Spaces:
Sleeping
Sleeping
File size: 16,793 Bytes
300df0f | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 | """
Hierarchical legal document parser — T1.1 (Người B)
Design
------
- Input : clean_html string produced by Người A (T0.4)
- Output: ParseResult with flat list of Segment objects in document order
- Pure Python + BeautifulSoup. No Neo4j, no embedding, no I/O.
- Stateless: create once, call parse() many times.
State machine tracks:
current_phan → current_chuong → current_muc → current_dieu
→ current_khoan → current_diem
Priority order (must be checked top-to-bottom per line/element):
1. Phần (only in Bộ luật)
2. Chương
3. Mục (between Chương and Điều)
4. Điều
5. Khoản (only valid inside a Điều)
6. Điểm (only valid inside a Khoản)
"""
from __future__ import annotations
import re
import logging
from typing import Optional
from .models import HierarchyType, Segment, ParseResult
logger = logging.getLogger(__name__)
# ---------------------------------------------------------------------------
# Regex catalogue — all patterns anchored to start of stripped text
# ---------------------------------------------------------------------------
# Phần — Part (Bộ luật only): "Phần thứ nhất", "Phần I"
RE_PHAN = re.compile(
r"^Phần\s+(?:thứ\s+\w+|[IVX]+)",
re.UNICODE | re.IGNORECASE,
)
# Chương — Chapter: "Chương I", "Chương II.", "CHƯƠNG III"
RE_CHUONG = re.compile(
r"^Chương\s+([IVXLCDM]+)\s*[.:]?\s*(.*)?$",
re.UNICODE | re.IGNORECASE,
)
# Mục — Section: "Mục 1.", "Mục 2. Tên mục"
RE_MUC = re.compile(
r"^Mục\s+(\d+)[.:]?\s*(.*)?$",
re.UNICODE | re.IGNORECASE,
)
# Điều — Article: "Điều 5.", "Điều 10:", "điều 3 ", "Ðiều 1", "Điều thứ 1", "Điều 155a"
RE_DIEU = re.compile(
r"^[ĐĐð][iíìĩị]ều\s+(?:thứ\s+)?(\d+[a-z]*)[.\s:]\s*(.*)?$",
re.UNICODE | re.IGNORECASE,
)
# Khoản — Clause: "1. text", "1a. text" — ONLY valid after a Điều is active
RE_KHOAN = re.compile(
r"^(\d+[a-z]*)\.\s+(.+)$",
re.UNICODE | re.IGNORECASE,
)
# Điểm — Point: "a) text", "b) text" (after Khoản)
RE_DIEM = re.compile(
r"^([a-zđ])\)\s+(.+)$",
re.UNICODE,
)
# Điểm nhỏ — sub-point: "i) text", "ii) text", "iii) text" (rare)
RE_DIEM_NHO = re.compile(
r"^([ivxlcdm]+)\)\s+(.+)$",
re.UNICODE,
)
# Preamble markers — skip these blocks entirely
_PREAMBLE_MARKERS = [
"Căn cứ",
"Theo đề nghị",
"Xét đề nghị",
"Thực hiện",
"Quốc hội nước",
"Chính phủ nước",
]
# Signature / closing block — stop parsing after these appear
_CLOSING_MARKERS = [
"Nơi nhận:",
"TM. CHÍNH PHỦ",
"TM. BỘ",
"KT.",
"CHỦ TỊCH",
"BỘ TRƯỞNG",
"TỔNG CỤC TRƯỞNG",
]
# ---------------------------------------------------------------------------
# UID builder (must match T1.4 schema from Người A)
# ---------------------------------------------------------------------------
def build_uid(
doc_id: str,
hierarchy_type: HierarchyType,
dieu_idx: Optional[str] = None,
khoan_idx: Optional[str] = None,
diem_letter: Optional[str] = None,
) -> str:
"""
Build stable UID for a segment node.
"""
base = f"doc_{doc_id}"
if hierarchy_type == HierarchyType.DIEU:
return f"{base}_dieu_{dieu_idx}"
if hierarchy_type == HierarchyType.KHOAN:
return f"{base}_dieu_{dieu_idx}_khoan_{khoan_idx}"
if hierarchy_type == HierarchyType.DIEM:
return f"{base}_dieu_{dieu_idx}_khoan_{khoan_idx}_diem_{diem_letter}"
return f"{base}_{hierarchy_type.value.lower()}_{dieu_idx}"
# ---------------------------------------------------------------------------
# Parser
# ---------------------------------------------------------------------------
class LegalDocumentParser:
"""
Stateless hierarchical parser for Vietnamese legal documents.
Usage
-----
parser = LegalDocumentParser()
result = parser.parse(doc_id="42", clean_html="<p>Điều 1...</p>")
Notes for implementer (T1.1)
----------------------------
1. Use BeautifulSoup to extract text lines from clean_html.
Recommended: soup.find_all(['p', 'div', 'li']) for line iteration.
2. Strip each element's text before matching against regexes.
3. Keep track of current_dieu_idx, current_khoan_idx throughout iteration.
4. Attach table content (<table> elements) to the last active clause/article.
5. "Phần" is only found in Bộ luật — safe to skip detection for ND/TT.
6. Watch out for numbered preamble items (e.g., "1. Luật này...") that look
like Khoản — only activate Khoản detection AFTER a Điều is seen.
"""
def parse(
self,
doc_id: str,
clean_html: str,
*,
expected_article_count: Optional[int] = None,
loai_van_ban: str = "",
) -> ParseResult:
"""
Parse a single document's HTML into a flat list of Segments.
Parameters
----------
doc_id : str
Document identifier matching Document.id in Neo4j.
clean_html : str
Cleaned HTML string from Người A (T0.4).
Must have <b>/<strong> preserved for heading detection.
expected_article_count : int, optional
If provided, used by ConfidenceScorer to compute ratio.
Obtain from document metadata cross-references.
loai_van_ban : str
"Luật" | "Bộ luật" | "Nghị định" | "Thông tư" | "Thông tư liên tịch"
Affects: Phần detection (only for Bộ luật), preamble handling.
Returns
-------
ParseResult
Flat list of Segment objects in document order.
confidence_score is NOT set yet — call ConfidenceScorer.score() next.
TODO (T1.1): implement this method.
Replace the NotImplementedError below with the state machine.
"""
from bs4 import BeautifulSoup
result = ParseResult(doc_id=doc_id)
soup = BeautifulSoup(clean_html, 'html.parser')
# Find all block-level elements
block_tags_list = ['p', 'div', 'li', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6']
elements = soup.find_all(block_tags_list)
# Filter out elements that contain other block elements to avoid text duplication
leaf_elements = []
for el in elements:
has_block_child = el.find(block_tags_list) is not None
if not has_block_child:
leaf_elements.append(el)
current_phan: Optional[Segment] = None
current_chuong: Optional[Segment] = None
current_muc_title: Optional[str] = None
current_dieu: Optional[Segment] = None
current_khoan: Optional[Segment] = None
current_diem: Optional[Segment] = None
in_quote = False
for el in leaf_elements:
raw_html = str(el)
raw_text = el.get_text(separator=' ', strip=True)
# Normalize whitespace: replace all whitespace sequences (newlines, tabs) with a single space
text = " ".join(raw_text.split())
if not text:
continue
if _is_preamble(text):
continue
# Chỉ dừng (break) khi đã có Điều (Article) được parse.
# Tránh lỗi nhận diện nhầm "CHỦ TỊCH" / "BỘ TRƯỞNG" ở phần tiêu đề đầu văn bản.
if result.article_count > 0 and _is_closing(text):
break
# 0. Quote handling (T1.1 improvement)
# If we are inside a quoted block, skip structural detection (Điều, Khoản, Điểm)
# to avoid false positives when one article quotes another.
quote_marks = ['"', '“', '”', '«', '»']
if in_quote:
# Toggle if we see an odd number of quotes (closing the block)
count = sum(text.count(c) for c in quote_marks)
if count % 2 != 0:
in_quote = False
# Append to current active node
active_node = current_diem or current_khoan or current_dieu or current_chuong or current_phan
if active_node:
active_node.text_content += f"\n{raw_html}"
active_node.clean_text += f"\n{text}"
continue
# Check if this segment starts a quote (usually block quote)
if text.startswith('"') or text.startswith('“') or text.startswith('«'):
count = sum(text.count(c) for c in quote_marks)
if count % 2 != 0:
in_quote = True
# Append to current active node and skip structural checks for this line
active_node = current_diem or current_khoan or current_dieu or current_chuong or current_phan
if active_node:
active_node.text_content += f"\n{raw_html}"
active_node.clean_text += f"\n{text}"
continue
# 1. Phần
if RE_PHAN.match(text) and "luật" in loai_van_ban.lower():
current_phan = Segment(doc_id=doc_id, hierarchy_type=HierarchyType.PHAN, index=1, text_content=raw_html, clean_text=text, title=text)
# Reset lower levels
current_chuong = None
current_muc_title = None
current_dieu = None
current_khoan = None
current_diem = None
continue
# 2. Chương
m_chuong = RE_CHUONG.match(text)
if m_chuong:
roman = m_chuong.group(1)
title_text = m_chuong.group(2) or ""
result.chapter_count += 1
current_chuong = Segment(doc_id=doc_id, hierarchy_type=HierarchyType.CHUONG, index=result.chapter_count, text_content=raw_html, clean_text=text, roman_index=roman, title=title_text)
result.segments.append(current_chuong)
# Reset lower levels
current_muc_title = None
current_dieu = None
current_khoan = None
current_diem = None
continue
# 3. Mục
m_muc = RE_MUC.match(text)
if m_muc:
current_muc_title = text
# Reset lower levels
current_dieu = None
current_khoan = None
current_diem = None
continue
# 4. Điều
m_dieu = RE_DIEU.match(text)
if m_dieu:
dieu_val = m_dieu.group(1) # String to support 155a
title_text = m_dieu.group(2) or ""
uid = build_uid(doc_id, HierarchyType.DIEU, dieu_idx=dieu_val)
parent_uid = None
if current_chuong:
# Chapter uid is not really used for cross-reference, but we use index as id
parent_uid = build_uid(doc_id, HierarchyType.CHUONG, dieu_idx=current_chuong.index)
path = f"Điều {dieu_val}"
if current_chuong:
path = f"Chương {current_chuong.roman_index} / {path}"
current_dieu = Segment(
doc_id=doc_id,
hierarchy_type=HierarchyType.DIEU,
index=dieu_val,
path=path,
text_content=raw_html,
clean_text=text,
parent_uid=parent_uid,
uid=uid,
title=title_text,
section=current_muc_title
)
result.segments.append(current_dieu)
result.article_count += 1
# Reset lower levels
current_khoan = None
current_diem = None
continue
# 5. Khoản
m_khoan = RE_KHOAN.match(text)
if m_khoan and current_dieu:
khoan_val = m_khoan.group(1) # String to support 4a
uid = build_uid(doc_id, HierarchyType.KHOAN, dieu_idx=current_dieu.index, khoan_idx=khoan_val)
current_khoan = Segment(
doc_id=doc_id,
hierarchy_type=HierarchyType.KHOAN,
index=khoan_val,
path=f"{current_dieu.path} / Khoản {khoan_val}",
text_content=raw_html,
clean_text=text,
parent_uid=current_dieu.uid,
uid=uid
)
result.segments.append(current_khoan)
result.clause_count += 1
# Reset lower levels
current_diem = None
continue
# 6. Điểm
m_diem = RE_DIEM.match(text) or RE_DIEM_NHO.match(text)
if m_diem and current_khoan:
letter = m_diem.group(1)
uid = build_uid(doc_id, HierarchyType.DIEM, dieu_idx=current_dieu.index, khoan_idx=current_khoan.index, diem_letter=letter)
current_diem = Segment(
doc_id=doc_id,
hierarchy_type=HierarchyType.DIEM,
index=0,
path=f"{current_khoan.path} / Điểm {letter}",
text_content=raw_html,
clean_text=text,
parent_uid=current_khoan.uid,
uid=uid
)
result.segments.append(current_diem)
result.point_count += 1
continue
# Content
active_node = current_diem or current_khoan or current_dieu or current_chuong or current_phan
if active_node:
active_node.text_content += f"\n{raw_html}"
active_node.clean_text += f"\n{text}"
# Check if a quote started in the middle of this content segment
count = sum(text.count(c) for c in ['"', '“', '”', '«', '»'])
if count % 2 != 0:
in_quote = True
return result
def parse_batch(
self,
documents: list[dict],
*,
loai_van_ban: str = "",
) -> list[ParseResult]:
"""
Parse multiple documents.
Parameters
----------
documents : list of dict, each with keys:
- "doc_id": str
- "clean_html": str
- "expected_article_count": int (optional)
- "loai_van_ban": str (optional — overrides method param)
Returns
-------
list[ParseResult] — same order as input
"""
results = []
for doc in documents:
lvb = doc.get("loai_van_ban", loai_van_ban)
try:
r = self.parse(
doc_id=doc["doc_id"],
clean_html=doc["clean_html"],
expected_article_count=doc.get("expected_article_count"),
loai_van_ban=lvb,
)
except NotImplementedError:
raise
except Exception as exc:
logger.error("Parse failed for doc %s: %s", doc["doc_id"], exc)
r = ParseResult(doc_id=doc["doc_id"])
r.parse_errors.append(str(exc))
results.append(r)
return results
# ---------------------------------------------------------------------------
# Internal helpers (stubs — implement alongside parse())
# ---------------------------------------------------------------------------
def _is_preamble(text: str) -> bool:
"""Return True if the line is part of the preamble and should be skipped."""
return any(text.startswith(marker) for marker in _PREAMBLE_MARKERS)
def _is_closing(text: str) -> bool:
"""Return True if the line signals end of operative content."""
return any(marker in text for marker in _CLOSING_MARKERS)
def _strip_html_tags(html: str) -> str:
"""
Quick tag stripper (no BeautifulSoup) for single-line use.
For full documents, use BeautifulSoup.get_text().
"""
return re.sub(r"<[^>]+>", "", html).strip()
|