Study on similarity of simple questions based on the catering field

The similarity between sentences is a theoretical basis and key technology to the question answering system. The method presented in this paper is as follows. Firstly, the dependency question sets are obtained and the key words are extracted from the major components of the question sentences and the target question form the related libraries, and then the candidate question sets are obtained through comparing their field words and key words. Secondly, take the HowNet as the semantic knowledge resource to compute the semantic similarity; and lastly the sentences with the largest similarity are returned. This accuracy of the method to calculate the similarity between the simple sentences is up to 85% in catering field. The experimental results show that the approach presented in this paper achieves good results in specific areas.