import pandas as pd

df = pd.read_csv('labeled_data.csv')
df['label'] = (df['class'] != 2).astype(int)
df_sample = df[['tweet', 'label']].rename(columns={'tweet': 'text'})
df_sample.to_csv('toxic_data.csv', index=False)

print(f"✅ 完成！共 {len(df_sample)} 条，有害比例: {df_sample['label'].mean():.2%}")