princeton-nlp / MoQA

☕ MoQA: Multi-type Open-domain Question Answering

Check out our paper here!

Setup

You can set up the virual environment using conda:

conda env create -f environment.yml

Data

You can download the pre-processed data from here:

You can also find all the pre-processing scripts under ./preprocess.

Reproducing baselines

DPR

You can use the original DPR repo for training DPR. We provide the prediction files for the DPR-All model.

Extractive Reader

You can train the extractive reader with:

make run-reader

You can change the hyperparameters in the Makefile.

FiD Reader

You can use the original FiD repo for training FiD.

LLMs

You must first set the environment variable OPENAI_API_KEY to your API key.

For the LLM experiments:

make test-gpt3

You can change the model name and other configurations in the Makefile.

Classifier

make run-classifier

Contacts

For any questions, please feel free to reach out to Howard ([e-mail]).