GramTrans is a research-based system using innovative technology such as Constraint Grammar dependency parsing and dependency-based polysemy resolution. Some core ideas, methods and evaluation are provided in the following two reference articles:
- Bick, Eckhard (2007), Dan2eng: Wide-Coverage Danish-English Machine Translation, In: Bente Maegaard (ed.), Proceedings of Machine Translation Summit XI, 10-14. Sept. 2007, Copenhagen, Denmark. pp. 37-43
- Bick, Eckhard & Nygaard, Lars (2007). Using Danish as a CG Interlingua. A Wide-Coverage Norwegian-English Machine Translation System. In: Proceedings of the 16th Nordic Conference of Computational Linguistics. Tartu, Estonia. ISBN978-9985-4-0514-7
At an academical level, GramTrans has a close cooperation with the University of Southern Denmark. Thus, some of the language analysis tools (taggers and parsers) used in our MT system, are also used for teaching and corpus annotation within the framework of the VISL project at the Institute of Language and Communication, and though machine translation as such is a purely commercial project, there is a considerable amount of development synergy between GramTrans and SDU in various areas of language technology research.
VISL (Visual Interactive Syntax Learning), as its name indicates, was originally conceived as a research-supported grammar teaching initiative at its inception in 1996, where it received funding from CTU (Center for Technology-Supported Teaching). However, activities soon spanned a variety of research topics in computational linguistics, allowing internet-based teaching applications to draw on cutting-edge language technology. Today VISL offers manually revised langage data (treebanks) for 27 languages, which can be used through a variety of internet-based exercises, quizzes and educational games. In collaboration with the Ministry of Education the system has been adapted to cover not only the university level, but also different levels and types of primary and secondary schools. with a current focus on language awareness teaching. In cooperation with GrammarSoft, VISL has developed rule-based NLP tools (Natural Language Processing), such as taggers and parsers, for 9 different languages. These tools allow the grammatical annotation of large corpora and support application-oriented research within the general field of language technology. In an effort to make language data accessible to non-specialists, such as school teachers and students, VISL maintains a user-friendly corpus-search interface with corpora covering 12 languages and 1 billion words. Both the teaching and corpus interfaces are subject to repeated evaluation and adaptations to cater for the needs of individual teaching or research projects.
VISL has participated in a number of international joint projects in Computational Linguistics and Computer-Aided Language Learning, such as the PaNoLa Constraint Grammar projects, the Nordic Treebank Network and the Nomen Nescio NER project, all funded by the Nordic Council of Ministries, as well as bilateral projects for Portuguese, French, English etc.