griffiths, T.L. and Steyvers, M. (2004). Finding scientific topics. Proc. National Academy of Sciences, 101(1), Guo, L, Vargo, C.J., Pan, Z., Ding, W. and Ishwar, P. (2016). Big Social Data Analytics in Journalism and Mass Communication: Comparing Dictionary-Based Text Analysis and Unsupervised Topic Modeling. Journalism & Mass Communication Quarterly, online first. (2015).., 115(388), Iwata, T., Watanabe, S., Yamada, T. and Ueda, N. (2009). Topic Tracking Model for Analyzing Consumer Purchase Behavior. Proc. IJCAI 09, Iyengar, S. (1991). Is Anyone Responsible? How Television Frames Political Issues Chicago. University of Chicago Press. (2015) link=pds/publication&menu id=193&menu orderby=2&tabyn=n ( ) (2008)..,, (2014).... (2007). TV., 57, (2008) ,., Lee, Y.M. and Min, Y. (2011). Social Movements and Media Frame: Collective Action Frames of the Japanese comfort women issue. Media, Gender & Culture, 19, (In Korean). Li, W. and McCallum, A. (2006). Pachinko Allocation: DAG-Structured Mixture Models of Topic Correlations. Proc. ICML 06, NHK (2016) yoron/ pdf/ pdf Oh, D.-Y. (2011). Frame Difference Analysis on the News of Korean and Japanese newspapers regarding the 100th Anniversary of Japan s Colonization of Korea. Korean Journal of Journalism and Communication Studies, 55(1), (In Korean). (2010) 2., 26(2), Pak, H.J. (2016). News Reporting on Comfort Women: Framing, Frame Difference, and Frame Changing in Four South Korean and Japanese Newspapers, Journalism & Mass Communication Quarterly, online first. Pennacchiotti, M. and Popescu, A.M. (2011). Machine Learning Approach to Twitter User Classification. Proc. ICWSM 11, Semetko, H.A. and Valkenburg, P.M. (2000). Framing European Politics: A content Analysis of Press and Television News. Journal of communication, 50(2), Titov, I. and McDonald, R. (2008). A joint model of text and aspect ratings for sentiment summarization, Proc. ACL 08, Xu, Z., Ru, L., Xiang, L. and Yang, Q. (2011). Discovering User interest on Twitter with a Modified Author-Topic model. Proc. IEEE/WIC/ACM Int. Conf. WIIAT, 1, Weaver, D.H., Graber, D.A., McComs, M.E. and Eyal, C.H. (1981). Media Agenda-Setting in a Presidential Election. New York: Praeger. (2000).., 50, Zhao,W.X.,Jiang,J.,Weng,J.,He,J.,Lim,E.P., Yan, H. and Li, X. (2011). Comparing Twitter and Traditional Media Using Topic Models. In Advances in Information Retrieval (pp ). Berlin, Germany: Springer

Library and Information Science No

Library and Information Science No. 65 2011, p. 83 107. Episodic Frame and Thematic Frame in Television News and Newspapers: An Analysis of the Coverage of General Elections in Japan Yuko MATSUBA Shuichi