When Frequency Data Meet Dispersion Data in the Extraction of Multi-word Units from a Corpus: A Study of Trigrams in Chinese

摘要

One of the main approaches to extract multi-word units is the frequency threshold approach, but the way this approach considers dispersion data still leaves a lot to be desired. This study adopts Gries's (2008) dispersion measure to extract trigrams from a Chinese corpus, and the results are compared with those of the frequency threshold approach. It is found that the overlap between the two approaches is not very large. This demonstrates the necessity of taking dispersion data more seriously and the dynamic nature of lexical representations. Moreover, the trigrams extracted in the present study can be used in a wide range of language resources in Chinese.