Coarse-grained Candidate Generation and Fine-grained Re-ranking for Chinese Abbreviation Prediction

Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1881–1890, October 25-29, 2014, Doha, Qatar. c⃝2014 Association for Computational Linguistics Coarse-grained Candidate Generation and Fine-grained Re-ranking for Chinese Abbreviation Prediction Longkai Zhang Houfeng Wang Xu Sun Key Laboratory of Computational Linguistics (Peking University) Ministry of Education, China zhlongk@qq.com, wanghf@pku.edu.cn, xusun@pku.edu.cn Abstract Correctly predicting abbreviations given the full forms is important in many natu- ral language processing systems. In this paper we propose a two-stage method to find the corresponding abbreviation given its full form. We first use the contextual information given a large corpus to get ab- breviation candidates for each full form and get a coarse-grained ranking through graph random walk. This coarse-grained rank list fixes the search space inside the top-ranked candidates. Then we use a sim- ilarity sensitive re-ranking strategy which can utilize the features of the candidates to give a fine-grained re-ranking and se- lect the final result. Our method achieves good results and outperforms the state-of- the-art systems. One advantage of our method is that it only needs weak super- vision and can get competitive results with fewer training data. The candidate genera- tion and coarse-grained ranking is totally unsupervised. The re-ranking phase can use a very small amount of training data to get a reasonably good result. 1 Introduction Abbreviation Prediction is defined as finding the meaningful short subsequence of characters given the original fully expanded form. As an example, “HMM” is the abbreviation for the correspond- ing full form “Hidden Markov Model”. While the existence of abbreviations is a common lin- guistic phenomenon, it causes many problems like spelling variation (Nenadi´c et al., 2002). The dif- ferent 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 and machine translation (MT) systems. Previous works mainly treat the Chinese ab- breviation generation task as a sequence labeling problem, which gives each character a label to in- dicate whether the given character in the full form should be kept in the abbreviation or not. These methods show acceptable results. However they rely heavily on the character-based features, which means it needs lots of training data to learn the weights of these context features. The perfor- mance is good on some test sets that are similar to the training data, however, when it moves to an un- seen context, this method may fail. This is always true in real application contexts like the social me- dia where there are tremendous new abbreviations burst out every day. A more intuitive way is to find the full- abbreviation pairs directly from a large text cor- pus. A good source of texts is the news texts. In a news text, the full forms are often mentioned first. Then in the rest of the news its corresponding abbreviation is mentioned as an alternative. The co-occurrence of the full form and the abbrevia- tion makes it easier for us …

Content truncated to 500 words