BERT For Classifying Intent in Ulos Fabric and Batak Culture Learning Chatbot
Syarifah Atika, Maya Silvi Lydia, Muhammad Anggia Muchtar · 2025
Indonesia is rich in cultural heritage, including its woven fabric known as "tenun." Each province has a unique tenun style, but North Sumatra is renowned for its Tenun Ulos. Ulos, a traditional cloth woven by Batak women, features patterns and colors with specific meanings and plays a vital role in Batak ceremonies and traditions. Preservation of Ulos fabric is important, typically achieved through interviews with cultural figures and literary sources. However, fragmented information, inconsistencies, and language barriers hinder effective knowledge dissemination. To address these issues, leveraging artificial intelligence (AI) technology, specifically chatbots presents a promising solution. A chatbot simulates human communication by providing quick, relevant responses through text or voice. The ability to distinguish between names and types of Ulos is essential to reducing confusion about Ulos terminology. For instance, Ulos Pargomgom, worn by elders, varies by occasion, using Ulos Ragidup for weddings and Ulos Suri-Suri Ganjang for naming ceremonies. Identifying users' specific purposes is critical in question-answering and dialogue systems. BERT (Bidirectional Encoder Representations from Transformers), a deep learning-based language model, excels in providing universal text representation, making it valuable for automated natural language processing. This research evaluates BERT for intent classification by training it on an Ulos and Batak Culture dataset to create an intent classification model. The model achieved an accuracy of 0.83 and an F1 score of 0.83 during training. End-to-end evaluation tested the chatbot's performance, resulting in an accuracy of 85% and an F1 score of 84%. These results indicate that while the model performs reasonably well, there is room for improvement.