Selection of chatbot architectures based on LLMs and Augmented Information Retrieval (AIR) for detecting problematic import items using the PROMETHEE multicriteria method

• Grant number: 25/02051-9
• Support Opportunities: Scholarships in Brazil - Master
• Start date: April 01, 2025
• End date: March 31, 2027
• Field of knowledge: Engineering - Production Engineering - Operational Research
• Principal Investigator: Cristiano Morini
• Grantee: María Alejandra Asmat Reyes
• Host Institution: Faculdade de Ciências Aplicadas (FCA). Universidade Estadual de Campinas (UNICAMP). Limeira , SP, Brazil
• Associated research grant: 23/09754-0 - Customs compliance: policy design, actions, program evaluation, legislation review, initiatives to encourage compliance with regulations and good customs practices for imports into Brazil, AP.PP

Abstract

The project proposes the analysis and comparison of chatbot architectures based on Large Language Models (LLMs) and Augmented Information Retrieval (AIR) to detect potential problems in import items. The approach will use the PROMETHEE multicriteria method to identify the best architecture according to previously defined criteria, which include response accuracy, speed, costs and implementation complexity. This study is particularly important to improve automation in the screening and inspection of imported goods, helping agencies such as the Brazilian Federal Revenue Service to optimize processes and reduce errors in the customs context.