MUSER

How to set up

• Python3.7
• Install all packages in requirement.txt.

pip3 install -r requirements.txt

Index a list of documents:

python multi_step_retriever.py --index example_docs.jsonl

Train NLI models

• Run train.py and specify what Transformer model you would like to fine tune:

python train.py --bert_type bert-large --check_point 1

Option "--check_point 1" means that we will use the checkpoint technique during training.

The trained model (that has the best performance on the dev set) will be saved to directory output/.

Test the performance of the trained models

• To test the performance of a trained model, run the command below:

python test_trained_model.py --bert_type bert-large

1. Download the model weights and extract them into the output/nli_model folder:

• NLI model

Main experiment setup parameters

	PolitiFact	Gossipcop	Weibo
Sequence_length	512	512	512
Max_encoder_length	512	512	512
Min_decoder_length	64	64	64
Max_decoder_length	128	128	128
Embedding_dimension	200	200	200
k(number of paragraphs retrieved)	30	30	30
MSR	0.3	0.3	0.3
$lambda$	0.9	0.9	0.9
Retrieve_steps	2	3	3
Batch_size	64	64	32
Maximum_epochs	10	10	10
Vocabulary_size	30522	30522	21128
Learning_rate	1e-5	1e-5	1e-5