Jointly Predicting Predicates and Arguments in Neural Semantic Role Labeling

ACL 2018  ·  Luheng He, Kenton Lee, Omer Levy, Luke Zettlemoyer ·

Recent BIO-tagging-based neural semantic role labeling models are very high performing, but assume gold predicates as part of the input and cannot incorporate span-level features. We propose an end-to-end approach for jointly predicting all predicates, arguments spans, and the relations between them. The model makes independent decisions about what relationship, if any, holds between every possible word-span pair, and learns contextualized span representations that provide rich, shared input features for each decision. Experiments demonstrate that this approach sets a new state of the art on PropBank SRL without gold predicates.

PDF Abstract ACL 2018 PDF ACL 2018 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Result Benchmark
Semantic Role Labeling (predicted predicates) CoNLL 2005 He et al. 2018 + ELMo F1 86.0 # 2
Semantic Role Labeling CoNLL 2005 He et al. (2018) F1 82.5 # 15
Semantic Role Labeling CoNLL 2005 He et al. (2018) + ELMo F1 86.0 # 13
Semantic Role Labeling (predicted predicates) CoNLL 2005 He et al. (2018) F1 86.0 # 2
Semantic Role Labeling (predicted predicates) CoNLL 2005 He et al. 2018 F1 82.5 # 5
Semantic Role Labeling (predicted predicates) CoNLL 2012 He et al. 2018 + ELMo F1 82.9 # 4
Semantic Role Labeling (predicted predicates) CoNLL 2012 He et al. 2018 F1 79.8 # 7
Semantic Role Labeling OntoNotes He et al., F1 85.5 # 12
Semantic Role Labeling OntoNotes He et al. F1 82.1 # 16

Methods


No methods listed for this paper. Add relevant methods here