A survey of Domain Adaptation in Machine Translation: Towards a refinement of domain space

Domain adaptation is a recurring problem in Artificial Intelligence in general and Machine Translation (MT) specifically. A system crafted to deal with one particular type of problem often fails when subjected to another, even a closely related one. In MT this is manifested by a system's inability to translate different types of texts (from different domains) with the same confidence. The paper addresses the characterization of language domains and their treatment in the MT literature. We will visit classical linguistic theory as well as cognitive linguistics, and suggest that a refinement of the types of domains along more dimensions can be useful when simultaneously adapting to multiple domains.