ABSTRACT

Parallel corpora ideally consist of texts in one language and translations of those same texts in another language. One option open to the professional translator (and even more so to the translation agency) is to construct their own parallel corpora, using source and target texts of previous translation jobs. Translation memories can themselves be considered parallel corpora, but since they contain pairs of segments rather than pairs of texts, and only one translation per segment, they offer less contextual information, and none on frequency or alternatives. Thus constructing and using comparable corpora may develop awareness of cross-cultural similarities and differences; parallel corpora may develop awareness of strategic alternatives and the role of context. However the interfaces offered by the major search engines do not facilitate this, and there are obvious copyright implications attached to any process which involves automatic downloading (rather than consultation) of documents.