further substantiates the suitability of example-based translation machine for domain-specific translation. Their system is based on examples written in controlled language, a subset of natural language whose grammar and lexicon are restricted in order to minimize ambiguity and complexity. In return, their system outperforms both RBMT and system translation memory systems in evaluation, largely due to the fact that RBMT suffers from greater complexity in fine-tuning its system to support controlled language, while system memory translation requires extremely large volumes of training text which is hard to come by for controlled language. An example-based translation machine system, on the other hand, can be developed with smaller amounts of training examples.