Verbs observed: A corpus-driven pedagogic grammar

This paper describes the outcome of a project to code the complementation patterns of all the verbs in Collins COBUILD English Dictionary (1995). COBUILD stands for 'Collins and Birmingham University international Language Database'. The coding is based on the Bank of English corpus at COBUILD and uses a simple notation based on words and word-classes rather than traditional functional categories. The result of this exercise is a list of verb patterns, with a complete list of all the verbs in a corpus of 250 million words that have each pattern. It is found that the verbs that share a pattern fall into groups based on meaning. This grammar is the first pedagogic grammar to integrate syntax and lexis using corpus data. The grammar is used to explore traditional grammatical categories such as Object, Complement, etc. These are found to be inadequate to account for the actual behaviour of verbs. Finally, the paper explores the possibility of using a pattern grammar to analyse naturally-occurring discourse.