Statistical versus knowledge-based machine translation

In this report, some of the distinguished researches who are actively working on the problem of machine translation discuss their significant works on this field. Yorick Wilks identifies the reasons behind the success of the statistics approaches and argues why they must be integrated with knowledge-based approaches. Ken Church describes the basic idea behind statistical machine translation. Sergie Nireburg, maintains that machine translation is too complex to be handled by statistics and that a pure knowledge-based approach will eventually win out. Lastly, Eduard Hovy, argues that integrated approaches are inevitable and that the precise nature of the integration will depend on the specific application task.