Medical Entity and Relation Extraction from Narrative Clinical Records in Italian Language

Applying Natural Language Processing techniques enables to unlock precious information contained in free text clinical reports. In this paper, we propose a system able to annotate medical entities in narrative records. Considering that existing NLP systems mainly concern entity recognition in English language, we propose an NLP pipeline to manage clinical free text in Italian. The overall architecture includes a spell checker, sentence detector, word tokenizer, part-of-speech tagger, dictionary lookup annotator, and parsing rules annotator. Essentially, it uses a rule-based approach to extract relevant concepts regarding patient's conditions, administered medications, or performed procedures, detecting their attributes, negated forms, and relations expressions. The indexing of the documents allows the user to retrieve relevant information, increasing his/her medical knowledge.