Akademik Veri Yönetim Sistemi
7th International Conference on Intelligent and Fuzzy Systems, INFUS 2025, İstanbul, Türkiye, 29 - 31 Temmuz 2025, cilt.1530 LNNS, ss.601-610, (Tam Metin Bildiri)
This paper presents a novel recommendation system designed to leverage the rich, yet underutilized, information contained in user-generated reviews. By integrating advanced natural language processing (NLP) methods and graph database modeling using Apache AGE, the proposed system automatically extracts and rates the diverse experiences offered by different factors often overlooked or insufficiently addressed by conventional recommendation engines. Key stages in the methodology include web scraping to gather large-scale review datasets, text normalization and sentiment analysis for unstructured data, and structured graph modeling to map user preferences to specific hotel attributes and amenities. The sample usage in this study is the hotel search system and general recommendation systems. The key element in these systems is experience and giving results that best match the user based on the selected experiences based on artificial intelligence. A distinctive feature of the proposed approach is the “experience-based” filtering mechanism, which allows the system to pinpoint nuanced attributes (e.g., bed comfort, noise levels, family-friendliness) that collectively form the essence of a user’s desired hotel experience. The graph-centric design fosters more accurate, context-aware recommendations by representing and linking highly interconnected data points—covering sentiments, attributes, and user behavior. Experimental evaluations on a dataset of thousands of real-world hotel reviews demonstrate significant improvements in both accuracy (87%) and query performance (up to 35% faster) compared to traditional recommendation models. The system’s scalability, coupled with its enhanced personalization, indicates strong potential for broad application within and beyond the hospitality domain.