Cross-Lingual News Group Recommendation Using Cluster-Based Cross-Training

Computational Linguistics and Chinese Language Processing Vol. 13, No. 1, March 2008, pp. 41-60 41 © The Association for Computational Linguistics and Chinese Language Processing [Received June 30, 2007; Revised October 8, 2007; Accepted January 12, 2008] Cross-Lingual News Group Recommendation Using Cluster-Based Cross-Training Cheng-Zen Yang*, Ing-Xiang Chen*, and Ping-Jung Wu* Abstract Many Web news portals have provided clustered news categories for readers to browse many related news articles. However, to the best of our knowledge, they only provide monolingual services. For readers who want to find related news articles in different languages, the search process is very cumbersome. In this paper, we propose a cross-lingual news group recommendation framework using the cross-training technique to help readers find related cross-lingual news groups. The framework is studied with different implementations of SVM and Maximum Entropy models. We have conducted several experiments with news articles from Google News as the experimental data sets. From the experimental results, we find that the proposed cross-training framework can achieve accuracy improvement in most cases. Keywords: Cross-Lingual News Group Mapping, Cross-Training, Semantic Overlapping, Mapping Recommendation 1. Introduction As the Web becomes an abundant source of news information, it also becomes an important medium for people to learn recent tidings. To provide readers a convenient way of viewing a news event described by different news agencies, many Web news portals, such as AltaVista News and Google News, cluster news articles according to their relevance with consistent user interfaces. With such news clustering services, readers could easily acquire more details of an interesting news event from numerous reports. Ideally, they can simply click through an entry link to browse many related news reports without need of a cumbersome searching procedure. Nevertheless, if the news event is originally reported by foreign news agencies, the readers usually find that there are only few translated news articles and can only acquire an overview * Dept. of Computer Sci. and Eng., Yuan Ze University, 135 Yuan-Tung Rd., Chungli, 320, Taiwan. Tel.: +886-3-4638800 ext: 2361 Fax: +886-3-4638850. E-mail: {czyang,sean,pjwu}@syslab.cse.yzu.edu.tw 42 Cheng-Zen Yang et al. of the news event. If they want to find more related foreign news stories, they may generally get frustrated due to the following two reasons. First, the translated news articles seldom provide as much information as the original news articles. Second, the translation may add more interpretations that can mislead in the searching direction. The following example illustrates these situations. This news story, reported in BBC News [2006], is a good example to show these problems. The title of its English version is “First impressions count for web” and the article contains 15 paragraphs mainly focused on the impressions in a 20th of a second after first sight [BBC News 2006]. However, the title of its Chinese news story is “好網頁還需要讓讀 者一見鍾情” and may be translated into “Good web pages need to let readers fall in love at first sight”, which includes additional semantic information related to love. In addition, the Chinese news article has only 7 paragraphs. When …

Content truncated to 500 words