Discriminative Word Alignment (Internship)

Discriminative word alignment is a DCU School of Computing-funded internship, 4 months in duration, with 1 academic partner (Dublin City University).

The research project aims to explore the discriminative approach to word alignment, especially for morphologically-complex languages, such as Czech. The key techniques to be employed are structured prediction based on Perceptron and/or Support Vector Machines and the minimum weight edge cover in a bipartite graph used as the search algorithm. The feature set will include multiple association measures on word and morpheme levels, as well as positional and other features.

Please contact

• Dr. Pavel Pecina