Using Chinese Gigaword Corpus and Chinese Word Sketch in linguistic Research

Using Chinese Gigaword Corpus and Chinese Word Sketch in linguistic research Jia-Fei Hong, Chu-Ren Huang National Taiwan University, Graduate Institute of Linguistics, No. 1, Sec. 4, Roosevelt Road, Taipei, 106 Taiwan Institute of Linguistics 128, Section 2, Academia Road 115, Taipei, Taiwan, R.O.C {jiafei, churen}@gate.sinica.edu.tw Abstract. We explore the possibility of deeper linguistic research based on corpus and computational linguistic tools in this paper. In particular, we adopt Chinese Word Sketch, the application of Word Sketch Engine to Chinese GigaWord Corpus, for linguistic research. We apply Chinese Sketch Engine results to deeper linguistic account such as selectional restriction and event type selection. The study is based on the comparison of two basic verbs of ingestion: chi1 ‘to eat’ and he1 ‘to drink’. Key words: Chinese Word Sketch, selectional restriction, event type, Corpus-based linguistic research 1 Introduction In this paper, we explore the potential of Chinese Word Sketch (CWS) as a tool for deeper linguistic research. The CWS is a combination of the Chinese GigaWord Corpus (Huang et al. 2005) with the linguistic search tool of Word Sketch Engine (Kilgarriff et al. 2004). It is well documented that the Sketch Engine is a very powerful tool for extracting meaning grammatical relations given a sufficiently large corpus (Kilgarriff et al. 2004, Huang et al. 2005). We show in this paper how CWS can be applied to confirm and refine human lexical semantic work, as well as be applied to deeper linguistic account. The set of linguistic fact that we will account for involve the two basic verbs of ingestion chi1 ‘to eat’ and he1 ‘to drink’. We draw directly from the corpus-based manual analyses from Chinese Wordnet Group (CWN Group) to define the sense divisions for “chi1” and “he1”. Chinese Word Sketch Engine 183 helps us to verify and refine the manual analysis, especially with the event type and argument roles for “chi1” and “he1”. In addition, we are concerned with the semantic attributes shared or distributed among these argument roles. We further explore the linguistic accounts of selectional restriction and event type selection based on these distributions. The remaining part of this paper starts with an introduction to the GigaWord Corpus and Word Sketch Engine. We next explicate the scope and and goals of the current study. Thirdly, we detail the collocation data, as well as how they support and refine manual sense analysis. Fourthly, we show how the derived generalizations can be applied in deep linguistic account. Finally, we 2 conclude with predictions and future work. GigaWord Corpus The Chinese Gigaword Corpus contains about 1.1 billion Chinese characters, including more than 700 million characters from Taiwan’s Central News Agency, and nearly 400 million characters from China’s Xinhua News Agency. Before loading Chinese Gigaword into Sketch Engine, all the simplified characters were converted into traditional characters, and the texts were segmented and POS tagged using the Academia Sinica segmentation and tagging system (Huang et al., 1997). The segmentation and tagging was performed automatically with automatic and partially manual post-checking. The precision …

Content truncated to 500 words