Spaces:
Sleeping
Sleeping
File size: 4,124 Bytes
0e39d80 | 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 | """Resume/CV field extraction and validation."""
from __future__ import annotations
import re
from ml_utils.ocr import OcrResult, get_full_text, get_average_confidence
from ml_utils.extract import find_value_near_label, find_keyword, get_text_in_region, get_nearby_text
EMAIL_RE = re.compile(r"[a-zA-Z0-9._%+\-]+@[a-zA-Z0-9.\-]+\.[a-zA-Z]{2,}")
PHONE_RE = re.compile(r"(?:\+91[\s\-]?)?[6-9]\d{9}|\+?\d{1,3}[\s\-]?\(?\d{3}\)?[\s\-]?\d{3}[\s\-]?\d{4}")
LINKEDIN_RE = re.compile(r"linkedin\.com/in/[\w\-]+", re.I)
SECTION_HEADERS = {
"education": ["education", "academic", "qualification", "academics"],
"experience": ["experience", "work experience", "employment", "professional experience", "work history"],
"skills": ["skills", "technical skills", "core competencies", "technologies", "proficiencies"],
"projects": ["projects", "personal projects", "key projects"],
"certifications": ["certifications", "certificates", "awards", "achievements", "honors"],
"summary": ["summary", "objective", "about me", "profile", "career objective"],
}
def _extract_section(full_text: str, section_keywords: list[str]) -> str:
"""Extract text under a section header until the next section."""
lines = full_text.split("\n")
capturing = False
section_lines: list[str] = []
all_headers = [kw for kws in SECTION_HEADERS.values() for kw in kws]
for line in lines:
lower = line.lower().strip()
if any(kw in lower for kw in section_keywords) and len(lower) < 50:
capturing = True
continue
if capturing:
# Stop if we hit another section header
if any(kw in lower for kw in all_headers) and len(lower) < 50:
break
if line.strip():
section_lines.append(line.strip())
return "\n".join(section_lines)
def parse_resume_fields(ocr_results: list[OcrResult]) -> dict:
"""Extract resume fields from OCR results."""
full_text = get_full_text(ocr_results)
fields: dict = {}
# Name (usually the first/largest text at top)
top_text = get_text_in_region(ocr_results, y_start_pct=0.0, y_end_pct=0.15)
for line in top_text.split("\n"):
cleaned = re.sub(r"[^A-Za-z\s.]", "", line).strip()
if len(cleaned) >= 3 and " " in cleaned:
# Likely a name — not an email or phone
if not EMAIL_RE.search(line) and not PHONE_RE.search(line):
fields["name"] = cleaned
break
# Email
emails = EMAIL_RE.findall(full_text)
if emails:
fields["email"] = emails[0]
# Phone
phones = PHONE_RE.findall(full_text)
if phones:
fields["phone"] = phones[0]
# LinkedIn
linkedin = LINKEDIN_RE.search(full_text)
if linkedin:
fields["linkedin"] = linkedin.group(0)
# Sections
for section_key, section_kws in SECTION_HEADERS.items():
content = _extract_section(full_text, section_kws)
if content:
fields[section_key] = content
return fields
# ── Validator ───────────────────────────────────────────────────────────
def validate_resume(fields: dict, ocr_results: list[OcrResult]) -> tuple[float, list[str]]:
"""Validate resume extraction. Returns (validation_score_0_to_20, flags)."""
flags: list[str] = []
checks_total = 0
checks_passed = 0
for req in ["name", "email"]:
checks_total += 1
if fields.get(req):
checks_passed += 1
else:
flags.append(f"MISSING_{req.upper()}")
optional_filled = sum(1 for k in ["phone", "education", "experience", "skills"]
if fields.get(k))
checks_total += 2
checks_passed += min(2, optional_filled)
if ocr_results:
avg_conf = get_average_confidence(ocr_results)
if avg_conf < 0.4:
flags.append("LOW_OCR_CONFIDENCE")
validation_ratio = checks_passed / max(checks_total, 1)
return validation_ratio * 20.0, flags
|