<![CDATA[Research title duplication is a crucial problem that frequently occurs in the academic environment, particularly in the governance of the Institute for Research and Community Service (LPPM). The current manual title verification process is considered inefficient, time-consuming, and highly prone to human error, potentially allowing proposals with substantial similarity to pass. This study aims to design and develop a web-based LPPM Information System integrated with a smart automatic title similarity detection feature for a case study at Universitas Muhammadiyah Karanganyar. The proposed solution is the development of a distributed system using a microservices architecture, integrating a Laravel framework web service and a Python (FastAPI) Natural Language Processing (NLP) computational service via REST API communication protocol. The software development method adopts Agile Development to accommodate requirement changes iteratively. The primary contribution of this research is the creation of an academic validation instrument based on a distributed architecture that isolates the text computational workload from the main server. The application of the Cosine Similarity algorithm combined with Term Frequency-Inverse Document Frequency (TF-IDF) weighting and the Sastrawi lexical library is used to calculate text similarity precisely. Black Box functional testing using the Equivalence Partitioning technique shows that the API integration runs without errors. Given the research focus on system integration, the algorithm evaluation was conducted using a controlled synthetic dataset simulating various text manipulation techniques. The test results show that the system can precisely identify basic syntactic modifications. However, the system transparently acknowledges a computational limitation (a drop in detection to 12.74%) in recognizing semantic similarities purely based on synonyms. The implementation of this technology is expected to help LPPM administrators validate the originality of proposals efficiently, objectively, and in real-time.]]>
