Accurate and structured postal address data is essential for logistics, geolocation services, and public administration. In countries like Chile, companies such as Correos de Chile and Starken frequently encounter delivery failures due to informal or incomplete address records in their databases. This paper presents a token-based retrieval model for Spanish-language postal addresses, designed to support normalization tasks in georeferenced information systems. The model employs lexical tokenization and a similarity scoring function to identify structured address records that match informal or partial user queries. Using a standardized Chilean address database, we implemented a retrieval pipeline in R and evaluated it through two experiments involving incomplete and reordered input forms. Results show that the model robustly retrieves semantically consistent entries across executions and shows token order invariance. Our proposed approach offers a lightweight, interpretable, and language-specific solution for address normalization, suitable for integration into postal or spatial data systems in resource-constrained or linguistically diverse environments.