Named Entity Recognition in Chinese News Comments on the Web

Proceedings of the 5th International Joint Conference on Natural Language Processing, pages 856–864, Chiang Mai, Thailand, November 8 – 13, 2011. c⃝2011 AFNLP Named Entity Recognition in Chinese News Comments on the Web Xiaojun Wana, Liang Zongb, Xiaojiang Huangc, Tengfei Ma, Houping Jia, Yuqian Wu and Jianguo Xiao Institute of Compute Science and Technology, The MOE Key Laboratory of Computational Linguistics, Peking University, Beijing 100871, China {awanxiaojun, chuangxiaojiang}@icst.pku.edu.cn bzongliang.cn@gmail.com Abstract News comment is a new text genre in the Web 2.0 era. Many people often write com- ments to express their opinions about recent news events or topics after they read news articles. Because news comments are freely written without checking, they are very dif- ferent from formal news texts. In particular, named entities in news comments are usu- ally composed of some wrongly written words, informal abbreviations or aliases, which brings great difficulties for machine detection and understanding. This paper ad- dresses the task of named entity recognition in Chinese news comments on the Web. We propose to leverage the entity information in the referred news article to improve named entity recognition in the news comments. Three different schemes are investigated to find useful entities in the news article for new feature generation in the CRFs model. Finally, a dictionary-based correction step is employed to further improve the results. We manually labelled a benchmark dataset with 60 pieces of news and 6000 comments downloaded from a popular Chinese news portal – www.sina.com.cn. The experimen- tal results on the dataset show that our method is effective for this special task. 1 Introduction Named entity recognition (NER) is one of the fundamental tasks in the field of natural language processing. It has been widely used in the areas of information retrieval, machine translation, question answering, and so on. In most literatures, named entities are defined as entity names (per- son names, location names and organization names). With the advent of MUC, CONLL, ACE and SIGHAN evaluations, NER has received much attention of the researchers and hence achieved great development. News comment is a new text genre in the Web 2.0 era, and many people often write and post comments to express their opinions on recent news events or topics after they read news arti- cles. As the roles which people play on the inter- net have gradually changed from acquirers to suppliers, news comments have become one of the most valuable information resources. For ex- ample, on one of the popular Chinese news por- tals – sina.com.cn, every piece of hot news is associated with over 500 comments. Named en- tity recognition is the basis of many other news comments understanding and mining applications, including entity relation extraction, opinion holder and target extraction in news comments, and so on. Because news comments are freely written by different persons with different education back- grounds and writing styles, they are very differ- ent from formal news texts. In particular, Chinese news comments have the following properties: 1) The texts in news comments are …

Content truncated to 500 words