Zusammenfassung:
This article presents a comprehensive approach to probabilistic linkage of German company data using Machine Learning and Natural Language Processing techniques. Here, the long-running ifo Institute surveys are linked to financial information in the Orbis database by addressing the unique challenges of company data linkage, such as corporate structures and linguistic nuances in company names. Compared to a previous linkage, the approach achieves improved match rates and is able to re-evaluate existing matches. This article contributes best practice advice for company data linkage and serves as a documentation for the resulting research dataset.