Table 2:

Features used by the (a) SpatialIndicator, (b) Trajector, and (c) Landmark methods.

(a) SpatialIndicator features	(b) Trajector features
fI1 The SpatialIndicator candidate’s words. fI2 The part-of-speech of the candidate. fI3 If the next token is also a SpatialIndicator candidate. fI4 The word distance to the nearest Anatomy. fI5 The word distance to the nearest Disorder. fI6 The dependency distance to the nearest Anatomy. fI7 The dependency distance to the nearest Disorder. fI8 The words between the candidate and the nearest Anatomy. fI9 The words between the candidate and the nearest Disorder. fI10 The dependency path between the candidate and the nearest Anatomy. fI11 The dependency path between the candidate and the nearest Disorder. fI12 The types of the UMLS terms on the left and right of the candidate, along with the candidate itself. For example, Example (1) would be  Disorder-in-Anatomy. Example (3) contains both  Disorder-to-Anatomy and Anatomy-of-Anatomy.	fT1 The Trajector candidate’s words. fT2 The word distance between the candidate and its SpatialIndicator. fT3 The dependency distance between the candidate and its SpatialIndicator. fT4 The words between the candidate and its SpatialIndicator. fT5 The dependency path between the candidate and its SpatialIndicator.
	(c) Landmark features
	fL1 The Landmark candidate’s parts-of-speech. fL2 The equivalent of fT2. fL3 The equivalent of fT3. fL4 The equivalent of fT4. fL5 The equivalent of fT5.