Table 2 The reduction in the UMLS gold standard when Yetisgen-Yildiz and Pratt semantic type filtering is used
From: Is automatic detection of hidden knowledge an anomaly?
Train - Test - Eval | Orig |GS| | Y-P |GS| |
|---|---|---|
2006 - 2010 - 2013 | 10,237 | 275 |
2007 - 2011 - 2014 | 8,851 | 486 |
2008 - 2012 - 2015 | 5,476 | 739 |
2009 - 2013 - 2016 | 9,040 | 235 |
2010 - 2014 - 2017 | 24,772 | 649 |