Identifying semantic relations in a specialized corpus through distributional analysis of a cooccurrence tensor