open book question answering

OBQA Link to Dataset

1. Parser and Tagger are taken from https://github.com/tdozat/Parser-v3.

2. Parser and Tagger are trained using CoNLL 2018 dataset :

   • Git clone the repo. Create the Data Directory : data/CoNLL18/UD_English-EWT. Save both the Datasets and the embeddings.
   • Training Data : http://universaldependencies.org/conll18/
   • Word2Vec Embeddings : https://lindat.mff.cuni.cz/repository/xmlui/handle/11234/1-1989 . Only English Word2Vec embeddings are needed.
   • Environment : TensorFlow=1.4, Scipy, Matplotlib, Psutil, Python=3.6, Pandas, Conllu
   • Training Command : python main.py train ParserNetwork / TaggerNetwork
   • Run Model Command : python main.py --save_dir=\$PATH_TO_NETWORK run \$INPUTFILE --output_dir=\$OUTPUTDIR
   • Key Point to Note: CoNLLU format needs to be adhered strictly, Tabs between columns.
   • Trained Models to be pushed at a location : [DropboxLocation]

1. Present in folder ir, with its own ReadMe

1. Present in notebooks and folders

1. Present in filterir

1. Runner and scorer are present