Semantic textual relatedness: A hybrid method

Abstract

Semantic relatedness measure play an important roles in Natural Language Processing (NLP) tasks. By using the knowledge bases and current methods, the semantic relatedness measure could be done. This time, we implement the hybrid method in measuring semantic relatedness between the pair of word. Hybrid method is one of the most popular method that used to measures semantic relatedness. Hybrid method combine several methods in order to optimize correlation coefficient. We will combine two method that are often used. These method are path-based and gloss-based method. Path-based methods is a method based on taxonomy that were defined in knowledge bases. While the measurement of gloss-based methods are based on description and the conceptual meaning of the word. Knowledge bases that were used is WordNet and the selection of hybrid method is intended to prove whether there is an increase in correlation by combining those methods. Based on test scenario that were conducted, hybrid method gives higher correlation compared with the one that only used one method. The correlation result between words in hybrid method is 0.337.