Text Summarization of Batak Toba Language Using Bidirectional Auto-Regressive Transformers Model

This developed an automatic text summarization system for the Batak Toba language using the Bidirectional Auto-Regressive Transformers (BART) model. The system aims to simplify lengthy Batak language texts, making them more accessible and understandable for readers, especially those frequently engaging with Batak language content. To ensure summarization accuracy, the model was also trained on a dataset specific to Batak language contexts, allowing it to effectively capture linguistic nuances. During development, the system utilized natural language processing (NLP) techniques to process data. The data used in this study is not updated in real-time, making performance dependent on manually updated datasets. The research methodology includes dataset collection from various sources, data preprocessing, model training, model testing, and performance evaluation using ROUGE metrics. Model evaluation yielded F1 scores of 0.8437 for ROUGE-1, 0.8065 for ROUGE-2, and 0.8437 for ROUGE-L, indicating that the developed system successfully produces coherent text summaries. For further development, expanding the dataset and improving model quality are recommended to support more optimal text summarization that meets user needs.