Implementation of Transition-based Model for Chinese Information Extraction

We provide sample sentences from ACE05 for illustration purposes.

To reproduce our study, replace the 'test_sample.txt' file with full dataset.

Please refer to https://catalog.ldc.upenn.edu/ldc2006t06 for the full dataset acquirement.

Requirement: PyTorch (tested on 0.4), Python(v3)

Setup

run train.py to start training

Configuration

Configurations of the model and training are in config.py

Some important configurations:

  • if_gpu: whether to use GPU for training
  • input_dropout: whether to add dropout layer
  • if_pretrained: whether to using Glove pretrained embeddings