Место издания:University of Bologna, Forlì Campus Bologna, Italy
Первая страница:50
Последняя страница:50
Аннотация:This paper presents novel approaches to phrase alignment for example-based machine translation. We use matching of delimiters instead of word matching while determining fragment borders. We follow a monotonic machine translation approach, for which we develop an efficient and flexible partial reordering that allows introducing different reordering constraints. We have invented a procedure using the on-line translation engine. The engine translates the source text to the target one sentence-by-sentence. Then the Google-translated text is aligned with the human-translated text. It is not a trivial procedure because two translations may have different number of sentences and sentence borders may not coincide in both. For the alignment we use a dynamic programming method. We adopted number of coincided words in two translated sentences as a measure of proximity between them. Otherwise the measure of proximity is the Lowenstein distance between two same language sentences. Such an approach could be used for different language pairs within the scope of the Google translator engine. The method was implemented as a free accessible procedure coded in PHP language. As a side effect we have used this procedure to find the translation of psychological terms to find the proper Russian term for internationally adopted English one.