Skip to main content
. 2015 Dec 16;10(12):e0143274. doi: 10.1371/journal.pone.0143274

Fig 2. The hybrid automated-manual text mining pipeline.

Fig 2

(A) The three most powerful automated biomedical-tagging engines, BioIE, Whatizit and MEDIE have specific limitations. BioIE only tags relationships between biomedical entities, and Whatizit only tags bio-entities. MEDIE tags both bio-entities and their relationships. In the given example, all three search engines failed to tag the bio-entities “ROS” and “A(2)R” (circled), which are obvious to a human reader. Red circles denote terms that the automated text mining algorithms failed to recognise. (B) The hybrid data processing pipeline combines automated text mining (BioIE, Whatizit and MEDIE) and manual text collection. Bio-entities are annotated with BioIE, Whatizit, MEDIE and PubTator.