Extractive Summarization of Text Using Supervised and Unsupervised Techniques

The information has grown at a rapid pace in the recent past. The overwhelming amount of data available may be a lot for the consumers to absorb and retain. Various methods have been developed over the years to tackle the issue of over-burdening consumers. The research has led to the topic of text summarization. Text summarization is the method of producing a concise version of text by retaining the most relevant information. The paper focuses on the extractive summarization of single documents. Majority of the previous studies revolve around unsupervised methods while the supervised methods mainly comprise of models trained on the DUC datasets. The proposed work in this paper provides a comparison of supervised and unsupervised methods. The base of the supervised model is on a convolutional neural network (CNN) whereas the base of the unsupervised method is on K-Means clustering. The proposed model is used to evaluate the BBC News Summary dataset. The performance is a comparison with the previously proposed systems.