UC Irvine

The greatest sources of language data for natural language processing are held by the machine translation development community. That data is potentially more in demand than the MT-systems themselves. The defensive attitude of not making these data available for further development is damaging the natural evolution in the field. Activation generates users and those in turn the number of systems to be bought. However, that activation is stalled primarily by the cost of building an MT-system, i.e. the lack of language data available, and secondly by the fact that the potential buyers of machine translation systems lack the knowledge needed for tuning the system to fit the in-house environment.