Collaborative Ranking-Based Text Summarization Using a Metaheuristic Approach

Abstract

In the present work, a novel approach for improvement in automatic text summarization has been proposed. We introduce a different model for summarization problem by exploiting the strengths of different techniques like metaheuristic approaches and collaborative ranking. First, the sentences of document are scored via two methods. Method one assigns the weight to each text feature using a new metaheuristic approach ‘Jaya’  and scores the sentences by linearly combining these feature scores with their optimal weights. Method two scores the sentences by simple averaging the scores of each text feature. Moreover, the sentences are then ranked according to these scores which generates two sets of ranks for the documents. To calculate the final ranking of sentences, the concept of collaborative ranking has been adopted. The implementation of the proposed approach has been done in Python 3.5 in Anaconda environment. The experiments are performed on the DUC 2002 dataset using the co-selection-based performance parameter. We show empirically that the proposed method is viable and effective for extractive text summarization.