Sketching the Dependency Relations of Words in Chinese

The 2014 Conference on Computational Linguistics and Speech Processing ROCLING 2014, pp. 139-152  The Association for Computational Linguistics and Chinese Language Processing Sketching the Dependency Relations of Words in Chinese Meng-Hsien Shih and Shu-Kai Hsieh Abstract We proposes a language resource by automatically sketching grammatical relations of words based on dependency parses from untagged texts. The advantage of word sketch based on parsed corpora is, compared to Sketch Engine (Kilgarriff, Rychly, Smrz, & Tugwell, 2004), to provide more details about the different usage of each word such as various types of modification, which is also important in language pedagogy. Although some language resources of other languages have attempted to sketch words based on parsed data, in Chinese we have not seen a resource for dependency sketch of words in customized texts. Therefore, we propose such a resource and evaluate with Chinese Sketch Engine (Huang et al., 2005) in terms of corresponding thesaurus function. Keywords: Dependency grammar, Grammatical relation, NLP tools/resources. 1. Introduction Syntagmatic relational information has been the focus of the interface studies of syntax and semantics. With the rapid development of corpora, various corpus query, profiling and visualization tools have emerged quickly over the past years. Among these tools, Word Sketch Engine (Kilgarriff et al., 2004; Huang et al., 2005) has provided an effective approach to quantitatively summarize grammatical and collocation behavior 1. The provided functions include Concordancer, Word Sketch, Sketch Diff, Thesaurus, and other web corpus crawling and processing tools. Previous literatures have revealed that corpus linguistics has benefited greatly from Chinese Sketch Engine (Hong & Huang, 2006). Although proprietary, Word Sketch Engine system is popular among corpus linguists and language teachers because of its functions in language analysis. However, the construction of Sketch Engine is time-consuming due to the  Graduate Institute of Linguistics, National Taiwan University, Taipei, Taiwan E-mail: {simon.xian, shukai}@gmail.com 1http://www.sketchengine.co.uk 139 manually edited sketch grammar. Here we propose an alternative approach to sketch the grammar profile of words automatically from a text corpus. The paper is organized as follows: Section 2 reviews the current design of related language resources. Section 3 describes the proposed method of sketching words in a parsed corpus. Section 4 presents the results from the proposed approach and evaluation. In the final section, we have a brief conclusion for this paper. 2. Review Word Sketch Engine (WSE) provides a set of corpus query tools that aims to help users reveal linguistic patterns in language use. Among these tools, word sketch function gains the most popularity and has widely applied in the studies of corpus linguistics and language pedagogy (Kilgarriff, 2007). Given the preprocessed corpus data, the available WSE system in most languages makes use of regular expressions to extract grammatical information from a POS-tagged corpus. The so-called sketch grammars, mostly manually crafted by linguists, describe the relation between a target word and its dependent, constrained on the surrounding context. In its design of grammar engineering, the sketch grammars are used for finite-state shallow parsing to extract the different grammatical …

Content truncated to 500 words