A Semantic Disambiguation Algorithm Based on HowNet

摘要

This paper proposes a word sense disambiguation algorithm based on How Net. This algorithm makes full use of abundant instances library which provided by How Net to carry on preliminary disambiguation. If it does not match with the database, then the algorithm will extract the keywords inherent defined to ambiguities through the dependency grammar. Finally, combined with the explicit and implicit network relationship in How Net, this paper give different weights to different semantic relations and adjust the weights of the sememes through the appropriate algorithm. By this way it eliminates the ambiguity for ambiguous words.