Fig. 4: Performance difference between Relatives + global + HP and control models on i2b2 test with an increasing number of training samples per abbreviation.
From: Automatically disambiguating medical acronyms with ontology-aware deep learning
The x-axis marks abbreviations with an average less than or equal to X training sentences per expansion. a Number of abbreviations that have an average of less than X training sentences per expansion. b Difference in macro (blue) and micro (red) accuracy for these abbreviations between our best and control models.
