Text Summarization based on Feature Extraction using GloVe and B-GRU

In the day-to-day life, huge amount of data needed Automated Text Summarization (ATS) methods to identify the useful information. The traditional summarization methods are very complex as it consumes more time for people. In order to solve this issue lot of sentence scoring or ranking methods for labelling the input text is introduced for the summary. Therefore, Feature Extraction (FE) with deep learning techniques are emerging as a solution. However, the Bi-Gated Recurrent Unit (B-GRU) has the limitation in missing the features. To update with a more appropriate solution, the novel approach is incorporated in FE with sentence similarity and word feature vectors separately using Global Vectors for Word Representation (GloVe). This approach also adds B-GRU with sliding windows for more FE with attention layers. This makes the ranking for the important sentence that is included in the summary. The experimental results are implemented in arXiv and PubMed datasets. The proposed method outperforms with the various several state-of-the-art models. This method gains the remarkable Recall-Oriented Understudy for Gisting Evaluation (ROUGE)-l, ROUGE-2 and Recall-Oriented Understudy for Gisting Evaluation Longest (ROUGE-L) scores in both the datasets. It shows the proposed method have the potential with the informative summaries as it has good precision, Recall and F-score.