toxic_comment_classification

Identify toxicity in online comments.

Dataset

Data for this project has been picked from Kaggle.

• Download data from kaggle

• Unzip the files and save .csv in 'data/' folder

Requirements

• Run following command to load required libraries from requirements.txt file:

pip install -r requirements.txt

• Download pretrained GLoVe embeddings (glove.840B.300d) from here or here and save to 'data/' folder.

• Ensure file names specified in config.yaml is consistent with your training and embedding file names

Run script

• Choose preferable settings from config.yaml before initiating traning:

  • load_pretrained_embeddings_from_disk has been defaulted to False, change to True if you want to avoid unpacking glove embeddings for each subsequent run

  • Update random_seed to maintain reproducibility of multiple experiments

  • run main.py