Abstract

Discontinuous constructions in spoken Chinese varieties: Extraction and contrastive covarying collexeme analysis

---

Abstract

Linguistic theory has long been interested in the co-occurrence of forms in discourse. While most computational corpus-based studies of collocations and multi-word expressions (MWEs) examine contiguous strings of co-occurring words, e.g. kick the bucket or take a look, some have examined co-occurrence between non-contiguous words, e.g. turn the heater off, too heavy to use (e.g. Watanabe 2021, Dunn 2017). This project extends such methodology to Chinese discontinuous constructions, focusing on constructions where a preverbal or earlier particle (which I call nonfinal particles (NFPs)) tends to co-occur with a final particle (FP). I examine particle frames from the POS-tagged HKCanCor (mostly copresent conversations), plus the Mandarin CallHome corpus of telephone calls (Liu et al. 2004), which was POS-tagged and parsed using spaCy (Honnibal & Montani 2017). I first extract candidate combinations with a hybrid of existing methods, pruning results by statistically selecting only pairs with strong evidence for attraction/repellence while controlling the false discovery rate. I then perform a covarying collexeme analysis (Stefanowitsch & Gries 2005): I examine how NFPs and FPs co-occur within utterances (treated as constructions), through measures of attraction, repellence and productivity. I then assess how well these statistics support the common claim in the literature that particle frames are much more common in Cantonese than Mandarin (Tang 2015).

Keywords: Cantonese, Mandarin, discontinuous construction, comparative syntax, collostructional analysis, collocations, corpus linguistics, syntax, discontinuous collocations, information theory

How to Cite:

Lai, R., (2024) “Discontinuous constructions in spoken Chinese varieties: Extraction and contrastive covarying collexeme analysis”, Society for Computation in Linguistics 7(1), 324–328. doi: https://doi.org/10.7275/scil.2216