scripts/05_train_baselines.py

"""Script 05: Train all baselines for comparison.

Three baselines, all on the same train/val/test splits as the Proposed model:
  - Baseline 1: TF-IDF + LogReg          (text only, overall 3-class)
  - Baseline 2: BERT-overall fine-tune   (text only, overall 3-class)
  - Baseline 3: BERT-ACSA (no meta)      (text only, per-aspect 3-class)
                  ^- this is the key ablation for the paper:
                     Baseline 3 vs Proposed isolates the value of
                     metadata fusion.
"""
import argparse
import json
import sys
from pathlib import Path

import pandas as pd

sys.path.insert(0, str(Path(__file__).resolve().parent.parent))

from src.utils import setup_logging
from src import config as cfg
from src.baselines import train_tfidf_baseline
from src.trainer import train_bert_overall, train_acsa


def main():
    parser = argparse.ArgumentParser()
    parser.add_argument("--skip_tfidf", action="store_true")
    parser.add_argument("--skip_bert_overall", action="store_true")
    parser.add_argument("--skip_acsa_no_meta", action="store_true")
    parser.add_argument("--epochs", type=int, default=cfg.DEFAULT_EPOCHS)
    parser.add_argument("--batch_size", type=int, default=cfg.DEFAULT_BATCH_SIZE)
    parser.add_argument("--bert_name", default=cfg.BERT_MODEL_NAME,
                        help="HuggingFace model name (default: config.BERT_MODEL_NAME)")
    args = parser.parse_args()

    setup_logging()
    train_df = pd.read_parquet(cfg.TRAIN_PATH)
    val_df = pd.read_parquet(cfg.VAL_PATH)
    test_df = pd.read_parquet(cfg.TEST_PATH)

    # Baseline 1
    if not args.skip_tfidf:
        print("=" * 60)
        print("Baseline 1: TF-IDF + Logistic Regression (overall 3-class)")
        print("=" * 60)
        _, m = train_tfidf_baseline(train_df, val_df, test_df)
        print(json.dumps({k: v for k, v in m.items() if not k.endswith("_report")},
                         indent=2))

    # Baseline 2
    if not args.skip_bert_overall:
        print("=" * 60)
        print("Baseline 2: BERT fine-tune (overall 3-class)")
        print("=" * 60)
        train_bert_overall(train_df=train_df, val_df=val_df,
                           bert_name=args.bert_name,
                           epochs=args.epochs, batch_size=args.batch_size)

    # Baseline 3 — KEY ablation for the paper
    if not args.skip_acsa_no_meta:
        print("=" * 60)
        print("Baseline 3: BERT-ACSA (per-aspect, NO metadata) — ablation")
        print("=" * 60)
        train_acsa(train_df=train_df, val_df=val_df,
                   bert_name=args.bert_name,
                   epochs=args.epochs, batch_size=args.batch_size)


if __name__ == "__main__":
    main()