fix: dedup only same-author duplicates

중복 제거를 텍스트만 보고 하던 것을 작성자까지 보게 고친다. 저자가 같으면
같은 사람 글이 두 번 적재된 것이라 침해가 아니지만, 저자가 다르면 그 중복
자체가 침해 후보다. 앞서는 이를 구분하지 않아 256건을 모두 뺐고, 그중
17건이 서로 다른 작성자의 글이었다.

같은 작성자 그룹만 제외하도록 목록을 다시 만들어(239건) 재실행하니 검사
대상 6104건, 침해의심 106건이다. 80 -> 106 의 증가분은 되살린 17건 전부와,
그 17건이 색인에 돌아오며 새로 매칭된 9건이다. 빠진 건은 없다.
106건 전수 확인 결과 매칭 상대의 작성자는 모두 다르다.

판정 사유 표기도 바로잡는다. 행 단위 대표값으로 재현하면 서로 다른 후보가
각기 다른 조건을 충족한 3건이 '사유 미상' 으로 빠졌다. 후보 판정재료를
쓰도록 고쳤다.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

reports/dedup_exclude_same_author.txt

@@ -0,0 +1,239 @@
+seg-d3fbe94228ab881296b6
+seg-751c3589b594278c083a
+seg-5fc0614ce99a6618c8df
+seg-e9d6cfe7d8c32ac76685
+seg-5661be946f8b631f15a5
+seg-075a0e34ae4297d7a690
+seg-50db79ea6d2727c2294b
+seg-247b665d9d4ee02b33f9
+seg-44984b4a4e98fa4a523c
+seg-07581bfbc6837377918b
+seg-bff759c20055d96d1f4b
+seg-f5c5bca93e58b039a022
+seg-f5d329cc2b30ae71f105
+seg-4d6b39d38d78d612702c
+seg-0edf0ca49a261c64c385
+seg-55dde76d300a6ae8c3de
+seg-ffdde9a61951d9ebb1a3
+seg-410f0ebe300e69c627b9
+seg-1245e54f50bc293ccaf6
+seg-8331dded31e15a735156
+seg-6f4f844cfa81f266abce
+seg-8c6cacccd47732ef2df3
+seg-0f3197e20379dd548cc3
+seg-9a98e4bbf3c132bb9b5b
+seg-23ded3d3143f049f3082
+seg-e37c0908ca2aa3f55503
+seg-24f8107434fc48c067b1
+seg-5d4f783c3645437c6070
+seg-5274f32e15cf95d178d0
+seg-d243f0147bf6b556f572
+seg-a467d7d4e7c1fb53da80
+seg-cc9363262d7bb9159306
+seg-5d7f16828df7db4a2b76
+seg-54022db9a1d633a4045d
+seg-d0b69f55d3f934f87b5f
+seg-bdb4188636fcf0d96ca4
+seg-288dc6c99554be21fa18
+seg-cf0cdf656e0fb06ef873
+seg-eb77d16bc244ad727b84
+seg-30d7e35a031a46cc71e0
+seg-a42a048a63356af43586
+seg-5ff84bb1c602e8410ed7
+seg-bbd6204bfd92ce23ae88
+seg-d1830d7e5b83dce69c15
+seg-bf88dc382fa57030f641
+seg-be8c1f14df73996ff3fc
+seg-22a696af220d87a59378
+seg-ac6d6a1c976b9d03b8a1
+seg-d383082fce6037b85412
+seg-3ef361f6e3f6d8238949
+seg-dde5178ca77977d2af77
+seg-430ce38f4aeaec2ee30d
+seg-dc3ed111c198333b41a6
+seg-62dbfefbb87ae7edffb5
+seg-70c32fa5c3fa064d8517
+seg-38ce69b80bff6bf91bad
+seg-aa5a97c20e56d3e06a18
+seg-a538c6ed532265042c2c
+seg-8b80b03f94f0ff0a09fc
+seg-ef8b9a90adaabae27467
+seg-9371d04bedc46ff3307a
+seg-2e774b13ce5649bbef44
+seg-7a544a5709a515eebe9f
+seg-f5b34bfd2d2808b30d57
+seg-a5827f00cc888d6bd553
+seg-e3e107d84c4705a9cce1
+seg-53fe28079bf8e65376da
+seg-8c2c713c44ef651f5138
+seg-d182c9b4a909dd0a7bbb
+seg-93ceac7aa9a9e5a06ce1
+seg-a398d393288e13d66de3
+seg-52cb8da0dc336319f732
+seg-ae11015b4ae9b766d5c3
+seg-30b280d1c2859a5ae036
+seg-9c13ba45ae839ebfc44d
+seg-5600dc312d9effe82322
+seg-cd270243863ef7841dfb
+seg-cf8f63bf5fe850471401
+seg-7507defdcb9416205954
+seg-f7e0090d229fc1a308b0
+seg-e0d851a207476f87c789
+seg-bd5662c91ab35270516b
+seg-be04976635524964ce62
+seg-fa44f8a2635163b24fb7
+seg-62d41db76b35c50fc80c
+seg-943fd1f2254a41d56795
+seg-9dcd69c2d11e6cca179e
+seg-7d35c0b29a378389e352
+seg-76b19ec66f94dffaf00a
+seg-e66251e5a312bc798e2e
+seg-a9b38449e85ed466047e
+seg-abf6d9be2a35224b9e09
+seg-aab80bc4d6a9f419c723
+seg-bd7b1fd2bc26447c9910
+seg-8da64f864be5944aedef
+seg-587d9bf5b0eca24fe707
+seg-9eb5bd09ab63980559b4
+seg-ee19e6aaf1b089d3e641
+seg-8a7b17ec37d128136ac9
+seg-4504aa1d788d212d2df5
+seg-b98614064b81140e0fe4
+seg-966a6311fb356d92b6d1
+seg-bc2e4bb72583ecf07179
+seg-576943b2975470189554
+seg-f773aa3058f06ae79b47
+seg-f4be6c5c541c902d642b
+seg-a3fca07ea170e36417e2
+seg-c5352306e93499c0a97b
+seg-7c843b92edcd5b2a9fd9
+seg-81c3bd5d41f4a068de6e
+seg-8f78bae75bb369f94503
+seg-e0de5524dfe1159eb363
+seg-ef54adc86f4a0a297118
+seg-bedd4f81482541351bb7
+seg-e811cffbcc76695b6e6f
+seg-c3b58a34306a6a4e5b59
+seg-bf75b74deb39ccf1303a
+seg-c8d5a76ff7e035ef22b6
+seg-b8637db3c35177137a3f
+seg-ccaf5c07904036512566
+seg-7419176fd92cceea9686
+seg-b72eea0366a31e987386
+seg-69ea27fd2881c7627ded
+seg-f7a461e0a15fb18ba98f
+seg-896990d33e817107c0bc
+seg-8cecf2cdee3da79b1b05
+seg-f885a8e028dadc75d8e2
+seg-e83d73a7381c3c9d52e1
+seg-a8facc3763ded9d2ddb1
+seg-fafe437f7069e5fd7722
+seg-7bf545545753af04309f
+seg-af0a392936f76d3f0bbc
+seg-5956baedf58f802290ca
+seg-8f64481ef86947095f9c
+seg-68644acb7b741ed0040f
+seg-f5a3280bab5b83a6716b
+seg-e00351c4a290824c305c
+seg-881fe2194708c84ff2bc
+seg-905c5684472d150fbaf8
+seg-9977b26f515658764743
+seg-f25987b145b277afd1bb
+seg-8620bf9b61e70258c169
+seg-b94936f18a528d1d1484
+seg-78b533ef1ae0d6edae2d
+seg-f4dd4bb7814c681310be
+seg-ef082f2347172e8cecf4
+seg-8ad3bcfb1e19146479aa
+seg-d9651090b7a6aa2f0430
+seg-e1e78ffdc76fd9ffaf9e
+seg-cff3db09ecfb8d7fc0d7
+seg-a6355ffad5bd71af53c6
+seg-d53dda8d81b2adf120e8
+seg-67563b9afa606c5587c7
+seg-c43016a4caaf3143cd98
+seg-a0837f25f895120a065f
+seg-c9da95e71f1e2fabd0ec
+seg-fbb309c63c26989c8576
+seg-c29909ee9a58cb26b01a
+seg-819aecc9f54fd9b55be2
+seg-d0a0f26c8be62bfb2d0c
+seg-fd26051079a4bf040532
+seg-98a5697519e5f97e249a
+seg-d2751017aa736359f7d4
+seg-74fed19b12664e50be9c
+seg-c23c2fc3639889509e9e
+seg-c36dff9dde138672ddee
+seg-e3c87966f3b3f2bb6841
+seg-e4771ac5792eeb914b07
+seg-cd775fabaf6cb4994cdf
+seg-7a1c0983f84940c7201b
+seg-9d79705b707d9e32b117
+seg-b0c16c55fa9e8277349d
+seg-d57e517755b00919daca
+seg-79b6ba59c8781f00ce0a
+seg-f6c8922e9d20de2076ff
+seg-c93928d4782eb4c6afe8
+seg-93bcd6a33398bd7280ec
+seg-f832f4d3a11932fd556a
+seg-d6810579afc033bd1b14
+seg-b6ed3f352a7847b0b8a1
+seg-a7358497f1e85ad51ff8
+seg-e17e714288dc109c5baa
+seg-a1655a23565dc471aa47
+seg-d5fe534027eb9c1e85fd
+seg-ab243504653f9648e7bf
+seg-dfec7d5d1b119a418561
+seg-bf99840a40b62052271b
+seg-a985637ec6ab3d46d555
+seg-a9413062361aa4f345c4
+seg-e57636f9636d23ef4039
+seg-9554999648a9aadc31bd
+seg-c62f06ad5656a4599475
+seg-b49e6da1a3a6b0eba049
+seg-d7063bf4d41293c25c90
+seg-ccf95d993d1c472e27e9
+seg-b7f22ef0b4697779bde5
+seg-bbce92de4976f5dc93a1
+seg-f03532de0097e52edb9f
+seg-f70fdab6a38ff5140308
+seg-c23074b3203f30fbba5c
+seg-d6ea9afd3697d2aaefe1
+seg-d07c0d0cd0d0aea3cfc9
+seg-eead2327f235ebd6efe7
+seg-9fd1b8cd453902242690
+seg-a82cd00ce71260f0350a
+seg-a5a838ab500c1da7e11a
+seg-a311b6e8f8c03dc79c1e
+seg-fff1369f0ea4f9f17516
+seg-d3dbae3326a84db54529
+seg-ed34c62c55263092b07e
+seg-d35338514edde20b18b5
+seg-aa007367cea2c002f094
+seg-d0b236f8748a269bdecc
+seg-ed816d3c437985f03ea5
+seg-debed50cad3a7946a2cf
+seg-eaacca81f4591e799863
+seg-fda2d67cf5a23cb91b3e
+seg-e7ad9915d5e81d735cb0
+seg-df4cfebb17d4574beaf6
+seg-c0e7b6315ea41307ea0d
+seg-cb9c60dd9ccad9606a72
+seg-fb900d525ce3b5fac463
+seg-e75049a453021a03ff30
+seg-e01166955fe0dadec87d
+seg-fccf864c1bd61629218c
+seg-d8ecd0403ec0c7ddbdbb
+seg-fecff1ad96f117b29fb9
+seg-ea129bba581d8ab62b7a
+seg-da7ac80c29ff726d58ab
+seg-dd7f88929af72ef109e2
+seg-fdc6d5861c7d7e2edee9
+seg-f923d6974271cb91fb5a
+seg-f1c0cf14a231160b7f3e
+seg-e97c69c482cd66aa04cd
+seg-f552457bdb5adcf726fe
+seg-ecbbb6b9ee769091fdb5
+seg-f9da50c0714c3e8eca0a
+seg-ec05c1d86aacdb87b00f
+seg-f97a6835ca56b8bcf03d

scripts/build_report_xlsx.py

@@ -489,7 +489,22 @@ COVERAGE_MIN = 0.30
 
 
 def judgment_reasons(row: dict, longest_span: int) -> list[str]:
-    """어느 조건으로 의심 판정이 났는지. detector.py 의 OR 조건과 같은 기준."""
+    """어느 조건으로 의심 판정이 났는지. detector.py 의 OR 조건과 같은 기준.
+
+    엔진은 후보 하나하나에 조건을 적용한다. 행 단위 대표값(결합유사도·
+    union_coverage)으로 재현하면 서로 다른 후보가 각기 다른 조건을 충족한
+    경우를 놓치므로, 후보 판정재료가 있으면 그것을 쓴다.
+    """
+    candidates = row.get("후보 판정재료")
+    if candidates:
+        reasons = []
+        if any(c["combined"] >= SCORE_MIN for c in candidates):
+            reasons.append("①유사도")
+        if any(c["longest"] >= SPAN_MIN for c in candidates):
+            reasons.append("②연속일치")
+        if any(c["coverage"] >= COVERAGE_MIN for c in candidates):
+            reasons.append("③커버리지")
+        return reasons
     reasons = []
     if (row.get("결합유사도") or 0) >= SCORE_MIN:
         reasons.append("①유사도")