Comparative linguistic analysis with Firthian collocations: Cases of synonym differentiation and proficiency assessment

This paper proposes a novel approach to collocation-based comparative linguistic analysis. The approach defines Firthian collocation as a range of collocates parameterized with positional information and association strength, and facilitates Firthian-collocation-based comparative analyses with a suite of quantitative tools including visualization, semantic abstraction, and similarity computation. The application of this approach is showcased in the synonym differentiation of four adverbs—actually, genuinely, really, and truly—and the collocation accuracy measurement of L2 learners that uses Chinese Learners English Corpus as the learners’ corpus and British National Corpus as the reference corpus. These case studies demonstrate that the approach not only enables intuitive and holistic inference of syntactic and semantic functions of node words, but also supports large-scale quantitative investigation of lexical knowledge.