CIRF-IVC-ADSC2026_ - Text Summarization Using Deep Learning Transformers

The increase in the number of digital texts in various fields including news articles and scientific papers has resulted in the need for text summarization systems. The Transformer model has helped in improving the results for text summarization systems through its recent inclusion in the systems. The model requires a lot of computational resources for its functioning. The authors have proposed a lightweight model for text summarization based on the concept of a hybrid fusion approach. The extractive part of the model helps in identifying the important sentences in the text through the BERT-Tiny model along with weak supervision using the ROUGE model. The abstractive part of the model is based on the DistilBART model as the main concept for the model since the model generates text through its sequence-to-sequence method. The results obtained from the models are then fed into the T5-Small model for generating the summary based on the important information in the two summaries. The proposed framework is evaluated on a multi-domain hybrid dataset consisting of news articles, scientific papers, legal documents, and conversational dialogues. Experimental results demonstrate that the proposed fusion pipeline maintains low computational complexity while achieving ROUGE-1, ROUGE-2, and ROUGE-L scor