Development of Polibatam P2M Interactive Geovisualization Platform Using AI-Assisted ETL Pipeline and Structured Onboarding Systems

Abstract

Politeknik Negeri Batam's Pusat Penelitian dan Pengabdian Kepada Masyarakat (P2M) manages community service programs across the Riau Islands, yet program documentation remains confined to text-based listings on static web pages without geographic context. This study integrates AI-assisted data extraction, geospatial visualization, and usability-oriented interface design into a unified platform to address both data scalability and user accessibility challenges. An interactive geovisualization platform was developed to transform unstructured P2M journal articles into an explorable, map-based multimedia experience. An AI-assisted Extract–Transform–Load (ETL) pipeline using GPT-3.5-turbo was implemented to automatically extract, categorize, and geocode 48 community service programs from 49 PDF documents published between 2019–2024, scaling from an initial 5 manually curated entries. The platform was built using Next.js and Leaflet.js with a mobile-first responsive design, featuring marker clustering, category-based filtering, interactive lightbox, and a structured two-tier onboarding system grounded in Cognitive Load Theory. Usability evaluation using Task Completion Rate (TCR) with 5 participants and System Usability Scale (SUS) with 23 respondents yielded a TCR of 95.0% and a SUS score of 67.5, indicating highly effective task completion and acceptable usability at the industry benchmark average. The results demonstrate the feasibility of combining AI-driven data extraction with interactive web mapping to enhance institutional knowledge discovery and geographic awareness of community service impact.