Initial commit

extract/experience.py

@@ -0,0 +1,159 @@
+from __future__ import annotations
+
+import re
+from dataclasses import dataclass
+from datetime import date
+from typing import Dict, List, Optional, Tuple
+
+# Month maps (EN + RU)
+MONTHS = {
+    "jan": 1, "january": 1, "янв": 1, "январ": 1,
+    "feb": 2, "february": 2, "фев": 2, "феврал": 2,
+    "mar": 3, "march": 3, "мар": 3, "март": 3,
+    "apr": 4, "april": 4, "апр": 4, "апрел": 4,
+    "may": 5, "май": 5,
+    "jun": 6, "june": 6, "июн": 6, "июнь": 6,
+    "jul": 7, "july": 7, "июл": 7, "июль": 7,
+    "aug": 8, "august": 8, "авг": 8, "август": 8,
+    "sep": 9, "september": 9, "сен": 9, "сент": 9,
+    "oct": 10, "october": 10, "окт": 10, "октя": 10,
+    "nov": 11, "november": 11, "ноя": 11, "ноябр": 11,
+    "dec": 12, "december": 12, "дек": 12, "дека": 12,
+}
+
+PRESENT_RE = re.compile(r"\b(present|now|current|настоящее время|по н\.в\.|по настоящее)\b", re.I)
+
+# Direct "X years" patterns
+DIRECT_YEARS_RE = re.compile(r"(\d+(?:[.,]\d+)?)\s*(?:\+?\s*)?(?:years?|yrs?|лет|года|год)\b", re.I)
+
+# Dates like 03.2019, 2019, Jan 2020, янв 2020
+MMYYYY_RE = re.compile(r"\b(0?[1-9]|1[0-2])[./-](\d{4})\b")
+YYYY_RE = re.compile(r"\b(19\d{2}|20\d{2})\b")
+MON_YYYY_RE = re.compile(r"\b([A-Za-z]{3,9}|[А-Яа-я]{3,9})\.?\s*(\d{4})\b")
+
+# Range separators
+RANGE_RE = re.compile(r"(?P<a>.+?)\s*(?:—|–|-|to|по)\s*(?P<b>.+?)$", re.I)
+
+@dataclass
+class ExpResult:
+    years: Optional[float]
+    confidence: float
+    debug: Dict
+
+def _clamp_years(y: float) -> Optional[float]:
+    if 0.0 <= y <= 45.0:
+        return y
+    return None
+
+def _parse_mon(mon: str) -> Optional[int]:
+    m = mon.strip().lower()
+    m = re.sub(r"[^\wа-я]+", "", m, flags=re.I)
+    # allow prefixes: "январ", "феврал"
+    for k, v in MONTHS.items():
+        if m.startswith(k):
+            return v
+    return None
+
+def _as_ymd(y: int, m: int) -> date:
+    return date(y, m, 1)
+
+def _parse_one_date(s: str) -> Optional[date]:
+    s = s.strip()
+    if PRESENT_RE.search(s):
+        today = date.today()
+        return date(today.year, today.month, 1)
+
+    m1 = MMYYYY_RE.search(s)
+    if m1:
+        mm = int(m1.group(1))
+        yy = int(m1.group(2))
+        return _as_ymd(yy, mm)
+
+    m2 = MON_YYYY_RE.search(s)
+    if m2:
+        mon = _parse_mon(m2.group(1))
+        yy = int(m2.group(2))
+        if mon:
+            return _as_ymd(yy, mon)
+
+    m3 = YYYY_RE.search(s)
+    if m3:
+        yy = int(m3.group(1))
+        return _as_ymd(yy, 1)
+
+    return None
+
+def _merge_intervals(intervals: List[Tuple[date, date]]) -> List[Tuple[date, date]]:
+    if not intervals:
+        return []
+    intervals = sorted(intervals, key=lambda x: (x[0], x[1]))
+    merged = [intervals[0]]
+    for s, e in intervals[1:]:
+        ls, le = merged[-1]
+        if s <= le:
+            merged[-1] = (ls, max(le, e))
+        else:
+            merged.append((s, e))
+    return merged
+
+def _months_between(a: date, b: date) -> int:
+    # month-level difference (inclusive-ish): b >= a
+    return (b.year - a.year) * 12 + (b.month - a.month)
+
+def extract_experience(text: str) -> ExpResult:
+    debug: Dict = {"direct_matches": [], "ranges": [], "intervals": []}
+
+    # 1) Direct years
+    directs = []
+    for m in DIRECT_YEARS_RE.finditer(text):
+        try:
+            v = float(m.group(1).replace(",", "."))
+            if 0 <= v <= 45:
+                directs.append(v)
+                debug["direct_matches"].append({"match": m.group(0), "value": v})
+        except Exception:
+            pass
+    if directs:
+        years = _clamp_years(max(directs))
+        return ExpResult(years=years, confidence=0.90, debug=debug)
+
+    # 2) Ranges in lines: try to detect "start - end"
+    intervals: List[Tuple[date, date]] = []
+    for line in text.splitlines():
+        ln = line.strip()
+        if len(ln) < 7:
+            continue
+        # require range separator
+        if not any(x in ln for x in ("—", "–", "-", " to ", " по ")):
+            continue
+        rr = RANGE_RE.match(ln)
+        if not rr:
+            continue
+        a = rr.group("a")
+        b = rr.group("b")
+        da = _parse_one_date(a)
+        db = _parse_one_date(b)
+        if da and db:
+            if db < da:
+                da, db = db, da
+            # cap extremely old
+            if da.year < 1990:
+                continue
+            intervals.append((da, db))
+            debug["ranges"].append({"line": ln, "start": da.isoformat(), "end": db.isoformat()})
+
+    intervals = _merge_intervals(intervals)
+    debug["intervals"] = [{"start": s.isoformat(), "end": e.isoformat()} for s, e in intervals]
+
+    if not intervals:
+        return ExpResult(years=None, confidence=0.0, debug=debug)
+
+    total_months = 0
+    for s, e in intervals:
+        total_months += max(0, _months_between(s, e))
+    years = round(total_months / 12.0, 2)
+    years = _clamp_years(years) if years is not None else None
+
+    # confidence depends on amount of evidence
+    conf = 0.70 if total_months >= 12 else 0.55
+    return ExpResult(years=years, confidence=conf, debug=debug)