Enhancing Abstractive Multi-Document Summarization with Bert2Bert Model for Indonesian Language

Abstract

This study investigates the effectiveness of the proposed Bert2Bert and Bert2Bert+Xtreme models in improving abstract multi-document summarization for the Indonesian language. This study uses the transformer model as a basis for developing the proposed Bert2Bert and Bert2Bert+Xtreme models. The results of the model evaluation with the Liputan6 dataset using ROUGE-1, ROUGE-2, ROUGE-L, and BERTScore show that the proposed models have slight improvements over previous research models with Bert2Bert being better than Bert2Bert+Xtreme. Despite the challenges posed by limited reference summarization for Indonesian documents, content-based analysis using readability metrics, including FKGL, GFI, and Dwiyanto Djoko Pranowo revealed that the summaries generated by Bert2Bert and Bert2Bert+Xtreme are at a moderate readability level, which means they are suitable for adult readers and in line with the target audience of the news portal.