Multilingual corpora: models, methods, uses

Abstract

This paper gives an account of our work on multilingual corpora at the University of Oslo. Different corpus models are presented, in particular the bi-directional translation model used in building the English-Norwegian Parallel Corpus (ENPC). The steps in building the ENPC are briefly outlined, with some comments on problems encountered. Most of the paper is concerned with the use of this sort of corpus. There is a brief introduction to the search program developed for the ENPC. The main point of the paper is to show that the bi-directional corpus model makes it possible to carry out contrastive studies and simultaneously control for translation effects. Bengt Altenberg’s notion of mutual correspondence is introduced, with reference to his study of adverbial connectors in English and Swedish, based on the sister corpus of the ENPC, the English-Swedish Parallel Corpus. As an illustration of translation effects, there are some comments on the distribution of two verbs of posture in the ENPC: Norwegian stå and its English cognate stand. The difference is sharpest in original texts, while the distribution in the translations is clearly tinged by the source texts.