Abstract

Abstract: Prior work on computing semantic relatedness of words focusedon representing their meaning in isolation, effectivelydisregarding inter-word affinities. We propose a large-scale data mining approach to learning word-word relatedness,where known pairs of related words imposeconstraints on the learning process.Our method, called CLEAR, is shown to significantly outperformpreviously published approaches. The proposed method is basedon first principles, and is generic enough to exploit diversetypes of text corpora, while having the flexibility to imposeconstraints on the derived word similarities. We also makepublicly available a new labeled dataset for evaluating wordrelatedness algorithms, which we believe to be the largest suchdataset to date.