Summarizer: A Single Document text summarization using Hybrid Approach

Abstract

Summarization task is a process of diminishing the size of the input text, without altering its overall meaning, by selecting only the significant parts of the input text as an output. Summary of a text can be obtained either by an abstractive approach or an extractive approach. Our system is a hybrid system (that is the combination of abstractive and extractive approach). The first stage of this system is to generate an extract summary using statistical features and a semantic feature. Emotions play a vital role in representing the emotional relationship of the writer thus we have used emotion as a semantic feature. The second stage of this system is to pass the extractive summary to the Novel language generator, which is designed by combining WordNet, Lesk algorithm and part-of-speech tagger, in order to transform extractive summary into abstractive summary, resulting in a hybrid summarizer.  Evaluation of our summarizer was done using DUC 2007 data set as an input and the output i.e., summaries generated by this system were compared to the MS-Word summarizer generated summaries. We achieved significant results after comparison.

Keywords: Summarization, abstract summary, extract summary, hybrid summarization, machine learning techniques, emotion analysis, WordNet

Cite this Article
Mahira Kirmani, Nida Manzoor Hakak. Summarizer: A Single Document Text Summarization using Hybrid Approach. Journal of Advancements in Robotics. 2018; 5(2): 17–27p.