The problem of finding similarity between natural language sentences is crucial for many applications in Natural Language Processing (NLP). Moreover, accurate calculation of similarity between sentences is highly needed. Many approaches depends on word-to-word similarity to measure sentences similarity. This paper proposes a new approach to improve accuracy of sentences similarity calculation. The proposed approach combines different similarity measures in calculation of sentences similarity. In addition to traditional word-to-word similarity measure the proposed approach exploits sentences semantic structure. Discourse representation structure (DRS) which is a semantic representation for natural sentences is generated and used to calculated structure similarity. Furthermore, word order similarity is measured to consider order of words in sentences. Experiments show that exploiting structural information achieves good results. Moreover, the proposed method outperforms the current approaches on Pilot standard benchmark dataset achieving 0.8813 peasron correlation with human similarity.