Enhancing multilingual and biomedical named entity recognition using Wikidata semantic relations

Named Entity Recognition (NER) is currently a key technique is knowledge engineering, particularly in the context of biomedical informatics. In this context, Wikidata has been used as a Knowledge Graph to drive named entity recognition for various applications such as news tracking and question answering. Current Wikidata NER Systems are mainly based on the labels, aliases and classes of Wikidata items. In this research paper, we propose a new approach to augment and validate the named entity recognition of a type of entities based on Wikidata semantic relations. We evaluate our semantic relation-based multilingual named entity recognition algorithm by applying it on a corpus of 8,705 titles of biomedical research publications about drugs extracted from the Wikidata knowledge graph.