D2_run.sh

#!/bin/sh

source ~/anaconda3/etc/profile.d/conda.sh
conda activate ./env

# Preprocess tweets
python3 src/preprocess_olid.py \
    --file data/olid-training-v1.0.tsv \
    --train_ids data/train_ids.txt \
    --val_ids data/val_ids.txt \
    --split_punctuation True \
    --remove_apostraphes True \
    --remove_hashtags True
    
# Create feature embeddings for train/val tweets
python3 src/featurize_tweets.py \
    --preprocessed_data data/pp_olid-training-v1.0.tsv \
    --train_data data/clean_train_english.tsv \
    --val_data data/clean_val_english.tsv \
    --embedding_path data/glove.twitter.27B.200d.w2vformat.txt \
    --emoji_embedding_path data/emoji2vec_200d.txt \
    --embedding_size 200 \
    --train_vectors data/train_vectors.npy \
    --val_vectors data/val_vectors.npy

# Binary classification of train/val tweets as offensive/non-offensive
python3 src/classify_tweets.py \
    --preprocessed_data data/pp_olid-training-v1.0.tsv \
    --train_data data/clean_train_english.tsv \
    --val_data data/clean_val_english.tsv \
    --train_vectors data/train_vectors.npy \
    --val_vectors data/val_vectors.npy \
    --val_output_csv outputs/D2_val_preds.csv

# Evaluation script
python3 src/eval.py \
    --val_output_csv outputs/D2_val_preds.csv \
    --output_path results/D2_scores.out