DIP-BR: Dataset of Intellectual Property in Brazil

Technology transfer between research institutions (ICTs) and companies is a crucial yet challenging aspect of the innovation ecosystem. While patent analysis is a strategic tool for understanding these collaborations, its application in Brazil is constrained by data quality. A previously developed dataset, containing 256,909 patents from the National Institute of Industrial Property (INPI) since 2018, possesses a limitation for this analysis: patent holder names are stored as inconsistent plain text, preventing an accurate mapping of relationships and collaborations. To overcome this limitation, we introduce DIP-BR, an open research Dataset of Intellectual Property in Brazil, developed through a comprehensive enrichment pipeline. Specifically, we apply a deduplication algorithm to standardize names and integrate machine learning techniques to automatically classify each patent holder as a research institution, company, or individual. The main contribution of this work is the resulting enriched dataset, which is further structured using network modeling and clustering. We demonstrate the dataset's value by identifying significant collaboration patterns between ICTs and companies in key domains like Human Necessities, Chemistry, and Electricity. This enriched resource serves as a foundational tool for analyzing trends and visualizing the dynamics of innovation in Brazil.