Evaluating Semantic similarity is widely used in areas range from Psychology, Linguistics, Cognitive Science to Artificial Intelligence. This paper means to the merely use of HowNet to evaluate Information Content as the semantic similarity of two terms or word senses. While the conventional ways of measuring the IC of word senses must depend on both an ontology like WordNet and a large corpus, the experiment proves that the semantic similarity measured in this method is easier to calculate and more closely with human judgments, as HowNet has an elaborate way to represent descriptive object.