Corruption is a phenomenon in most countries worldwide, creating obstacles to social, political, and economic development. This research aims to utilize social media platforms like Twitter to gather data on crimes in Peru against public administration and, through an annotation guide and a data-driven annotation process, develop a reliable corpus for classifying crimes into 12 categories. Additionally, this study seeks to increase the availability of Spanish-language datasets, facilitating the identification and classification of corruption-related topics in Peru. The dataset construction process involved extracting text from Twitter using tools such as Snscrape, Tweepy, and manual extraction, resulting in 3,692 records. The annotation method employed included designing an annotation process and creating an annotation guide. Three iterations achieved a final Cohen’s Kappa coefficient of 0.922, indicating an almost perfect agreement. The most frequent crime types identified were Influence Peddling, Collusion, Illicit Enrichment, and Bribery.