Linear-Time Span-Based Neural Constituency Parser

Link to the paper. A DP beam-search extension of Mitchell Stern's span-based neural constituency parser, based on the code from his github.

Requirements and Setup

• Python 3.5 or higher.
• DyNet. We recommend installing DyNet from source with MKL support for significantly faster run time.

Training

A new model can be trained using the command python3 src/main.py train ... with the following arguments:

Argument	Description	Default
--numpy-seed	NumPy random seed	Random
--parser-type	beam-parse or chart	N/A
--tag-embedding-dim	Tag embedding dimension	50
--word-embedding-dim	Word embedding dimension	100
--lstm-layers	Number of bidirectional LSTM layers	2
--lstm-dim	Hidden dimension of each LSTM within each layer	250
--label-hidden-dim	Hidden dimension of label-scoring feedforward network	250
--dropout	Dropout rate for LSTMs	0.4
--model-path-base	Path base to use for saving models	N/A
--train-path	Path to training trees	data/02-21.10way.clean
--dev-path	Path to development trees	data/22.auto.clean
--batch-size	Number of examples per training update	10
--epochs	Number of training epochs	No limit
--checks-per-epoch	Number of development evaluations per epoch	4
--print-vocabs	Print the vocabularies before training	Do not print the vocabularies
--beamsize	Beam size	Infinity
--cross-span	Our new augmented loss	False
--nocubepruning*	Turns off cubepruning	False

*This argument only applies to the beam parser.

Any of the DyNet command line options can also be specified.

The training and development trees are assumed to have predicted part-of-speech tags.

For each development evaluation, the F-score on the development set is computed and compared to the previous best. If the current model is better, the previous model will be deleted and the current model will be saved. The new filename will be derived from the provided model path base and the development F-score.

To train our parser with beamsize 20 using the default hyperparameters, you can use the command:

python3 src/main.py train --parser-type beam-parse --model-path-base models/beam-parse-model --epochs 50 --beamsize 20 --cross-span

Alternatively, to train a chart parser (replicating stern et al. 2017), you can use the command:

python3 src/main.py train --parser-type chart --model-path-base models/chart-model

Evaluation

A saved model can be evaluated on a test corpus using the command python3 src/main.py test ... with the following arguments:

Argument	Description	Default
--model-path-base	Path base of saved model	N/A
--test-path	Path to test trees	data/23.auto.clean
--beamsize	Beam size (for separate decoding)	The trained model's beam size
--nocubepruning*	Turns off cubepruning	False

As above, any of the DyNet command line options can also be specified.

The test trees are assumed to have predicted part-of-speech tags.

As an example, after extracting the pre-trained top-down model, you can evaluate it on the test set using the following command:

python3 src/main.py test --model-path-base models/action-b020-model_dev=92.20