Node-Weighted Centrality Ranking for Unsupervised Long Document Summarization

Supervised methods have demonstrated superior performance to unsupervised methods in text summarization. However, supervised methods heavily rely on human-generated summaries, which can be costly and difficult to obtain in large quantities. They also face challenges in summarizing long documents due to input length restrictions. Graph-based methods are frequently employed in unsupervised text summarization owing to their capacity to examine interrelationships between. However, these methods usually depend on unique node weights, resulting in limited mapping capabilities and weak performance on long documents. To address these difficulties, this study proposes an unsupervised method that employs a graph model with augmented node weights with a novel centrality ranking algorithm. Comprehensive experiments on standard datasets demonstrate the effectiveness of the proposed method, which outperforms both unsupervised and supervised techniques when evaluated using the ROUGE metric.