Sentence Alignment as the Basis For Translation Memory Database
Sanja Seljan, Angelina Gašpar, Damir Pavuna · Repozitorij Filozofskog fakulteta u Zagrebu' at University of Zagreb (University of Zagreb) · 2007
Sentence alignment represents the basis for computer-assisted translation (CAT), terminology management, term extraction, word alignment and crosslinguistic information retrieval.Created out of the sentence alignment process, translation memory (TM) represents the basis for further research in translation equivalencies.Automatic sentence alignment, based on parallel texts, faces two types of problems: robustness and discrepancies between source and target texts in layout and omissions which have an influence on the accuracy of the alignment process.The aim of the paper is to present research on the sentence alignment process carried out on the Croatian-English parallel texts (laws, regulations, acts and decisions) and implemented by the alignment tool WinAlign 7.5.0 by SDL Trados 2006 Professional.The alignment process and its impact on the creation of translation memories is presented through comparison of translation memories that differ regarding the levels of expert intervention in the set up of the alignment program and preparation of the source text for the segmentation.Recommendations for further development using statistical analysis, automatic learning techniques and language knowledge are suggested.