Machine translation involves automatic translation of sentences in one language to another. The statistical approaches for doing machine translation have proved effective in recent years, specially for translating between European languages. However, its effectiveness in translating sentences from English and Indian languages needs to be explored further.

This shared tasks aims to collectively explore a variety of ways of combining statistical techniques with linguistic inputs to improve a baseline statistical machine translation system from English to Hindi.


Contest

In the contest, training data will be provided to the contestants. It will consist of English-Hindi parallel corpus. The contestants will have to train their systems on the data. A development corpus will also be provided to refine and improve their system. The final contest will be held on November, 2008 with the test data. A workshop will be held as a part of ICON to allow the short listed candidates to present their techniques and results.

        - Linguistic tools     - Not trained on test/devel set .

        - Lingu. Resources     - Not extracted from test/devel set .

        - Any annotated data   - Data not from test/devel set .