Machine Translation (MT) is the translation of text by a computer. There are two types of machine translation systems: rule-based and statistical systems. Rule-based systems use a combination of dictionaries and (grammar) rules. Statistical systems do not work with 'rules', but derive all information necessary to translate from large collections of texts. To produce high-quality translations, humans still need to intervene in the process either by making the input more suitable for MT (pre-editing) or changing the output (post-editing).
The course deals with the following topics:
- Challenges for MT
- Architecture of MT systems (rule-based MT, statistical MT and neural MT systems; interactive and adaptive systems)
- Evaluation of MT output (automatic vs. manual evaluation methods)
- Post-editing and post-editing tools
- Integration of MT in the translation workflow
- Creation and evaluation of a customized MT engine