Penerapan Model BERT pada Chatbot dalam Platform E-Commerce

Abstract

This study develops chatshop, a BERT-based e-commerce chatbot designed to assist customers in searching for pizza menus. The BERT model was chosen because of its ability to understand sentence context bidirectionally, thereby increasing the accuracy in detecting user intent. This chatshop allows users to find menus, get recommendations, and access price and stock availability information in real time. The evaluation was carried out using BLEU and ROUGE-L metrics, with a ROUGE-L F1 score of 32.91%. These results indicate that the chatbot is able to handle simple interactions well, but still needs improvement in answering more complex questions accurately and completely.