Intelligent Abstractive Text Summarization using Hybrid Word2Vec and Swin Transformer for Long Documents

Abstract

In this paper, a novel hybrid Swin-T transformer based automatic text summarization model is proposed for both short and long documents. It involves various phases namely data acquisition, preprocessing, Word to Vector conversion, semantic feature extraction using Swin-T transformer, clustering the similar sentences using K-medoid clustering and finally ranking the sentences and summary generation using Bi-directional Gated Recurrent Unit (Bi-GRU). This setup outperforms the existing state-of-the-art systems in terms of evaluation score named ROUGE score such as ROUGE 1, ROUGE 2 and ROUGE L for the short benchmark datasets as well as a long user created arXiv dataset.