MisRoBÆRTa: Transformers versus Misinformation

Article:

Ciprian-Octavian Truică and Elena-Simona Apostol. MisRoBÆRTa: Transformers versus Misinformation. Mathematics, 10(4):1-25(569), ISSN 2227-7390, February 2022 DOI: 10.3390/math10040569

Deployment:

Python v3.7.x

Packages:

• numpy
• pandas
• matplotlib
• scikit-learn
• simpletransformers
• sentence-transormers
• keras
• tensorflow
• pytorch
• bert-for-tf2
• tensorflow-hub
• transformers

To run MisRoBÆRTa use "FakeBERT.py" script as follows:

python MisRoBÆRTa.py FILE_NAME USE_CUDA NO_TESTS

Where:

• FILE_NAME - is the a csv file with 2 columns: content and label
• USE_CUDA - 0 - False, 1 - True
• NO_TESTS - how many test to perform

To run the transformers classificatin use "transfs_misinformation.py" script as follows:

python -u transfs_misinformation.py FILE_NAME MODEL_TYPE NO_GPU NO_TESTS

Where:

• FILE_NAME - is the a csv file with 2 columns: content and label
• MODEL_TYPE - the transformer model name given with lowercase, e.g., bart
• NO_GPU - the GPU to run the code on
• NO_TESTS - how many test to perform

To run the BART models either replace or copy the marked code in the "classification_model.py" and "bart_model.py" as follows:

For "classification_model.py" there are new lines added to this file

Place it here to overwrite:

$PYTHON_HOME/lib/python3.7/site-packages/simpletransformers/classification/

We higly recommand to add the lines marked with " # line for BART " in the existing "classification_model.py" file, and not overwrite the file.

For "bart_model.py" place this file here:

$PYTHON_HOME/lib/python3.7/site-packages/simpletransformers/classification/transformer_models/

To run FakeBERT use "FakeBERT.py" script as follows:

python FakeBERT.py FILE_NAME NO_TESTS

Where:

• FILE_NAME - is the a csv file with 2 columns: content and label
• NO_TESTS - how many test to perform