Ontology-Based Topic Model for Document Retrieval Systems in Information Technology

Abstract

Most current academic document retrieval systems for topic-based search rely on simple keyword matching or statistical topic modeling. In these methods, topics are formed either from sets of frequent keywords or from statistical clusters. While these approaches work in some contexts, they cannot fully capture the rich semantic meaning of topics as understood by human experts. This often leads to search results that fail to match the intended meaning of the topic, causing gaps between what users need and what the system returns. This study aims to overcome these limitations by developing a topic-based retrieval system that represents topics in a more semantically rich and human-aligned way. The system is designed to help Information Technology (IT) students search for topic-relevant materials—specifically English-language ebooks and research papers—from a curated faculty repository. We introduce C-ONTO, a structured knowledge model that includes topic names, descriptions, learning objectives, real-world application scenarios, and concept graphs describing internal semantic relationships. Documents are also modeled as concept graphs, enabling accurate semantic similarity calculations. An intelligent query analysis module interprets user intent and maps it to the system’s semantic structure. Testing with 425 real student queries in four IT domains shows that the system achieves 81.18% accuracy, outperforming keyword-based and statistical methods in precision, recall, coverage, and F1-Score. By aligning computational topic modeling with human understanding, the proposed system improves accuracy, semantic consistency, and educational relevance in academic document retrieval for information technology and related fields.