Syntactic performance analysis of English-Turkish translation tools

In this thesis, the syntactic distance between English-Turkish human translation and the most commonly used automatic translation tools, "Google Translate" and "Bing Translator", have been examined using the text similarity methods. In the study, source sentences are used taken from "MSR Paraphrase Corpus" which was created by the Microsoft Research company and from Tatoeba sentence dictionary which is on the Internet. Source sentences which are in English language are first translated to Turkish by professional translation firm. After that, same sentences are translated to Turkish by Google Translate and Bing Translator tools. These machine translations are compared with the human translation using the syntactic text similarity methods. (Minimum Edit Distance, Matching Words Ratio, Text Length Ratio, Sequence Ratio, Common Word Set Ratio, BLEU and METEOR). Then, stemming is applied to all sentences and they are compared again after stemming. Finally, 100 selected sentences are evaluated by human judgement.