Machine Translation is the process of converting one natural language to another by a computer without a human effort. There are some key difficulties in MT; since there are a variety of languages, grammars and alphabets, or there is not exactly one correct answer while doing translation MT is a challenging task.
There are 3 major types of MT:
- System-Rule Based Machine Translation
- Statistical Machine Translation
- Neural Machine Translation
1) System-Rule Based Machine Translation
This type of translation needs a dictionary or rule sets for both the source language and the target language. RBMT has more linguistics about both the source and the target languages to develop a syntactic & semantic analysis to do the translation, then the other methods.
Examples: Systran, Apertium, GramTrans
2) Statistical Machine Translation
This approach uses statistical methods based on the bilingual corpora to generate the translation. By using statistical methods, it aims to minimize the error of the translation and maximize the chance of the true results. SMT is a data-driven translation technique and it only needs a corpus of examples for both the source and the target language texts.
Examples: Google Translate (2006–2016), Microsoft Translator (up to 2016)
3) Neural Machine Translation
Neural Machine Translation (NMT) is a deep learning-based approach to generate the translation. This approach uses neural network models to learn a statistical model for machine translation. Compared to the other methods, NMT does not need a pipeline to achieve the result.
Examples: Google Translate (from 2016), Microsoft Translator (from 2016), Facebook Translation.