Collocation dimensions in academic English

Abstract

This study set out to identify and characterize extended collocational networks, or “collocational dimensions”, as they are represented in written academic English, through a bottom-up approach, following Berber Sardinha (2017). It has been established in the literature that collocational networks exist for individual node words hence collocational networks could exist for collocates across nodes, which would be evidenced in a dimension of collocation. Thus, this study set out to extend the concept of collocational networks by identifying the major sets of collocational patterns across nodes, or “collocational dimensions”, in academic writing. The Corpus of Academic Texts (CAT) was designed and compiled for this study. It comprises textbooks and journal articles, divided into five subcorpora, one per discipline, according to research discipline areas of the National Science Foundation (NSF): Biological Sciences; Computer and Information Science and Engineering; Engineering; Mathematical and Physical Sciences; and Social, Behavioral and Economic Sciences. To accomplish this goal, multivariate statistical procedures, such as exploratory factor analysis and cluster analysis were employed. The exploratory factor analysis led to the identification of 11 collocational dimensions, i.e., dimensions of extended collocational patterns, comprising a total of 10,474 unique collocations, namely: (1) Finance and Economics; (2) Classification; (3) Social and Cultural Development; (4) Exerting Influence and Shaping Outcomes; (5) Empirical Research Methods; (6) Local and Global Society; (7) Reporting Research; (8) Individuals in Society; (9) Gaining Knowledge and Developing Concepts; (10) Social Welfare; (11) Marking Time and Sectioning. The cluster analysis revealed five clusters, which reflected semantic groupings of nodes