File size: 13,045 Bytes
6dfa658
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
"""
Load and preprocess real legal datasets into a unified JSONL corpus.

Supported sources:
- Local exported Kaggle files (JSON, JSONL, CSV, TXT)
- HuggingFace dataset: Renicames/turkish-law-chatbot
"""

from __future__ import annotations

import csv
import json
import re
from pathlib import Path
from typing import Any, Dict, Iterable, List, Optional, Set, Tuple

from datasets import load_dataset


def normalize_whitespace(text: str) -> str:
    """Collapse repeated spaces/newlines into a clean single-space text."""
    return re.sub(r"\s+", " ", text).strip()


def first_non_empty(row: Dict, keys: List[str]) -> str:
    """Return first usable value from candidate keys."""
    for key in keys:
        value = row.get(key)
        if value is None:
            continue
        value_str = normalize_whitespace(str(value))
        if value_str:
            return value_str
    return ""


def extract_text_fields(row: Dict) -> tuple[str, str]:
    """
    Map different possible schemas to unified title/text.
    """
    title = first_non_empty(
        row,
        [
            "title",
            "madde_basligi",
            "kanun_adi",
            "document_title",
            "heading",
            "source",
        ],
    )
    # Prefer a single ``text``-like field if present (use before Q+A merge).
    text = first_non_empty(
        row,
        [
            "text",
            "content",
            "body",
            "madde_metni",
            "article_text",
            "context",
        ],
    )
    if text:
        return title, text

    # Turkish QA exports often use Soru + Cevap; English uses question + answer.
    question = first_non_empty(
        row,
        ["question", "soru", "Soru", "query", "input"],
    )
    answer = first_non_empty(
        row,
        ["answer", "cevap", "Cevap", "response", "output"],
    )
    if question or answer:
        merged = normalize_whitespace(f"{question} {answer}".strip())
        if not title and question:
            title = question[:200] + ("…" if len(question) > 200 else "")
        return title, merged

    text = first_non_empty(
        row,
        [
            "answer",
            "cevap",
            "response",
            "Cevap",
        ],
    )
    return title, text


def _row_lower_keys(row: Dict[str, Any]) -> Dict[str, Any]:
    """Map lowercase column names to values (HuggingFace uses e.g. Soru/Cevap)."""
    return {str(k).lower(): v for k, v in row.items()}


def print_hf_dataset_structure(dataset) -> None:
    """Print splits, column names, and a sample row for debugging."""
    print("\n[HuggingFace] Renicames/turkish-law-chatbot — dataset structure:")
    for split_name, split_data in dataset.items():
        cols = split_data.column_names
        print(f"  Split {split_name!r}: columns ({len(cols)}) = {cols}")
        if len(split_data) > 0:
            sample = split_data[0]
            print(f"  Sample row keys: {list(sample.keys())}")


# Keys that usually carry long legal / conversational text (lowercase).
_QUESTION_KEYS: Tuple[str, ...] = (
    "question",
    "soru",
    "query",
    "input",
    "prompt",
    "instruction",
)
_ANSWER_KEYS: Tuple[str, ...] = (
    "answer",
    "cevap",
    "response",
    "output",
    "completion",
    "targets",
)
_TEXT_SINGLE_KEYS: Tuple[str, ...] = (
    "text",
    "content",
    "body",
    "context",
    "document",
    "passage",
    "article",
    "madde_metni",
    "article_text",
    "raw",
    "message",
)
_TITLE_KEYS: Tuple[str, ...] = (
    "title",
    "heading",
    "source",
    "kanun_adi",
    "madde_basligi",
    "document_title",
    "topic",
)
_METADATA_KEYS: Set[str] = {
    "id",
    "idx",
    "index",
    "split",
    "label",
    "labels",
}


def _first_nonempty_from_lower(lower: Dict[str, Any], keys: Tuple[str, ...]) -> str:
    for key in keys:
        if key not in lower:
            continue
        s = normalize_whitespace(str(lower[key]))
        if s:
            return s
    return ""


def extract_hf_text_and_title(row: Dict[str, Any], split_name: str, row_index: int) -> Tuple[str, str]:
    """
    Build unified (title, text) from one HF row.

    Priority:
    1) If a direct ``text``-like field exists and is non-empty, use it.
    2) If question-like + answer-like fields exist (e.g. Soru + Cevap), merge them.
    3) Otherwise merge remaining string-like fields safely (dedupe parts).
    """
    lower = _row_lower_keys(row)

    # 1) Direct text column
    direct_text = _first_nonempty_from_lower(lower, _TEXT_SINGLE_KEYS)
    if direct_text:
        title = _first_nonempty_from_lower(lower, _TITLE_KEYS)
        if not title:
            title = f"HF {split_name} (text)"
        return title, direct_text

    # 2) Question + answer (Turkish dataset: Soru, Cevap)
    question = _first_nonempty_from_lower(lower, _QUESTION_KEYS)
    answer = _first_nonempty_from_lower(lower, _ANSWER_KEYS)
    if question or answer:
        text = normalize_whitespace(f"{question} {answer}".strip())
        title = _first_nonempty_from_lower(lower, _TITLE_KEYS)
        if not title and question:
            title = question[:200] + ("…" if len(question) > 200 else "")
        elif not title:
            title = f"HF {split_name} QA"
        return title, text

    # 3) Merge all non-metadata string fields (skip duplicates)
    parts: List[str] = []
    seen_lower: Set[str] = set()
    for key in sorted(lower.keys()):
        if key in _METADATA_KEYS:
            continue
        val = lower[key]
        if val is None:
            continue
        s = normalize_whitespace(str(val))
        if not s:
            continue
        s_low = s.lower()
        if s_low in seen_lower:
            continue
        seen_lower.add(s_low)
        parts.append(s)

    merged = normalize_whitespace(" ".join(parts))
    title = _first_nonempty_from_lower(lower, _TITLE_KEYS)
    if not title and parts:
        title = parts[0][:200] + ("…" if len(parts[0]) > 200 else "")
    elif not title:
        title = f"HF {split_name} row {row_index}"
    return title, merged


def _iter_json_records(path: Path) -> Iterable[Dict]:
    with path.open("r", encoding="utf-8") as f:
        data = json.load(f)

    if isinstance(data, list):
        for row in data:
            if isinstance(row, dict):
                yield row
    elif isinstance(data, dict):
        # Some files wrap records in a top-level key.
        for value in data.values():
            if isinstance(value, list):
                for row in value:
                    if isinstance(row, dict):
                        yield row


def _iter_jsonl_records(path: Path) -> Iterable[Dict]:
    with path.open("r", encoding="utf-8") as f:
        for line in f:
            line = line.strip()
            if not line:
                continue
            row = json.loads(line)
            if isinstance(row, dict):
                yield row


def _iter_csv_records(path: Path) -> Iterable[Dict]:
    with path.open("r", encoding="utf-8", newline="") as f:
        reader = csv.DictReader(f)
        for row in reader:
            yield dict(row)


def load_kaggle_exported_records(kaggle_dir: Path) -> List[Dict]:
    """
    Read local files from ``data/kaggle_export/`` (JSON, JSONL, CSV, TXT).

    Each record uses a distinct id ``KG_<index>`` (stable order while scanning files).
    Column detection uses ``extract_text_fields`` (text, content, question+answer, Soru+Cevap, etc.).
    """
    records: List[Dict] = []
    kg_index = 0

    supported_files = sorted(
        list(kaggle_dir.rglob("*.json"))
        + list(kaggle_dir.rglob("*.jsonl"))
        + list(kaggle_dir.rglob("*.csv")),
        key=lambda p: str(p),
    )

    for file_path in supported_files:
        try:
            if file_path.suffix.lower() == ".json":
                iterator = _iter_json_records(file_path)
            elif file_path.suffix.lower() == ".jsonl":
                iterator = _iter_jsonl_records(file_path)
            else:
                iterator = _iter_csv_records(file_path)

            for row in iterator:
                title, text = extract_text_fields(row)
                if not text:
                    continue
                rec_id = f"KG_{kg_index}"
                kg_index += 1
                records.append(
                    {
                        "id": rec_id,
                        "title": title or "Kaggle Turkish Law Record",
                        "text": text,
                    }
                )
        except Exception:
            continue

    for txt_path in sorted(kaggle_dir.rglob("*.txt"), key=lambda p: str(p)):
        try:
            with txt_path.open("r", encoding="utf-8") as f:
                for line in f:
                    line = normalize_whitespace(line)
                    if not line:
                        continue
                    rec_id = f"KG_{kg_index}"
                    kg_index += 1
                    records.append(
                        {
                            "id": rec_id,
                            "title": txt_path.stem,
                            "text": line,
                        }
                    )
        except Exception:
            continue

    return records


def load_hf_records(exclude_splits: Optional[Set[str]] = None) -> List[Dict]:
    """
    Load HuggingFace dataset: Renicames/turkish-law-chatbot.

    The public JSON uses Turkish keys ``Soru`` (question) and ``Cevap`` (answer);
    we detect columns case-insensitively and combine or merge as needed.
    """
    records: List[Dict] = []
    dataset = load_dataset("Renicames/turkish-law-chatbot")
    print_hf_dataset_structure(dataset)

    for split_name, split_data in dataset.items():
        if exclude_splits is not None and split_name in exclude_splits:
            continue
        for idx, row in enumerate(split_data):
            row_dict = dict(row)
            lower = _row_lower_keys(row_dict)

            title, text = extract_hf_text_and_title(row_dict, split_name, idx)
            text = normalize_whitespace(text)
            if not text:
                continue

            row_id = _first_nonempty_from_lower(lower, ("id", "uuid", "doc_id"))
            if not row_id:
                row_id = str(idx)
            title = normalize_whitespace(title) or f"HF {split_name}"
            records.append(
                {
                    "id": f"HF_{split_name}_{row_id}",
                    "title": title,
                    "text": text,
                }
            )

    return records


def deduplicate_records(records: List[Dict]) -> List[Dict]:
    """Drop duplicate (title, text) pairs."""
    unique: List[Dict] = []
    seen = set()
    for rec in records:
        title = normalize_whitespace(rec.get("title", ""))
        text = normalize_whitespace(rec.get("text", ""))
        rec_id = normalize_whitespace(rec.get("id", ""))
        if not text:
            continue
        key = (title, text)
        if key in seen:
            continue
        seen.add(key)
        unique.append({"id": rec_id or f"DOC_{len(unique)}", "title": title or "Untitled", "text": text})
    return unique


def save_jsonl(records: List[Dict], output_path: Path) -> None:
    output_path.parent.mkdir(parents=True, exist_ok=True)
    with output_path.open("w", encoding="utf-8") as f:
        for rec in records:
            f.write(json.dumps(rec, ensure_ascii=False) + "\n")


def build_real_corpus(
    output_path: Path,
    kaggle_dir: Optional[Path] = None,
    include_hf: bool = True,
    hf_exclude_splits: Optional[Set[str]] = None,
) -> List[Dict]:
    """
    Merge Kaggle (``KG_*`` ids) + HuggingFace (``HF_*`` ids) into one JSONL corpus.
    """
    kaggle_records: List[Dict] = []
    hf_records: List[Dict] = []

    if kaggle_dir is not None and kaggle_dir.exists():
        kaggle_records = load_kaggle_exported_records(kaggle_dir)
        print(f"[Corpus] Kaggle records loaded: {len(kaggle_records)} (prefix KG_)")

    if include_hf:
        hf_records = load_hf_records(exclude_splits=hf_exclude_splits)
        print(f"[Corpus] HuggingFace records loaded: {len(hf_records)} (prefix HF_)")

    all_records = kaggle_records + hf_records
    cleaned = deduplicate_records(all_records)
    save_jsonl(cleaned, output_path)
    print(f"[Corpus] Merged total (before dedup): {len(all_records)}")
    print(f"[Corpus] Final merged size (after dedup): {len(cleaned)}")
    print(f"[Corpus] Saved to: {output_path}")
    return cleaned


if __name__ == "__main__":
    project_root = Path(__file__).resolve().parent.parent
    default_output = project_root / "data" / "real_corpus.jsonl"
    default_kaggle = project_root / "data" / "kaggle_export"

    records = build_real_corpus(
        output_path=default_output,
        kaggle_dir=default_kaggle if default_kaggle.exists() else None,
        include_hf=True,
    )
    print(f"Done. Total valid records: {len(records)} -> {default_output}")