Indonesian Abstractive Text Summarization with Bidirectional Long Short-Term Memory (Bi-LSTM)
Dian Sa’adillah Maylawati, Karima Marwazia Shalih, Muhammad Ali Ramdhani, Cepy Slamet, Diena Rauda Ramdania · 2024
Every language is unique, including Indonesian text. In the development of text summarization, research is mostly oriented to extractive methods, which have an unreadable sentence structure. The abstractive summarization has a more natural summary, but the process is more challenging than the extractive method. Therefore, this research aims to produce an Indonesian abstractive text summary using Bidirectional Long Short-Term Memory (Bi-LSTM), which can produce abstractive summarization for English texts well. So, in this study, Bidirectional LSTM is used to improve the quality of the Indonesian text summary. Also, local attention is added to assist the model in predicting the summary. The results of evaluating the model on the Indonesian language article text show that the model can capture the essence of the summary text. The ROUGE evaluation shows that Bi-LSTM with local attention can produce an Indonesian abstractive summary well.