Abstract: Semantic relevancy computation is used to solve structural disambiguity in parsing syntactic. Semantic relevancy computation based on Hownet is proposed, based on semantic similarity computation. The method can compute the relevancy of different POS words using resources of Hownet, such as the examples, the relationship of horizontal and vertical primarily. Consider the explanatory relations of the two primarily to amend the primary distance, when computing the primary distance. According to the symmetry of semantic similarity, the result of relevancy is improved by computing the effect factor of examples. Experimental results show that the results are satisfactory.