Automatic Term and Collocation Extraction from English-Croatian corpus

Term and collocation bases represent valuable additional resources covering specific domain and frequently expressions, which then can be used in further research. The paper presents possible model of building terminology and collocation base, using statistical and linguistic approaches in order to gain experience in building of such resources for the English Croatian language pair. The aim of the paper is not to evaluate tools, but to give an insight into use of tools and to gain experience in building, training and testing of language resources. In the paper, two types of statistically-based term and collocation bases have been compared, created out of the legislative documentation and then filtered through language dependant linguistic patterns.