Apples and oranges: On comparing data from different corpora
Hans Lindquist, Magnus Levin · Corpus Linguistics and Linguistic Theory · 2000
This paper discusses problems arising when several different corpora are used in the same investigation to study e.g. differences between geographical varieties like British and American English, differences between speakers with varying educational or social background, differences between spoken and written language, text type variation, and developments over time. Examples are given from the study of agreement with collective nouns, prepositional usage, comparison of disyllabic adjectives, and variation between pronouns in -body and -one. Corpora used include LOBIFLOB, Brown/Frown, the London-Lund Corpus, the Longman Spoken American Corpus, CobildDirect, the BNC and newspaper CD-ROMs. It is demonstrated that with careful evaluation of the results, data from different corpora can be successfully combined to give a clearer picture of the area under investigation.