Constructing Chinese Abbreviation Dictionary: A Stacked Approach

Proceedings of COLING 2012: Technical Papers, pages 3055–3070, COLING 2012, Mumbai, December 2012. Constructing Chinese Abbreviation Dictionary: A Stacked Approach Longkai Zhang Sujian Li Houf eng Wang∗ Ni Sun X inf an Meng Key Laboratory of Computational Linguistics (Peking University) Ministry of Education, China zhlongk@qq.com, lisujian@pku.edu.cn, wanghf@pku.edu.cn, sunny.forwork@gmail.com, mxf@pku.edu.cn ABSTRACT Abbreviation is a common linguistic phenomenon with wide popularity and high rate of growth. Correctly linking full forms to their abbreviations will be helpful in many applications. For example, it can improve the recall of information retrieval systems. An intuition to solve this is to build an abbreviation dictionary in advance. This paper investigates an automatic abbreviation generation method, which uses a stacked approach for Chinese abbreviation generation. We tackle this problem in two stages. First we use a sequence labeling method to generate a list of candidate abbreviations. Then, we try to use search engine to incorporate web data to re-rank the candidates, and finally get the best candidate. We use a Chinese abbreviation corpus which contains 8015 abbreviation pairs to evaluate the performance. Experiments revealed that our method gave better performance than the baseline methods. KEYWORDS: Chinese Abbreviation Generation, Abbreviation Mining. ∗Corresponding author 3055 1 INTRODUCTION Abbreviation is defined as a short description of the original long phrase. For example, "ACL" is the abbreviation for the full form "Association for Computational Linguistics". While abbreviation is a common linguistic phenomenon, it causes many problems like spelling variation (Nenadi´c et al., 2002).The different writing manners make it difficult to identify the terms conveying the same concept, which will hurt the performance of many applications, such as information retrieval (IR) systems. In IR applications, one simple solution is to expand the original query by adding corresponding abbreviations to a search engine. For example, when using a search engine with an original query of "United States of America", a user will get more relevant results by expanding the query to include the abbreviation "USA." To achieve this we need to have an abbreviation dictionary, which is laborious to manually maintain because the number of abbreviations increases rapidly (Chang and Schutze, 2006). Therefore, it is helpful to automatically generate abbreviation from full forms. This leads to the idea of "abbreviation generation", i.e., finding the correct abbreviation for a full form. The generation of abbreviations in Chinese differs from that for English. The reason is that Chinese itself lacks many commonly considered features in English abbreviation generation methods (Pakhomov, 2002; Yu et al., 2006; HaCohen-Kerner et al., 2008; Ao and Takagi, 2005). Detailed differences between English abbreviation generation and Chinese abbreviation features are listed in TABLE 1. Due to these differences, specific attention should be paid to Chinese abbreviation generation. Feature English Chinese Word boundary YES NO Case sensitivity YES NO Table 1: Comparison between Chinese and English abbreviation generation with regards to features. Most of Chinese abbreviations are generated by selecting representative characters from the full forms1. For example, the abbreviation of "北京大学" (Peking University) is "北大" which is generated by selecting the …

Content truncated to 500 words