scripts/generate_synthetic.py

import os
import sys
import yaml
import argparse
import pandas as pd
import numpy as np
import random
from datetime import datetime, timedelta

# Import text generator
sys.path.append(os.path.dirname(os.path.abspath(__file__)))
from text_generator import generate_speech, PARTIES, SPEAKERS, CHAMBERS

def load_config(config_path):
    with open(config_path, "r", encoding="utf-8") as f:
        return yaml.safe_load(f)

def generate_ai_corpus(sample_size, models, doc_types, seed=42):
    """Generates a synthetic AI dataset representing text produced by different LLMs."""
    print(f"Generating synthetic AI corpus ({sample_size} samples) across models {models}...")
    random.seed(seed + 1000)
    np.random.seed(seed + 1000)
    
    data = []
    # AI speeches are typically newer (e.g. 2022-2026)
    start_date = datetime(2022, 11, 30) # ChatGPT release date
    end_date = datetime(2026, 5, 1)
    date_range_days = (end_date - start_date).days
    
    for i in range(sample_size):
        date_speech = start_date + timedelta(days=random.randint(0, date_range_days))
        model = random.choice(models)
        doc_type = random.choice(doc_types)
        speaker = random.choice(SPEAKERS)
        party = random.choice(PARTIES)
        chamber = random.choice(CHAMBERS)
        
        # Calculate legislature (16th or higher)
        leg = "16" if date_speech.year >= 2022 else "15"
        
        # Simulating standard prompt injected for generation
        prompt = f"Rédige un texte parlementaire en français de type '{doc_type}' sur un sujet de politique nationale. Style : {model}."
        
        speech_text = generate_speech(is_ai=True, ai_model=model, doc_type=doc_type, seed=i + 10000)
        
        # Clean text
        speech_text = " ".join(speech_text.split())
        
        data.append({
            "text": speech_text,
            "label_human_ai": 1,
            "source": model,
            "speaker": speaker,
            "party": party,
            "date": date_speech.strftime("%Y-%m-%d"),
            "chamber": chamber,
            "document_type": doc_type,
            "legislature": leg,
            "generation_prompt": prompt
        })
        
    return pd.DataFrame(data)

def main():
    parser = argparse.ArgumentParser(description="Generate synthetic AI parliamentary corpus.")
    parser.add_argument("--config", default="configs/config.yaml", help="Path to config file")
    args = parser.parse_args()
    
    config = load_config(args.config)
    raw_dir = config["paths"]["raw_dir"]
    os.makedirs(raw_dir, exist_ok=True)
    
    sample_size = config["data_collection"]["sample_size_ai"]
    models = config["synthetic_generation"]["models"]
    doc_types = config["synthetic_generation"]["document_types"]
    seed = config["data_collection"]["seed"]
    
    df_ai = generate_ai_corpus(sample_size, models, doc_types, seed)
    output_path = os.path.join(raw_dir, "ai_corpus.csv")
    df_ai.to_csv(output_path, index=False)
    print(f"Successfully generated and saved AI corpus to {output_path}")

if __name__ == "__main__":
    main()