Representasi Pengetahuan Ontologi untuk Klasifikasi Topik Penelitian pada Bidang Ilmu Informatika

Abstract

Research in informatics often involves multiple subdisciplines, making topic classification challenging. Typically, text classification in natural language processing requires training and testing. However, ontology-based classification eliminates the need for training data. Challenges in ontology-based classification include finding terms that lack similarity with ontology and ensuring accuracy in measuring data similarity with knowledge representation. To address this, the fastText method identifies term similarities between words and ontology, while the Wu-Palmer method measures semantic similarity and relationships within ontology. The research process includes preprocessing (Casefolding, Tokenizing, Stopword Removal, Lemmatization), Query Processing (Query Reduction, Duplicate Removal), Word Embedding with fastText, and Semantic Similarity measurement using Wu-Palmer. The dataset consists of 200 research studies from Fasilkom Unsri informatics students' final projects. The classification results show that 178 out of 200 topics were correctly classified, achieving an accuracy of 89.5%, demonstrating the system’s effectiveness.