Abstract
MACHINE TRANSLATION METHODS AND CURRENT STATE OF MACHINE TRANSLATION
In this study, machine translation’s emergence, development and the point which it stands today are evaluated, the commonly used machine translation methods in our present day are addressed over the popular machine translation systems (Google Translation, Microsoft Bing Translation, Yandex Translation).
Google Translation which was instructed over documents of United Nations containing 20 billion of words has used the method of “Rule-Based Machine Translation” and the method of “Statistical Machine Translation” in the early years, it has started to use the method of “Neural Machine Translation” that is closer to the human translation in the recent years. It uses this method in 9 languages as English, French, Deutsch, Spanish, Portuguese, Chinese, Japanese, Korean and Turkish.
Yandex Translation which provides translation service in more than 95 languages makes translation by analyzing the languages it works on by statistical methods, not by the language rules which have been already instructed. Yandex has developed this method in recent years and started to use a hybrid model (in combination of statistical machine translation and neural machine translation).
Microsoft Bing Translation offers text translation service in more than 70 languages, it uses the method of Neural Machine Translation in 60 languages and the method of statistical machine translation in 10 languages.
In this article, the statistical machine translation will be evaluated over Turkish-Kyrgyz examples, the neural machine translation will be evaluated over English-Turkish examples.
Keywords
Machine translation, rule-based machine translation, statistical machine translation, neural machine