add build_parquet_upload.py

build_parquet_upload.py

@@ -0,0 +1,137 @@
+"""
+Build parquet dataset + upload semua ke HF Hub (folder upload, no rate limit).
+"""
+
+import json
+import os
+import sys
+from pathlib import Path
+
+import pandas as pd
+from huggingface_hub import HfApi, create_repo, repo_exists
+
+REPO_ID = "fassabilf/ir2025-papers-text"
+TEXTS_DIR = Path("texts_2025")
+META_FILES = ["papers_metadata.json", "papers_metadata_extra.json"]
+PARQUET_PATH = "ir2025_papers.parquet"
+
+
+def build_parquet():
+    """Gabung metadata + full text jadi satu parquet."""
+    print("Building parquet...")
+
+    # Load all metadata
+    all_meta = []
+    for mf in META_FILES:
+        if Path(mf).exists():
+            with open(mf) as f:
+                all_meta.extend(json.load(f))
+
+    print(f"  Metadata: {len(all_meta)} papers")
+
+    # Build text lookup: (venue, slug) -> full_text
+    # PDF filename format dari download.py: {idx:03d}_{slug}.pdf
+    # Txt filename: {idx:03d}_{slug}.txt  (sama, cuma ekstensi beda)
+    import re as _re
+    texts_map = {}  # (venue, slug) -> text
+    txt_count = 0
+    for txt_path in sorted(TEXTS_DIR.rglob("*.txt")):
+        if txt_path.name == "all_papers.jsonl":
+            continue
+        try:
+            text = txt_path.read_text(encoding="utf-8")
+            venue = txt_path.parent.name
+            # Extract slug from filename: 001_some_title.txt -> some_title
+            name = txt_path.stem  # tanpa .txt
+            slug = _re.sub(r'^\d{3}_', '', name)  # hapus prefix 001_
+            texts_map[(venue, slug)] = text
+            txt_count += 1
+        except Exception:
+            pass
+
+    print(f"  Texts   : {txt_count} .txt files loaded")
+
+    # Build rows — match metadata ke text via slug
+    rows = []
+    matched = 0
+    for m in all_meta:
+        title = m.get("title", "")
+        venue = m.get("venue", "")
+        # Generate slug sama persis kaya download.py safe_filename
+        clean = "".join(c if c.isalnum() or c in " -_" else "" for c in title)
+        slug = clean.strip().replace(" ", "_")[:60]
+        full_text = texts_map.get((venue, slug), "")
+        if full_text:
+            matched += 1
+
+        rows.append({
+            "venue": venue,
+            "title": title,
+            "authors": "; ".join(m.get("authors", [])) if isinstance(m.get("authors"), list) else m.get("authors", ""),
+            "year": m.get("year", 2025),
+            "doi": m.get("doi", ""),
+            "source": m.get("source", ""),
+            "pdf_url": m.get("pdf_url", ""),
+            "abstract": m.get("abstract", ""),
+            "full_text": full_text,
+        })
+
+    print(f"  Matched : {matched} texts → metadata")
+
+    df = pd.DataFrame(rows)
+    df.to_parquet(PARQUET_PATH, compression="zstd", index=False)
+
+    size_mb = Path(PARQUET_PATH).stat().st_size / (1024 * 1024)
+    print(f"  Parquet : {PARQUET_PATH} ({size_mb:.1f} MB, {len(df)} rows)")
+    print(f"  Columns : {list(df.columns)}")
+    print(f"  With text: {(df['full_text'].str.len() > 0).sum()} papers")
+    return df
+
+
+def upload_all():
+    """Upload parquet + texts folder + metadata ke HF Hub."""
+    token = os.environ.get("HF_TOKEN") or os.environ.get("HUGGINGFACE_HUB_TOKEN")
+    if not token:
+        print("[ERROR] HF_TOKEN not set")
+        sys.exit(1)
+
+    api = HfApi()
+
+    # Buat repo kalo belum ada
+    if not repo_exists(REPO_ID, repo_type="dataset", token=token):
+        create_repo(REPO_ID, repo_type="dataset", token=token)
+        print(f"Repo created: {REPO_ID}")
+
+    # 1. Upload parquet file
+    if Path(PARQUET_PATH).exists():
+        print(f"\nUploading {PARQUET_PATH}...")
+        api.upload_file(
+            path_or_fileobj=PARQUET_PATH,
+            path_in_repo=PARQUET_PATH,
+            repo_id=REPO_ID,
+            repo_type="dataset",
+            token=token,
+            commit_message="add parquet dataset",
+        )
+        print("  ✓ parquet uploaded")
+
+    # 2. Upload metadata files saja (parquet udah include everything)
+    for mf in META_FILES:
+        if Path(mf).exists():
+            print(f"\nUploading {mf}...")
+            api.upload_file(
+                path_or_fileobj=mf,
+                path_in_repo=mf,
+                repo_id=REPO_ID,
+                repo_type="dataset",
+                token=token,
+                commit_message=f"add {mf}",
+            )
+            print(f"  ✓ {mf} uploaded")
+
+    print(f"\nDone: https://huggingface.co/datasets/{REPO_ID}")
+
+
+if __name__ == "__main__":
+    build_parquet()
+    upload_all()