Finding semantically related words is a first step in the direction of automatic ontology building. Guided by the view that similar words occur in similar contexts, we looked at the syntactic context of words to measure their semantic similarity. Words that occur in a direct object relation with the verb drink, for instance, have something in common (liquidity,...). Co-occurrence data for common nouns and proper names, for several syntactic relations, was collected from an automatically parsed corpus of 78 million words of newspaper text. We used several vector-based methods to compute the distributional similarity between words. Using Dutch EuroWordNet as evaluation standard, we investigated which vector-based method and which combination of syntactic relations is the strongest predictor of semantic similarity. 1

The task of automatically acquiring semantically related words have led people to study distributional similarity. The distributional hypothesis states that words that are similar share similar contexts. In this paper we present a technique that aims at improving the performance of a syntax-based distributional method by augmenting the original input of the system (syntactic co-occurrences) with the output of the system (nearest neighbours). This technique is based on the idea of the transitivity of similarity. 1