Using the Web to Disambiguate Acronyms

This paper proposes an automatic method for disambiguating an acronym with multiple definitions, considering the context surrounding the acronym. First, the method obtains the Web pages that include both the acronym and its definitions. Second, the method feeds them to the machine learner. Cross-validation tests results indicate that the current accuracy of obtaining the appropriate definition for an acronym is around 92% for two ambiguous definitions and around 86% for five ambiguous definitions.