Fig. 2
From: Mapping anatomical related entities to human body parts based on wikipedia in discharge summaries
The flowchart of mapping. Anatomical related entities are extracted from medical text first. Then the entities are normalized by using our synonyms dictionary or co-reference chains provided in the corpus of [51]. After that, we match the normalized entities with THBP to see if they are included. If the entities are successfully matched (e.g. lower limb), the results are considered final. If not (e.g. myocarditis), it turns to the external knowledge base to normalize entities and match them with THBP again