A novel hybrid system with hierarchical semantic conceptual dependency parsing for documents text summarization

Abstract

The rapid development of Internet technologies and the growing number of users who exchange a huge information in a different subject causes the emergence need to extract only the important and useful information from multiple documents. This extraction is text summarization which reduces the time in predicting the purpose of these documents and a mean to access effective and right decisions in important fields. The huge amount of information lead to the problem of memory overload. So, text summarization will minimize the effects of this problem. To overcome the problems of time consuming in extracting a proper information from multiple documents and reduce the redundant information which affect the processing overhead, a novel approach for summarizing information based on Hierarchical Conceptual Dependency of words and sentences representation presented in this paper. Moreover, concepts in text documents with score and token ranks are semantic networks representation of this text achieved for summarization. The proposed technique applied to documents d061j to d070f which are available to compare the results of any proposed approach in text summarization field with the experts output of summarization for the same documents indicated as a (reference summary). The proposed system was evaluated to check the accuracy of results and the tested dataset are Documents Understanding Conference (DUC 2002). Accuracy and efficiency of the output summarization evaluated using ROUGE–N and ROUGE-L metric. Hardware Resources used was Laptop Dell CORE "I3" with RAM 4 GB. Java software was the environment platform for this output. Comparing output results with the related algorithms depicted in this paper for the same target documents reflects that the similarity values between output summary via applying the proposed approach are more appropriate, accurate and minimum cost effective from the related metods.