Using a comparable corpus to investigate lexical patterning in English abstracts written by non-native speakers

This study is set in a broad context of development of courses and computational tools to aid Brazilian graduate students in writing scientific papers in English. The main focus is on experimental research papers from the disciplines of physics, pharmaceutical and computer sciences. One of our primary objectives is to give students feedback and raise their awareness of the most typical lexical patterns used by their academic discourse community while at the same time draw their attention to the various available alternatives. Errors related to lexical use are by far the most frequent errors made by Brazilian graduate students when writing academic English. The aim of this paper is to carry-out a corpus-based study to investigate potential differences in the collocational patterns of work in abstracts written by Brazilian graduate students as opposed to abstracts collected from published papers of the same discipline. Relevant differences were found between the two subcorpora. The results were validated by examining the identified lexical patterns in a reference corpus of English abstracts. We also identified various items other than work which may be used to refer to the study described in the abstract as well as other lexical variations within the lexical patterns analysed.