A TensorFlow implementation of Graph-based Biaffine Parser

Table of Contents

• Requirements
• GloVe
• Data Format
• Train a Supertagger
• Structure of the Code
• Jackknife POS Tagging
• Run a pre-trained TAG Parser
• Run demo
• Notes

Requirements

TensorFlow needs to be installed before running the training script. TensorFlow 1.0.0 or higher is supported. (TensorFlow 1.3.0 and 1.7.0 are tested)

GloVe

Our architecture utilizes pre-trained word embedding vectors, GloveVectors. Run the following:

wget http://nlp.stanford.edu/data/glove.6B.zip 

and save it to a sub-directory glovevector/.

Data Format

The biaffine parser takes as input a file in the Conllu+Supertag (conllustag) format, in which one column for supertags is added to the original conllu format at the end. See a sample.

Train a Parser

All you need to do is to create a new directory for your data in the conllustag format and a json file for the model configuration and data information. We provide a sample json file for the sample data directory. You can train a parser on the sample data by the following command:

python train_graph_parser.py sample_data/config_demo.json

After running this command, you should be getting the following files and directories in sample_data/:

Directory/File	Description
checkpoint.txt	Contains information about the best model.
sents/	Contains the words in the one-sentence-per-line format
gold_pos/	Contains the gold POS tags in the one-sentence-per-line format
gold_stag/	Contains the gold supertags in the one-sentence-per-line format
arcs/	Contains the gold arcs in the one-sentence-per-line format
rels/	Contains the gold rels in the one-sentence-per-line format
predicted_arcs/	Contains the predicted arcs in the one-sentence-per-line format
predicted_rels/	Contains the gold rels in the one-sentence-per-line format
Parsing_Models/	Stores the best model.
conllu/sample.conllustag_stag	Contains the predicted supertags in the conllustag format

Structure of the Code

File	Description
utils/preprocessing.py	Contains tools for preprocessing. Mainly for tokenizing and indexing words/tags. Gets imported to utils/data_process_secsplit.py
utils/data_process_secsplit.py	Reads training and test data and tokenize/index words, POS tags, stags, and characters.
utils/parsing_model.py	Contains the Parsing_Model class that constructs our LSTM computation graph. The class has the necessary methods for training and testing. Gets imported to bilstm_stagger_model.py. For more details, read README for utils.
utils/lstm.py	Contains tensorflow LSTM equations. Gets imported to utils/stagging_model.py.
graph_parser_model.py	Contains functions that instantiate the Parsing_Model class and train/test a model. Gets imported to graph_parser_main.py
graph_parser_main.py	Main file to run experiments. Reads model and data options.
scripts/train_graph_parser.py	Runs graph_parser_main.py in bash according to the json file that gets passed.

Run a pre-trained TAG Parser

We provide a pre-trained TAG parser. To run the pretrained parser on your data, first download the model and place the Pretrained_Parser directory in the demo directory. Then, run

• With tokenization

python demo/scripts/demo_model.py --infile demo/sents/test.txt --tokenize

• Without tokenization

python demo/scripts/demo_model.py --infile demo/sents/test.tokenized.txt

You can replace these files by your own data. It prints out predicted elementary trees, parses, and PTB-style fine-grained POS tags in the conllu format. Note that we put elementary trees in the UPOS column.

Run a demo TAG Parser

Enjoy playing around with our TAG demo parser online!

Notes

If you use this tool for your research, please consider citing:

@InProceedings{Kasai&al.18,
  author =  {Jungo Kasai and Robert Frank and Pauli Xu and William Merrill and Owen Rambow},
  title =   {End-to-end Graph-based TAG Parsing with Neural Networks},
  year =    {2018},  
  booktitle =   {Proceedings of NAACL},  
  publisher =   {Association for Computational Linguistics},
}