Automatisierte Identifikation und Lemmatisierung historischer Berufsbezeichnungen in deutschsprachigen Datenbeständen

Occupational information occurs in many historical sources. For a large number of research areas, not only standardization, but above all classification of these is a central prerequisite for analysis. In this article, the assignment of spelling variants to already defined generic names of occupations is referred to as lemmatization or normalisation, while the assignment of the normalised spelling and to a classification system is referred to as classification. In order to reduce manual effort, an algorithm for the automated lemmatization of historical, German-language occupational data is developed. The best result is achieved with a supervised machine learning approach. Overall, about 72 percent of the occupational data can be lemmatized, and about 98 percent of these assignments are correct.