Elhuyar Fundazioa - Language Services 

OpenMT 

Aims and general description

The main aim of the OpenMT project is to develop open-code architectures for machine translation by using hybrid methods and advanced semantic processors. The most important features of this project are as follows:

• Specification of open-code architectures for machine translation systems.
• Creation of a hybrid system combining different machine translation paradigms. By combining systems based on rules (Rule-Based Machine Translation, RBMT) statistical systems (SMT) and systems based on examples (Example-Based MT, EBMT).
• Integration of advanced processors of semantics and syntax. Furthermore, work will be conducted with Semantic Role Labeling systems and on Word Sense Disambiguation (WSD) by adapting to hypothetical translation selection and priorities. ?????
• Evaluation. A general tool for evaluation and a corpus will be created. The system will be evaluated on the basis of combined criteria and a range of languages and domains will be used. As is usual nowadays, we will be seeking to build a system that will fit human criteria as closely as possible.

The project is assisted by the Spanish Ministry for Education and Science within the "Proyectos I+D 2006-2009" grant application.

Participants:

• Elhuyar Fundazioa
• IXA Taldea (EHU/UPV)
• Grupo TALP(UPC)

Further information

• Alegria I., Díaz de Ilarraza A., Igartua J., Labaka G., Laskurain B., Lersundi M., Mayor A., Sarasola K., A. Casillas, X. Saralegi. 2008. Mixing Approaches to MT for Basque: Selecting the best output from RBMT, EBMT and SMT. MATMT2008 workshop: Mixing Approaches to Machine Translation.
• OpenMT project