Part of Speech Annotation of Intermediate Versions in the Keystroke Logged Translation Corpus

Translation process data contains noncanonical features such as incomplete word tokens, non-sequential string modifications and syntactically deficient structures. While these features are often removed for the final translation product, they are present in the unfolding text (i.e. intermediate translation versions). This paper describes tools developed to semi-automatically process intermediate versions of translation data to facilitate quantitative analysis of linguistic means employed in translation strategies. We examine the data from a translation experiment with the help of