Advanced Extractive Summarization of Indonesian Texts Using LSTM Models
Muhammad Imran Faisal, Shafira Halmahera, Vivin Octavia Cahyani, Abdul Aziz, Ashri Shabrina Afrah, Supriyono · 2024
In the current digital era, the volume of available information continues to increase, especially in the form of online articles. This surge presents a challenge for readers to quickly grasp the core information. This study aims to develop an advanced extractive summarization model for Indonesian texts using Long Short-Term Memory (LSTM) networks. The research process involves several key steps: preprocessing the text data, feature extraction using Word2Vec, training the LSTM model, and evaluating the generated summaries using ROUGE-1 metrics. The LSTM model is optimized with a learning rate of 0.001 and trained over 30–50 epochs. The results demonstrate that the LSTM model achieves superior performance in generating coherent and informative summaries. The highest accuracy attained by the model is 0.88 with the optimal performance observed at a learning rate of 0.001 and 50 epochs. Evaluation based on ROUGE-1 metrics indicates that the model can produce summaries with a high F-measure of 0.875, highlighting its effectiveness in automatic text summarization. This research underscores the significant potential of LSTM-based models in enhancing extractive text summarization for Indonesian articles. The findings contribute to the field by demonstrating the practical application of advanced machine learning techniques to address the challenge of information overload.