Parsing clinical text using the state-of-the-art deep learning based parsers: a systematic comparison

Table 2 Performance of deep learning based dependency parsers on the progress notes (%)

Parser	Corpus	Word embeddings	UAS	LS	LAS
Stanford parser	Penn TreeBank	Gigaword	75.76	84.23	71.21
	ProgressNotes	Gigaword	84.01	90.00	80.72
	ProgressNotes	MIMICIII	84.01	90.16	80.66
Bist-parser	Penn TreeBank	Gigaword	75.01	84.73	71.05
	ProgressNotes	Gigaword	82.26	89.69	78.94
	ProgressNotes	MIMICIII	81.78	89.31	78.42
Dependency-tf	Penn TreeBank	Gigaword	78.02
	ProgressNotes	Gigaword	76.72
	ProgressNotes	MIMICIII	77.09
jPTDP-parser	Penn TreeBank	Gigaword	75.51	73.47	60.24
	ProgressNotes	Gigaword	77.59	83.60	71.58
	ProgressNotes	MIMICIII	79.35	85.10	73.61

Highest performance in terms of each evaluation criterion is highlighted in boldface

ISSN: 1472-6947