Zur maschinenunterstützten Übersetzung von Patentschriften: Der Stellenwert der internen Rekurrenz

Abstract

The empirical study presented in this paper explores the relationship between the linguistic features of patent specifications as a text genre and the benefits to be gained from using
CAT tools for their translation. The database is a corpus of 60 complete original (i.e. non-translated) texts selected in accordance with the requirements of the task at hand. The study focuses on analysing patterns of intratextual recurrence that are typical of this genre with a view to examining their impact on the retrieval performance of commercial translation memory (TM) systems. Using the model developed by Reinke (2004) for comparing sentence content, it also looks at genre characteristics that may affect the usability of the search results. The test software was the Translator's Workbench by SDL/Trados (version 7.0.0). The results confirm that it makes sense from a linguistic point of view to use TM systems for patent translations for two reasons: first, because the corpus was found to contain a broad spectrum of recurring sentences, clauses and complex phrases, which is characteristic of this text genre, and second, because the analysis showed that the structural, argumentative and functional features of patent documents have a positive effect
on the usability of the search results. The study is also of practical interest because patents are a frequently translated LSP genre, which, despite its high degree of linguistic standardisation, has so far not been one of the areas where memory-based translation tools are routinely used.