Efficient Voting-Based Extractive Automatic Text Summarization Using Prominent Feature Set

摘要

Automatic text summarization (ATS) is the process of generating a summary by condensing text document by a computer machine. In this paper, we explored voting-based extractive approaches for text summarization. The main issue with most of the feature-based ATS methods is to find optimal feature weights for sentence scoring to optimize the quality of summary. Voting-based methods are sensitive to initial ranking process. We proposed reciprocal ranking-based sentence scoring approach that alleviates the feature weighting and initial ranking problem. The proposed approach uses a specific prominent set of features for initial ranking that further enhance the performance. Experimental results on Document Understating Conference 2002 data-set using ROUGE evaluation matrices shows that our proposed method performs better as compared to other voting-based methods.