(Tagged) Centroid-based Hierarchical Ordered Processing for Summarization

Text summarization is critical in condensing large bodies of text while preserving key information, making large amounts of data and information digestible. Traditional approaches, including graph-based algorithms and deep learning models, often involve complex pipelines that are computationally expensive and difficult to interpret, or require large amounts of pre-training. In this paper, we introduce \textbf{CHOPS} (Centroid-based Hierarchical Ordered Processing for Summarization), a lightweight and efficient extractive summarization method that segments text into manageable chunks, computes a centroid representation, and then selects representative sentences using cosine similarity. Additionally, we propose \textbf{T-CHOPS} (Tagged Centroid-based Hierarchical Ordered Processing for Summarization), an extension that retains references to sentence positions, enhancing transparency through traceability in summarization. Our approach demonstrates competitive performance in ROUGE and BERTScore evaluations, while significantly reducing computational costs compared to existing methods. By streamlining text summarization into an interpretable and efficient process, CHOPS and T-CHOPS offer practical solutions for real-time and resource-constrained applications.