Design And Development of an NLP-Based Chatbot For Information Services

Abstract

Pondok Pesantren Modern AL-Ikhlas Putri faces difficulty in providing information services that are accessible at any time, since existing information channels rely on staff who can only be reached during limited working hours, making it hard for students, teachers, prospective students' parents, and the general public to obtain quick answers outside those hours. This study aims to design and build an NLP-based information service chatbot that functions as a round-the-clock virtual assistant for answering questions about the pesantren's activities, curriculum, facilities, and general information. The system was developed using a three-tier architecture consisting of a user interface (React.js), a Backend-for-Frontend server (Node.js), and an AI/ML service (Flask). A Natural Language Processing approach was applied by leveraging the pre-trained IndoBERT model to transform user questions into vector embeddings, and Cosine Similarity was then used to measure the semantic similarity between a user's query and an FAQ-formatted knowledge base. Black box testing showed that the chatbot was able to return accurate and relevant answers at any time, including for varied sentence structures and questions containing typographical errors, confirming that the semantic retrieval mechanism functioned effectively. The resulting chatbot provides an innovative digital solution that makes information about the pesantren accessible around the clock, supporting digital transformation in pesantren-based education.