import pandas as pd
import ssl
import urllib.request

# 临时绕过 SSL 验证（仅用于下载公开数据）
ssl._create_default_https_context = ssl._create_unverified_context

print("下载 Davidson 数据集...")
url = "https://raw.githubusercontent.com/t-davidson/hate-speech-and-offensive-language/master/data/labeled_data.csv"

# 手动下载
response = urllib.request.urlopen(url)
df = pd.read_csv(response)

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%}")