Omni-word Feature and Soft Constraint for Chinese Relation Extraction

Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics, pages 572–581, Baltimore, Maryland, USA, June 23-25 2014. c⃝2014 Association for Computational Linguistics Omni-word Feature and Soft Constraint for Chinese Relation Extraction Yanping Chen† Qinghua Zheng† †MOEKLINNS Lab, Department of Computer Science and Technology Xi’an Jiaotong University, China ypench@gmail.com, qhzheng@mail.xjtu.edu.cn ‡Amazon.com, Inc. wzhan@amazon.com Wei Zhang‡ Abstract Chinese is an ancient hieroglyphic. It is inat- tentive to structure. Therefore, segmenting and parsing Chinese are more difficult and less accurate. In this paper, we propose an Omni- word feature and a soft constraint method for Chinese relation extraction. The Omni-word feature uses every potential word in a sentence as lexicon feature, reducing errors caused by word segmentation. In order to utilize the structure information of a relation instance, we discuss how soft constraint can be used to cap- ture the local dependency. Both Omni-word feature and soft constraint make a better use of sentence information and minimize the in- fluences caused by Chinese word segmenta- tion and parsing. We test these methods on the ACE 2005 RDC Chinese corpus. The re- sults show a significant improvement in Chi- nese relation extraction, outperforming other methods in F-score by 10% in 6 relation types and 15% in 18 relation subtypes. 1 Introduction Information Extraction (IE) aims at extracting syntactic or semantic units with concrete concepts or linguistic functions (Grishman, 2012; McCal- lum, 2005). Instead of dealing with the whole doc- uments, focusing on designated information, most of the IE systems extract named entities, relations, quantifiers or events from sentences. The relation recognition task is to find the rela- tionships between two entities. Successful recog- nition of relation implies correctly detecting both the relation arguments and relation type. Although this task has received extensive research. The per- formance of relation extraction is still unsatisfac- tory with a F-score of 67.5% for English (23 sub- types) (Zhou et al., 2010). Chinese relation extrac- tion also faces a weak performance having F-score about 66.6% in 18 subtypes (Dandan et al., 2012). The difficulty of Chinese IE is that Chinese words are written next to each other without de- limiter in between. Lacking of orthographic word makes Chinese word segmentation difficult. In Chinese, a single sentence often has several seg- mentation paths leading to the segmentation ambi- guity problem (Liang, 1984). The lack of delimiter also causes the Out-of-Vocabulary problem (OOV, also known as new word detection) (Huang and Zhao, 2007). These problems are worsened by the fact that Chinese has a large number of characters and words. Currently, the state-of-the-art Chinese OOV recognition system has performance about 75% in recall (Zhong et al., 2012). The errors caused by segmentation and OOV will accumulate and propagate to subsequent processing (e.g. part- of-speech (POS) tagging or parsing). Therefore, the Chinese relation extraction is more difficult. According to our survey, com- pared to the same work in English, the Chinese re- lation extraction researches make less significant progress. Based on the characteristics of Chinese, in this paper, an …

Content truncated to 500 words