Effective document summarization: a hybrid clustering approach using transformer model
Authors:
- Divya S.DSDivya S.
- Sripriya N.SNSripriya N.
Open Access
- 27 February 2026
- journal article
- Published by PeerJ in PeerJ Computer Science
- Vol. 12, e3201
Abstract
The rapid growth of web documents has led to the need for automatic document summarization so readers can quickly browse the document for information. The challenge in extractive summarization is extracting the relevant information while ignoring redundant content to generate a precise summary. This article introduces a hybrid clustering algorithm that ignores outliers to eliminate irrelevant content and group similar sentences. The proposed methodology combines the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) and Balanced Iterative Reducing and Clustering using Hierarchies (BIRCH) algorithms, with DBSCAN forming densely coupled clusters and BIRCH splitting these massive clusters into sub-clusters. This hybrid approach facilitates effective sentence selection from each cluster and overcomes the limitations found when using DBSCAN and BIRCH individually. This proposed method further addresses the arbitrary decision issues found in hierarchical clustering by carefully tuning the Epsilon (ε) and MinPts parameters of DBSCAN to manage larger datasets. The systematic evaluation of the proposed summarization model using hybrid clustering was performed using the CNN/DailyMail dataset and measured in terms of Recall-Oriented Understudy for Gisting Evaluation (ROUGE) metrics. The model achieved values of 0.429, 0.213, and 0.371 for ROUGE-1, ROUGE-2, and ROUGE-L, respectively. The proposed summarization model outperforms the state-of-the-art unsupervised and supervised automatic text summarization models.Keywords
This publication has 29 references indexed in Scilit:
- An attention enhanced sentence feature network for subtitle extraction and summarizationExpert Systems with Applications, 2021
- A hybrid deep learning architecture for opinion-oriented multi-document summarization based on multi-feature fusionKnowledge-Based Systems, 2021
- A Two-stage Chinese text summarization algorithm using keyword information and adversarial learningNeurocomputing, 2021
- A Hierarchical Structured Self-Attentive Model for Extractive Document Summarization (HSSAS)IEEE Access, 2018
- Cluster Analysis of Rural, Urban, and Curbside Atmospheric Particle Size DataEnvironmental Science & Technology, 2009
- Multi-document summarization using cluster-based link analysis2008
- Hierarchical Clustering Algorithms for Document DatasetsData Mining and Knowledge Discovery, 2005
- The anatomy of a large-scale hypertextual Web search engineComputer Networks and ISDN Systems, 1998
- BIRCHACM SIGMOD Record, 1996
- Indexing by latent semantic analysisJournal of the American Society for Information Science, 1990