test.py

import pandas as pd

dest_file = "ALL_mtet.txt"
import os 
final_dict = {}

def english_or_vietnam(prompt):
    if "vietnamese" in prompt.lower() or "tiếng việt" in prompt.lower():
        return "vi"
    else:
        return "en"
def clean_text(text):
    new_text = text.replace("['", "").replace("']", "").replace("\t", " ").replace('["', "").replace('"]', "")
    return new_text
count = 0 

'''
deduplicate vietnamese source
'''
for filename in os.listdir("./dataset"):
    print (filename)
    parquet_file = os.path.join("./dataset", filename)
    # Replace 'your_file.parquet' with the actual file path
    df = pd.read_parquet(parquet_file)
    # Save the DataFrame to an Excel file

    for i in range (0, len(df['response'])):
        prompt = df["prompt"][i]
        response = str(df['response'][i]).replace("\t", " ")
        source =  clean_text(str(df['source'][i]))
        if english_or_vietnam(prompt) == "vi":
            # source is en
 
            final_dict[response] = source
        else:
            # source is vi
            if source in final_dict:
                count += 1 
            final_dict[source] = response

print ("duplicate count: ", count)
with open(dest_file, 'w', encoding = 'utf-8') as f:
        # df.to_excel('output_file.xlsx', index=False)
    for key, value in final_dict.items():
        f.write(key + '\t' + value + '\n')