Cosine Similarity Based Non-Redundant Summarizer

Abstract

As summarization inherently assign more weights to the more important sentences in an IT news article, this may improve the precision and recall score of the article. This paper uses the redundant reducing algorithm to improve the summarization technique. Redundancy in summaries was reduced to different levels and its effect on summarization meaning was investigated. This paper presents the summarization system: an efficient method to extract the most important and non-redundant “n” sentence segments based on sum of similarity. In this system, the characteristics are listed as follows: 1) using preprocessing to delete the additional information that won't turn up in the summarization; 2) getting sentence segment by syntax; 3) redesigning the vector similarity between a pair of sentences by using sum of similarity. Because of all of above properties, experimental results show that this system’s approach compares favorably with auto summarizer of MS word.