Jusletter IT

This paper explores the merits of a pattern- and rule-based approach for the automated extraction of punishments from Dutch criminal cases in courts of first instance. Automated extraction of case outcomes leverages the increasing amount of information becoming available through digital technologies and aids the creation of big data sets for work in legal informatics and AI & Law. This work addresses domain-specific challenges, in particular that Dutch criminal case decisions may impose a single combined sentence for multiple facts or impose multiple sentences in the same decision. Manual evaluation of the developed method shows that the use of interpretable methods is a viable approach in the legal domain.