AN ARTIFICIAL INTELLIGENCE-BASED INFORMATION SYSTEM FOR RESEARCH DUPLICATION DETECTION USING DEEP LEARNING AND DATA CLEANSING TECHNIQUES

Abstract

This research focused on the design and development of an information system to detect duplication in research proposal submissions. The system was developed by applying a pre-built model and integrating advanced techniques, specifically Data Cleansing, Text Embedding, and Cosine Similarity. The research objectives were to: 1) develop an information system for detecting research duplication, and 2) evaluate the system's performance. The testing process utilized a dataset of 10 research projects, with the collected data analyzed using semantic similarity criteria, expressed as percentages, and content analysis.

The results indicate that the web-based information system effectively meets user requirements. The implementation of Data Cleansing significantly reduced data noise and enhanced processing accuracy. Furthermore, the application of Text Embedding and Cosine Similarity enabled the system to detect duplication based on contextual and semantic meaning with greater precision than traditional keyword matching methods. Performance evaluations demonstrated that the system could accurately analyze and rank research topic duplication in accordance with established standards. Consequently, the system is suitable for organizational implementation, serving as an effective tool for systematic research data management and supporting administrative decision-making for new research funding allocations, thereby optimizing the management of institutional budgets and research resources.