/joint_stackptr

Joint POS Tagging with Dependency parsing with StackPtr Model

Primary LanguagePython

Stack Pointer Network for CoNLL2018 Shared Task

This is the code for our paper: Joint Learning of POS and Dependencies for Multilingual Universal Dependency Parsing

The code is based on NeuroNLP2, we are very appreciate the contribution of Xuezhe Ma.

Requirements

Python 2.7, PyTorch >=0.3.0, Gensim >= 0.12.0

Running the experiments

Universal Dependency Parsing

To train a Joint-POS-Stack-Pointer parser, simply run

./example/run_stackPtrParser_CoNLLU.sh

Remeber to setup the paths for data and embeddings.

You can train multiple model in different language by

./train.py

and evaluate multiple model by

./run.py