Erna Daniati, Aji Prasetya Wibawa, Wahyu Sakti Gunawan Irianto
Event extraction from narrative texts is an essential task in natural language processing (NLP) that enables the identification and categorization of significant events within a story. This study focuses on event extraction in Andersen's fairy tales, utilizing a zero-shot approach with advanced deep learning models: BERT (Bidirectional Encoder Representations from Transformers) and Bi-LSTM (Bidirectional Long ShortTerm Memory). Traditional event extraction methods often require a large amount of labeled data for training, which can be resource-intensive and time-consuming. In contrast, the zeroshot learning approach allows the model to identify events without the need for task-specific training data, making it more flexible and applicable across various domains. This research leverages the pre-Trained BERT model to capture contextual information and the Bi-LSTM model to enhance sequence learning, effectively identifying events in the text. The experiment demonstrates the effectiveness of this combined model on Andersen's fairy tales, achieving promising results in extracting narrative events such as actions, character interactions, and key plot moments. These findings suggest that zero-shot event extraction using BERT and Bi-LSTM can serve as a robust approach for automated event detection in literary and historical texts, with potential applications in digital humanities, storytelling analysis, and content summarization. © 2025 IEEE.
Universitas Negeri Malang, Department of Electrical Engineering and Informatic, Malang, Indonesia
