C.E.L_Slide_test2/tests/matching/phase6.py

"""Phase 6 — 4개 유닛 × 7개 방법 통합표
가로 7개: 키워드 3 (TF-IDF, Char, Kiwi+BM25) + 구조 1 + 의미 3 (청킹+sbert, 청킹+E5, Cross)

데이터 소스:
- Phase 2 MD: 키워드 3 + 구조 + SBERT-청크 (청킹+sbert)
- Phase 4 MD: E5-청크 (청킹+E5), Cross-원본 (Cross)
"""
import re
from pathlib import Path

HERE = Path(__file__).parent

TARGET_UNITS = [
    ("6",  "1. (MDX 1) 팝업 — DX와 BIM의 구분",      "18"),
    ("12", "2. (MDX 2) 2.2 DX 시행 주체별 기대효과",    "14"),
    ("13", "3. (MDX 03) 1. DX 시행을 위한 필수요건 — 기술·사람·자연",  "13"),
    ("14", "4. (MDX 03) 2. Process 혁신과 Product 변화", "29"),
]

# (소스 파일, 소스 방법명, 표시 이름)
METHOD_MAP = [
    ("phase2", "TF-IDF",       "TF-IDF"),
    ("phase2", "Char-ngram",   "Char 3-gram"),
    ("phase2", "Kiwi+BM25",    "Kiwi+BM25"),
    ("phase2", "Structural",   "구조 메타"),
    ("phase2", "SBERT-청크",    "청킹+sbert"),
    ("phase4", "E5-청크",       "청킹+E5"),
    ("phase4", "Cross-원본",    "Cross"),
]


def load_unit_blocks(md_name):
    md = (HERE / md_name).read_text(encoding="utf-8")
    parts = re.split(r"\n## (\d+)\. ", md)
    return {parts[i]: parts[i + 1] for i in range(1, len(parts), 2)}


blocks_p2 = load_unit_blocks("MATRIX_PHASE2.md")
blocks_p4 = load_unit_blocks("MATRIX_PHASE4.md")


def extract_top3(block_text, method_name):
    for line in block_text.split("\n"):
        if line.startswith(f"| {method_name} |"):
            cells = line.split(" | ")
            if len(cells) >= 4:
                return [c.strip().rstrip("|").strip() for c in cells[1:4]]
    return ["-", "-", "-"]


def mark_correct(cell, correct_id):
    if f"**{correct_id}**" in cell:
        return "⭐ " + cell
    return cell


def get_top3(source, method, unit_num):
    blocks = blocks_p2 if source == "phase2" else blocks_p4
    block = blocks.get(unit_num, "")
    return extract_top3(block, method)


# 리포트 조립
lines = []
lines.append("# Phase 6 — 4개 유닛 × 7개 방법 통합표")
lines.append("")
lines.append("**콘텐츠 4개:**")
for _, display_name, correct_id in TARGET_UNITS:
    lines.append(f"- {display_name} (정답 Frame **{correct_id}**)")
lines.append("")
lines.append("**방법 7개**: 키워드 3 + 구조 1 + 의미 3 (청킹+sbert / 청킹+E5 / Cross)")
lines.append("")
lines.append("정답 프레임은 ⭐ 표시")
lines.append("")

header = ["콘텐츠"] + [m[2] for m in METHOD_MAP]
lines.append("| " + " | ".join(header) + " |")
lines.append("|" + "|".join(["------"] * len(header)) + "|")

for num, display_name, correct_id in TARGET_UNITS:
    method_tops = {}
    for source, src_name, disp in METHOD_MAP:
        method_tops[disp] = get_top3(source, src_name, num)

    header_cell = f"**{display_name}**<br>정답 = Frame **{correct_id}**"
    lines.append(f"| {header_cell} | " + " | ".join([""] * len(METHOD_MAP)) + " |")

    for rank_idx, rank_label in enumerate(["**1순위**", "**2순위**", "**3순위**"]):
        row = [rank_label]
        for _, _, disp in METHOD_MAP:
            cell = method_tops[disp][rank_idx]
            row.append(mark_correct(cell, correct_id))
        lines.append("| " + " | ".join(row) + " |")

lines.append("")
lines.append("---")
lines.append("")
lines.append("## 방법별 1순위 정답 집계")
lines.append("")
lines.append("| 방법 | MDX1 팝업(18) | MDX2 2.2(14) | MDX03-1(13) | MDX03-2(29) | 정답 합계 |")
lines.append("|------|---------------|--------------|-------------|-------------|----------|")

for source, src_name, disp in METHOD_MAP:
    results = []
    for num, _, correct_id in TARGET_UNITS:
        top3 = get_top3(source, src_name, num)
        top1 = top3[0]
        results.append(f"**{correct_id}**" in top1)
    marks = ["✓" if r else "✗" for r in results]
    score = sum(results)
    lines.append(f"| {disp} | {marks[0]} | {marks[1]} | {marks[2]} | {marks[3]} | **{score}/4** |")

out_path = HERE / "MATRIX_PHASE6.md"
out_path.write_text("\n".join(lines), encoding="utf-8")
print(f"완료: {out_path}")

# 검증 출력
print("\n=== 검증: 각 유닛 1순위 ===")
for num, display_name, correct_id in TARGET_UNITS:
    print(f"\n[{display_name}] 정답={correct_id}")
    for source, src_name, disp in METHOD_MAP:
        top3 = get_top3(source, src_name, num)
        m = re.search(r"\*\*(\d{2})\*\*", top3[0])
        fid = m.group(1) if m else "-"
        mark = "✓" if fid == correct_id else "✗"
        print(f"  {disp:15s} (src={source}/{src_name}) → {fid} {mark}")