Event-Detection-and-Domain-Adaptation-with-Convolutional-Neural-Networks

code re-implement model in paper Event Detection and Domain Adaptation with Convolutional Neural Networks

Preprocessing data:

• Transform format data from ACE format to json format from here: https://github.com/nlpcl-lab/ace2005-preprocessing.git
• build vocabularies : re-build event vocab(34-classes), entity vocab( BIO format include pad label, id=0), word vocab( include pad token with id=0, unknow token with id =1) or just use vocabs from data folder
• use load_data_json and window_encode functions in utils.py to build data that will be fed into model

Train model:

• hyperparameters of model are stored in Config class
• just run file runs.py

References:

Detection and Domain Adaptation with Convolutional Neural Networks, Thien Huu Nguyen, Ralph Grishman