Word Sense Disambiguation as an Integer Linear Programming Problem

摘要

We present an integer linear programming model of word sense disambiguation. Given a sentence, an inventory of possible senses per word, and a sense relatedness measure, the model assigns to the sentence's word occurrences the senses that maximize the total pair-wise sense relatedness. Experimental results show that our model, with two unsupervised sense relatedness measures, compares well against two other prominent unsupervised word sense disambiguation methods.