Convolutional Neural Network based for Automatic Text Summarization

Abstract

In recent times, the apps for the processing of a natural language has been formed and generated through the use of intelligent and soft computing methods that allow computer systems to practically mimic practices related to the process of human texts like the detection of plagiarism, determination of the pattern as well as machine translation, Thereafter, Text summarization serves as the procedure of abridging writing within consolidated structures. 'Automatic text summarization' or the ATS is when a computer system is used to create a text summarization. In this study, the researchers have introduced a novel ATS system, i.e., CNN-ATS, which is a convolutional neural network that enables to Automatic text summarization using a text matrix representation. CNN-ATS is a deep learning system that was used to evaluate the improvements resulting from the increase in the depth to determine the better CNN configurations, assess the sentences, and determine the most informative one. Sentences deemed important are extracted for document summarization. The researchers have investigated this novel convolutional network depth for determining its accuracy during the informative sentences selection for each input text document. The experiment findings of the proposed method are based on the Convolutional Neural Network that uses 26 different configurations. It demonstrates that the resulting summaries have the potential to be better compared to other summaries. DUC 2002 served as the data warehouse. Some of the news articles were used as input in this experiment. Through this method, a new matrix representation was utilized for every sentence. The system summaries were examined by using the ROUGE tool kit at 95% confidence intervals, in which results were extracted by employing average recall, F-measure and precision from ROUGE-1, 2, and L.