Machine Translation for Freelance Translators
This study aims to show that the rapidly advancing machine translation technology can be useful for freelance translators if it is understood sufficiently and implemented correctly to their workflows. This fact has become clearer with the advances in Neural Machine Translation (NMT) which, seemingly, has brought translation into/from morphologically rich languages such as Turkish up to a higher and usable quality according to automatic evaluation metrics (BLEU, TER etc.). For this reason, Turkish became one of the first languages to be selected for Google NMT.
Taking Turkish as an example of morphologically rich languages, this study will firstly compare neural machine translation and statistical machine translation (SMT) through automatic evaluation metrics and human evaluations. To extend the research data, many different text types such as medical, legal, sports, localization and subtitling will be used. This will provide a thorough insight about which system to select for some language pairs. And based on these findings, MT useage cases and available translation environment tools (TEnTs) supporting MT will be explained and discussed by a special focus on freelance translators in the industry especially because up until now, it has been assumed that MT is an expensive technology and freelance translators are only passive users of its output or the production streamline suggested by large LSPs. It will be shown that freelance translators can benefit from the available technologies and free and open source machine translation solutions such as MTradumàtica.
Finally, it is true that NMT has created a big hype both in translation studies and translation industry. The study will try to demystify the myths about NMT and provide insights about its real position considering the findings of the current study on Turkish as well as other recent studies on different language pairs.