/spines

Tagging with information from shallow/unary constituents

Primary LanguagePython

Tagging with constituent information

Currently we combine unary constituents with the POS tag of the word. And we train an LSTM tagger using Dynet.

Dependencies:

Producing a data set:

python tags.py data/10sentences.sample

This will produce a file data/10sentences.sample.supertags

Training:

python dynet_bilstm_tagger.py --train data/10sentences.sample.supertags --dev data/10sentences.sample.supertags --model model