Extractive text summarization is a fundamental task to distill documents into concise summaries. It select sentences on the basis of their importance and create a summary out of the selected sentences. This work proposes a framework to BERT (Bidirectional Encoder Representations from Transformers) with weighted pooling for extractive summarization on datasets comprising legal case documents and scientific articles. Weighted pooling enhances BERT’s ability to discern sentence importance, thereby improving the selection of salient sentences for summary generation. We evaluated summary using three evaluation metrices Rouge-L, BERT and BLEU score. Experimental results demonstrate that the proposed method effectively produces coherent and informative summaries compared to other existing approaches such, BERT and RoBERTa.