Moses has become a standard for open source statistically-based machine translation systems. For many European language pairs, the Europarl corpus has been an accompanying material for training and testing. However, Europarl is not available for Norwegian, and there has not been a comparable corpus for experimenting with machine translation between Norwegian and other languages. In the later years, Språkbanken has published several sets of translated texts of EU laws.
The goal of this project is first to edit and sentence align these texts such that they can be processed by a system like Moses. Then the goal is to experiment with Moses on these texts and evaluate the results. From there on, the project is open and could proceed in several directions: Consider adding additional material and test the system on other domains, consider specialized modules for language specific differences, e.g. compounds, or experimenting with other approaches to the translation process.
The candidate should have a good knowledge of Norwegian and English and should have taken the course INF5820 Language technological applications.