Disambiguating Coordinations Using Word Distribution Information

In this paper we present some heuristics for resolving coordination ambiguities. This type of ambiguity is one of the most pervasive and challenging. We test the hypothesis that the most likely reading of a coordination can be predicted using word distribution information from a generic corpus. The measures that we use are: the relative frequency of the coordination in the corpus, the distributional similarity of the coordinated words, and the collocation frequency between the coordinated words and their modifiers. The heuristics that we present based on these measures have varying but useful predictive power. They also take into account our view that many ambiguities cannot be effectively disambiguated, since human perceptions