The subcategorization of English adverbs: A feature-based clustering approach
Christina Sanchez‐Stockhammer, Antony R. Unwin · 2022
The category of the adverb in the English language is notoriously heterogeneous, and Crystal (1995: 211) even considers it a “dustbin category” which combines such disparate members as the manner adverb happily, the intensifier very, the comparative more and the postmodifier indeed. In order to determine the most appropriate subclassification of the category of the adverb in English, the present contribution presents original research based on a dataset of the 2500 most frequent words in the British National Corpus (Sanchez 2008). The 206 adverbs in this high-frequency sample were coded with regard to their decomposability into semantic components, word-family integration, language of origin, age, underlying word formation process , suffix type used and semantic class. Using a tree plot, we first investigated whether the adverbs in our dataset are more similar to the lexical or the grammatical parts of speech, but with no conclusive evidence. We then used a recent clustering approach (consensus clustering; cf. Chiu 2018) and an innovative visualization in the form of an adaptation of parallel coordinate plots for multivariate categorical data to determine whether cluster analyses can be used to automatically subcategorize words assigned to the traditional category of the adverb into meaningful subcategories. We did indeed find a linguistically meaningful categorization into three clusters that are distinguished with regard to the word formation type characteristic of the group members, namely simplex adverbs (e.g. next), -ly suffixations (e.g. regularly) and other complex word formations (e.g. meanwhile). Our results indicate that the presence or absence of an adjectival base is a much better criterion for the subcategorization of English adverbs than the property of permitting inflection.