The diversification of Web search results has been known as an important factor of improving Web search efficiency and user satisfaction. In this paper,the diversification problem is formalized into a maximization problem of facet coverage,and a novel diversification method named KDM is proposed. In KDM,first,keywords representing document facets are extracted from the retrieved documents related to the query. Then,the document facet novelty is calculated according to the co-occurrence and description ability of the keywords. Finally,the documents are re-ranked by considering both the novelty and the relevance to provide diversified search results for users. Experimental results indicate that KDM outperforms other existing approaches in terms of diversification ability.