Figure 2. Graphic Representation of Text Input to Searchable Output Vector.

Step 1 of the process is to input unstructured text from a head computed tomography report; step 2, identify clinically relevant injury terms present in the text using the named entity recognition output in long-data format; step 3, identify pertinent injury, negation, uncertainty and property terms from the long data; step 4, produce an output summary of these terms.