Automatic summarization model based on clustering algorithm.

Area of Science:

• Natural Language Processing
• Artificial Intelligence
• Information Retrieval

Background:

• Extractive document summarization often results in semantically redundant summaries due to sentence selection methods.
• Existing methods struggle to balance informativeness with conciseness.

Purpose of the Study:

• To propose a novel model for extractive document summarization that reduces semantic redundancy.
• To enhance summary quality by selecting diverse sentences from the source document.

Main Methods:

• Utilizing K-means clustering algorithm to identify and select semantically different sentences.
• Improving the BERT model for more effective sentence scoring and selection.
• Evaluating the model on the CNN/DailyMail datasets using ROUGE scores.

Main Results:

• The proposed model significantly reduces semantic redundancy in generated summaries.
• Achieved improved accuracy and reduced repetition compared to six baseline methods.
• Demonstrated effectiveness against traditional and state-of-the-art deep learning models.

Conclusions:

• The integration of clustering algorithms with an enhanced BERT model effectively addresses semantic redundancy in extractive summarization.
• This approach yields more accurate and less repetitive document summaries.
• The findings validate the proposed method's superiority in generating high-quality summaries.