Cross-Lingual Topic Alignment in Time Series Japanese / Chinese News

Cross-Lingual Topic Alignment in Time Series Japanese / Chinese News Shuo Hu Yusuke Takahashi Liyi Zheng Takehito Utsuro Graduate School of Systems and Information Engineering, University of Tsukuba, Tsukuba, 305-8573, JAPAN Masaharu Yoshioka Graduate School of Information Science and Technology, Hokkaido University, Sapporo, 060-0808, Japan Noriko Kando National Institute of Informatics, Tokyo 101-8430, Japan Tomohiro Fukuhara National Institute of Advanced Industrial Science and Technology Tsukuba, 305-8568 Japan Hiroshi Nakagawa Information Technology Center, University of Tokyo, Tokyo 113-0033, Japan Yoji Kiyota NEXT Co., Ltd., Tokyo, 108-0075, Japan Abstract Among various types of recent information explosion, that in news stream is also a kind of serious problems. This paper studies is- sues regarding topic modeling of information flow in multilingual news streams. If some- one wants to find differences in the topics of Japanese news and Chinese news, it is usu- ally necessary for him/her to carefully watch every article in Japanese and Chinese news streams at every moment. In such a situation, topic models such as LDA (Latent Dirichlet Allocation) and DTM (dynamic topic model) are quite effective in estimating distribution of topics over a document collection such as articles in a news stream. Especially, as a topic model, this paper employs DTM, but not LDA, since it can consider correspondence be- tween topics of consecutive dates. Based on the results of estimating distribution of topics in Japanese / Chinese news streams, this paper proposes how to analyze cross-lingual align- ment of topics in time series Japanese / Chi- nese news streams. 1 Introduction Among various types of recent information explo- sion, that in news stream is also a kind of serious problems. This paper studies issues regarding topic modeling of information flow in multilingual news streams. If someone wants to find differences in the topics of Japanese news and Chinese news, it is usu- ally necessary for him/her to carefully watch every article in Japanese and Chinese news streams at ev- ery moment. In such a situation, topic models such as LDA (Latent Dirichlet Allocation) (Blei et al., 2003) and DTM (dynamic topic model) (Blei and Lafferty, 2006) are quite effective in estimating distribution of topics over a document collection such as articles in a news stream. Especially, as a topic model, this paper employs DTM, but not LDA, since it can con- sider correspondence between topics of consecutive dates. In DTM, we suppose that the data is divided by time slice, for example by date. DTM models the documents (such as articles of news stream) of each slice with a K-component topic model, where the k-th topic at slice t smoothly evolves from the k-th topic at slice t − 1. Based on the results of estimating distribution of topics in Japanese / Chinese news streams, this pa- per proposes how to analyze cross-lingual alignment of topics in time series Japanese / Chinese news streams. The overall flow of the proposed frame- work is illustrated in Figure 1. In order to bridge the gaps between the two languages, namely, …

Content truncated to 500 words