A Hybrid Approach to Single Document Extractive Summarization

Extractive summarization is a technique to compress the text in such a manner that only the important original sentences representing the text are considered for the summary. Creating a concise and non-redundant summary is not an easy task. There are a number of techniques which are already available for creating the summaries like graph-based techniques, feature-based techniques, fuzzy logic, etc. In this paper we propose a hybrid approach using graph-based technique (PageRank), MMR along with K-means clustering and feature-based extraction technique along with the fuzzy logic to create the extractive summaries. The experiments show that our hybrid approach gives either better results or comparable results than the individual techniques.