Resource Efficient Abstractive Text Summarization in Indonesian with ALBERT

In the era of information overload, there is an urgent need for advanced text summarization techniques. The demand for quick and effective news summarization calls for innovative approaches to streamline information extraction processes. This research presents a new method for text summarization in Indonesian by applying the resource-efficient ALBERT (A Lite BERT) model. Our model has achieved notable ROUGE scores, which include ROUGE-1 score of 0.45, ROUGE-2 score of 0.40 and ROUGE-L score of 0.44 in summarizing tasks. This performance was attained after only 50 minutes of training, demonstrating its efficiency in comparison to classical models. This efficiency highlights how ALBERT can accelerate training without sacrificing effectiveness. The study emphasizes the potential of resource-efficient models like ALBERT to overcome computational constraints associated with training large language models, contributing significantly to the field of text summarization.