DEVELOPMENT OF INDONESIAN CONSTITUTION QUESTION ANSWERING SYSTEM

The Constitution is the highest guideline for running the wheels of the state and government. In practice, society may not fully understand the constitution because of a lack of education and access to information. There-fore, in the case of Indonesia, it is important to increase the constitutional awareness of Indonesian citizens to effectively guard the practice of govern-ment. However, understanding and materials related to the constitution are usually only available in schools and universities. Thus, this research devel-ops a question-answering system for a more accessible understanding of the Indonesian Constitution by using TF-IDF, Cosine Similarity, and Jaccard Similarity. Legal experts collected the dataset from Indonesian law. A user can deliver a question through a web application and an instant messaging bot (Telegram). It uses the Mean Reciprocal Rank to assess the QA system. TFIDF+Cosine similarity performance is slightly better than Jaccard simi-larity. Human evaluation generates worse performance than self-evaluation because users can ask different words for the same meaning question. The application is expected to accelerate the constitutional understanding of society because it is easier and more accessible for society.