Polytechnic and design, Vol. 12 No. 4, 2024.
Professional paper
https://doi.org/10.19279/TVZ.PD.2024-12-4-19
BALANCING COST, SPEED AND RELEVANCE IN ARTIFICIAL INTELLIGENCE SYSTEMS FOR TOURISM
Tin Popović
orcid.org/0009-0006-7538-942X
; Bulb Technologies, Ulica Grada Vukovara 23, Zagreb, Croatia
*
* Corresponding author.
Abstract
The tourism sector faces growing demands for real-time personalization and intelligent decision-making, which traditional systems often struggle to meet. This study presents the development of a scalable AI concierge system that combines large language models (LLMs) with Retrieval-Augmented Generation (RAG) architectures to enhance recommendation quality in tourism applications. We evaluate multiple RAG configurations to deliver personalized suggestions for accommodations, attractions, and travel-related queries. The system is designed with modular retrieval components that enable flexible adaptation to user inputs and contextual relevance. Performance is assessed using a composite RCT (Relevance–Cost–Time) index, which captures trade-offs between answer quality, speed, and operational cost. Experimental results highlight the strengths and limitations of each approach, providing practical guidance for designing AI-driven tourism assistants that balance personalization with computational efficiency.
Keywords
AI in tourism; RAG architecture; large language models; personalization; RCT index
Hrčak ID:
335822
URI
Publication date:
2.6.2025.
Visits: 444 *