Skip to main content

Table 26 Parsing accuracy of constituency parsers, evaluated on their generated dependency correspondences

From: A corpus of full-text journal articles is a robust evaluation tool for revealing differences in performance of biomedical natural language processing tools

Parser

Dev Set

Dev – WSJ model

Berkeley Parser

(Micro) LAS

76.97

60.21

(Micro) UAS

88.11

70.66

(Micro) LS

83.13

72.68

(Macro) LAS

80.34

65.19

(Macro) UAS

91.04

75.57

(Macro) LS

84.98

75.63

Stanford Parser

(Micro) LAS

72.13

58.42

(Micro) UAS

83.22

68.83

(Micro) LS

80.12

72.12

(Macro) LAS

75.87

62.10

(Macro) UAS

86.57

71.98

(Macro) LS

82.40

73.85