Constituency Grammar Induction On Ptb

Metrics

Max F1 (WSJ)

Results

Performance results of various models on this benchmark

Model Name	Max F1 (WSJ)	Paper Title	Repository
PRPN	38.1	Neural Language Modeling by Jointly Learning Syntax and Lexicon	-
Ensemble (Selective MBR)	-	Ensemble Distillation for Unsupervised Constituency Parsing	-
ReCAT	-	Augmenting Transformers with Recursively Composed Multi-grained Representations	-
ON-LSTM	49.4	Ordered Neurons: Integrating Tree Structures into Recurrent Neural Networks	-
DIORA (+PP)	56.2	Unsupervised Latent Tree Induction with Deep Inside-Outside Recursive Auto-Encoders
Parse-Focused (NT=30)	68.4	Structural Optimization Ambiguity and Simplicity Bias in Unsupervised Neural Grammar Induction	-
Ensemble (Generative MBR)	71.9	Ensemble Distillation for Unsupervised Constituency Parsing	-
inside-outside co-training + weak supervision	66.8	Co-training an Unsupervised Constituency Parser with Weak Supervision	-
Compound PCFG	60.1	Compound Probabilistic Context-Free Grammars for Grammar Induction	-
TN-PCFG (p=500)	61.4	PCFGs Can Do Better: Inducing Probabilistic Context-Free Grammars with Many Symbols	-
ON-LSTM (tuned)	50.0	Ordered Neurons: Integrating Tree Structures into Recurrent Neural Networks	-
DIORA	49.6	Unsupervised Latent Tree Induction with Deep Inside-Outside Recursive Auto-Encoders
URNNG	52.4	Unsupervised Recurrent Neural Network Grammars	-
GPST(left to right parsing)	-	Generative Pretrained Structured Transformers: Unsupervised Syntactic Language Models at Scale	-
Neural PCFG	52.6	Compound Probabilistic Context-Free Grammars for Grammar Induction	-
Parse-Focused (NT=4500)	70.3	Structural Optimization Ambiguity and Simplicity Bias in Unsupervised Neural Grammar Induction	-
S-DIORA	63.96	Unsupervised Parsing with S-DIORA: Single Tree Encoding for Deep Inside-Outside Recursive Autoencoders	-
DP in rank space	-	Dynamic Programming in Rank Space: Scaling Structured Inference with Low-Rank HMMs and PCFGs	-
Hashing (Parserker 2)	64.1	On Eliciting Syntax from Language Models via Hashing	-
PRPN (tuned)	47.9	Neural Language Modeling by Jointly Learning Syntax and Lexicon	-

0 of 23 row(s) selected.