DeepDeepParser

Code and data preparation scripts for the paper Robust Incremental Neural Semantic Graph Parsing, Jan Buys and Phil Blunsom, ACL 2017.

Prerequisites

See Dependencies.md

Data preparation

To extract DMRS and EDS graphs from DeepBank (requires the LOGON environment and full original data):

scripts/extract-deepbank.sh

To extract DMRS and EDS graphs from the SDP release of DeepBank (does not require the LOGON environment):

scripts/extract-deepbank-sdp.sh 

Pre-processing (constructs lexicon, runs Stanford CoreNLP, constructs graph linearizations/oracle transition sequences):

scripts/preprocess.sh

Training

Train the transition-based parser:

python rnn/parser.py --decode_dev --decode_train --use_hard_attention_arc_eager_decoder --predict_span_end_pointers  --data_dir [data_dir] --train_dir [working_dir] --embedding_vectors [embedding_file] --train_name train --dev_name dev --singleton_keep_prob 0.5 --size 256 --input_embedding_size 256 --output_embedding_size 128 --tag_embedding_size 32 --use_encoder_tags --input_drop_prob 0.3 --output_drop_prob 0.3 --initialize_word_vectors 

where data_dir contains the pre-processed files for training.

Word embeddings are initialized with pre-trained structured skip-gram embeddings: sskip.100.vectors

Decoding

A pre-trained EDS model is available here

Decode with the parser (transition-based model):

python rnn/parser.py --decode --decode_dev --use_hard_attention_arc_eager_decoder --predict_span_end_pointers --data_dir [data_dir] --train_dir [working_dir] --dev_name [filename] --size 256 --input_embedding_size 256 --output_embedding_size 128 --tag_embedding_size 32 --use_encoder_tags --input_drop_prob 0.3 --output_drop_prob 0.3 --checkpoint_file model.ckpt

where data_dir contains the pre-processed files for decoding (filename.en, filename.ne, filename.pos) as well as a buckets file, and working_dir contains the model (checkpoint) file.

Suggested buckets:

24 77
37 133
52 201

Post-processing

Restore lemmas and constants and convert to output graph formats.

python mrs/linear_to_mrs.py [data_dir] [filename] [working_dir] output -arceagerbuffershift -unlex -withendspan