A comparison of NER tools w.r.t. a domain-specific vocabulary

Abstract

In this paper we compare several state-of-the-art Linked Data Knowledge Extraction tools, with regard to their ability to recognise entities of a controlled, domain-specific vocabulary. This includes tools that offer APIs as a Service, locally installed platforms as well as an UIMA-based approach as reference. We evaluate under realistic conditions, with natural language source texts from keywording experts of the Städel Museum Frankfurt. The goal is to find first hints which tool approach or strategy is more convincing in case of a domain specific tagging/annotation, towards a working solution that is demanded by GLAMs world-wide.