ConsumerBR: Brazilian Portuguese Consumer Complaint Corpus

ConsumerBR is a large-scale corpus of consumer complaints and company responses in Brazilian Portuguese, collected from the public consumer dispute resolution platform Consumidor.gov.br. The dataset comprises over 3.1 million consumer–company interactions recorded between May 2021 and April 2025, including complaint narratives, company responses, consumer satisfaction ratings, resolution status, temporal information, and location metadata. To ensure compliance with ethical and legal requirements, the dataset includes anonymized versions of all textual fields, where personally identifiable information (PII) has been detected and replaced with standardized placeholders. Raw (non-anonymized) texts are not distributed. The corpus supports a wide range of NLP research tasks, such as complaint classification, sentiment analysis, dialogue modeling, response generation, outcome prediction, and studies on fairness and robustness in institutional communication. Usage is restricted to research and academic purposes. Users must agree not to attempt re-identification, not to redistribute the data for non-research purposes, and to comply with applicable data protection regulations, including the Brazilian General Data Protection Law (LGPD).