Generation of non-redundant summary based on sum of similarity

This paper proposes an efficient method to extract the most important and non-redundant sentence segments based on sum of similarity. The new characteristics of our method are listed as follows: 1) we use preprocessing to delete the additional information (comma parenthesis) that won´t turn up in the summarization; 2) redesign the vector similarity between a pair of sentences by using sum of similarity; 3) in order to maximize topic diversity, we use a strikingly different redundancy reducing ways other than MMR.. Experimental results show that our approach compares favorably with some other summary systems.