There are several thousand languages in the world and there is a wealth of knowledge and information in all these languages. To make this knowledge available to all, it needs to be translated to other languages. There is only a limited amount that can be translated by human translators. Machine translation is translation of text from one language to another by computers.
There are two main approaches to MT. The first is rule-based and is inspired from linguistic insights. The second is statistical MT, which is mainly based on ideas from probability and statistics. There can also be a hybrid approach, which combines these two approaches.
The purpose of this course will be to introduce in sufficient detail the techniques for machine translation to those who have little or no background in this area, but have some background in Computer Science or languages/Linguistics. We will try to prepare the participants completing this course to initiate or take part in building MT systems for their own (or someone else's) languages. There will be emphasis on showing MT in practice, not just in theory.
The whole course will consist of some seven lectures of around two or three hours each. Each lecture will be followed by a two hour lab session. These lectures and labs will be covered over a period of seven days. During the course of their lectures, the instructors will refer to actual working MT systems to provide context for the concepts presented in the class. Home assignments can be given if the participants desire. Through labs, the participants will be expected to get some hands on experience with building MT systems, using both the approaches mentioned above.
Application deadline: November 10, 2016
Selection notification: November 15, 2016