Three Layer Automatic Text Summarization System

Text summarization system is an important and widely needed application. In this paper we introduce a novel method in order to summarize texts. Our approach is different from existing approaches in that we study the text structure and find local coherent sections in the text. For this reason we look for coherent parts by comparing semantic relatedness of all the paragraphs in a document, all the sentences in every paragraph and finally all the words comprising a sentence. In this process, we search for rich semantic sentences in strong paragraphs and documents. We also consider title Scores to get more similar sentences to the theme of the documents for final summary. DUC 2002 dataset is used in order to test our system of summarizing. Experimental results show that our approach significantly outperforms the best DUC summarization methods and the most of the state-of-the-art approaches.

کلیدواژه ها:

Text summarization ، Extractive Summary ، Abstractive Summary ، Similarity measure