Comparing Semantic Relatedness between Word Pairs in Portuguese Using Wikipedia

Abstract

The growth of available data in digital format has been facilitating the development of new models to automatically infer the semantic similarity between word pairs. However, there are still many natural languages without sufficient resources to evaluate measures of semantic relatedness. In this paper we translated word pairs from a well-known baseline for evaluating semantic relatedness measures into Portuguese and performed a manual evaluation of each pair. We compared the correlation with similar datasets in other languages and generated LSA models from Wikipedia articles in order to verify the pertinence of each dataset and how semantic similarity conveys across languages.