feat(rectification): add adaptive question selector

2026-09-03 23:48:14 +08:00
parent 49cd66c537
commit 0be6fcfb4c
12 changed files with 741 additions and 32 deletions
@@ -0,0 +1,156 @@
+from __future__ import annotations
+
+from scripts.active_rectification_questions import build_questionnaire
+from scripts.active_rectification_selector import select_next_questions
+
+
+def test_selector_asks_one_question_and_ranks_by_separation() -> None:
+    questionnaire = build_questionnaire("2001-02-03 10:20", uncertainty_minutes=30)
+
+    selection = questionnaire["selection"]
+
+    assert len(selection["selected_questions"]) == 1
+    assert selection["selected_questions"][0]["id"] == questionnaire["questions"][0]["id"]
+    assert selection["selected_questions"][0]["why_asked"]
+    assert selection["selected_questions"][0]["candidate_ids_distinguished"]
+    assert selection["selected_questions"][0]["technique_routes"]
+    assert selection["ranking"]
+
+
+def test_selector_skips_non_discriminating_questions() -> None:
+    questionnaire = {
+        "question_bank": [
+            {
+                "id": "neutral_only",
+                "domain": "fine_timing",
+                "prompt": "几乎同时吗？",
+                "sensitivity": ["KP_cusp"],
+                "scoring_map": {
+                    "A": {"cluster": "neutral", "points": 0},
+                    "B": {"cluster": "neutral", "points": 0},
+                    "C": {"cluster": "neutral", "points": 0},
+                    "D": {"cluster": "neutral", "points": 0},
+                },
+            },
+            {
+                "id": "usable_question",
+                "domain": "career",
+                "prompt": "工作是否变动？",
+                "sensitivity": ["D10"],
+                "scoring_map": {
+                    "A": {"cluster": "career_up", "points": 2},
+                    "B": {"cluster": "career_up", "points": 1},
+                    "C": {"cluster": "career_down", "points": -2},
+                    "D": {"cluster": "neutral", "points": 0},
+                },
+            },
+        ],
+        "candidate_scan": {"candidate_count": 61},
+    }
+
+    selection = select_next_questions(questionnaire, {}, limit=1)
+
+    assert selection["selected_questions"][0]["id"] == "usable_question"
+    assert any(item["question_id"] == "neutral_only" and item["skipped"] for item in selection["ranking"])
+
+
+def test_selector_changes_domain_after_uncertainty() -> None:
+    questionnaire = build_questionnaire("2001-02-03 10:20", uncertainty_minutes=30)
+    first = questionnaire["questions"][0]["id"]
+
+    selection = select_next_questions(
+        {
+            "question_bank": questionnaire["questions"],
+            "candidate_scan": questionnaire["candidate_scan"],
+        },
+        {first: "D"},
+        limit=1,
+    )
+
+    assert selection["selected_questions"]
+    assert selection["selected_questions"][0]["id"] != first
+    assert selection["selected_questions"][0]["domain"] != questionnaire["questions"][0]["domain"]
+
+
+def test_selector_stops_when_no_answer_can_improve_separation() -> None:
+    questionnaire = {
+        "question_bank": [
+            {
+                "id": "fine_only",
+                "domain": "fine_timing",
+                "prompt": "先内后外吗？",
+                "sensitivity": ["KP_cusp"],
+                "scoring_map": {
+                    "A": {"cluster": "neutral", "points": 0},
+                    "B": {"cluster": "neutral", "points": 0},
+                    "C": {"cluster": "neutral", "points": 0},
+                    "D": {"cluster": "neutral", "points": 0},
+                },
+            }
+        ],
+        "candidate_scan": {"candidate_count": 61},
+    }
+
+    selection = select_next_questions(questionnaire, {}, limit=1)
+
+    assert selection["selected_questions"] == []
+    assert selection["stop"] is True
+    assert selection["stop_reason"] == "no_answer_can_improve_separation"
+
+
+def _minute_question(question_id: str, domain: str, layer: str) -> dict[str, object]:
+    return {
+        "id": question_id,
+        "domain": domain,
+        "prompt": "这个虚构事件是否发生？",
+        "sensitivity": [layer],
+        "positive_cluster": f"{question_id}_yes",
+        "negative_cluster": f"{question_id}_no",
+        "factual_reliability": 0.9,
+        "domain_priority": 1,
+        "scoring_map": {
+            "A": {"cluster": f"{question_id}_yes", "points": 2},
+            "B": {"cluster": f"{question_id}_yes", "points": 1},
+            "C": {"cluster": f"{question_id}_no", "points": -2},
+            "D": {"cluster": "neutral", "points": 0},
+        },
+    }
+
+
+def test_selector_changes_with_remaining_candidate_window() -> None:
+    questions = [
+        _minute_question("relationship_split", "relationship", "D9"),
+        _minute_question("career_split", "career", "D10"),
+    ]
+    relationship_window = {
+        "candidate_count": 3,
+        "transitions": [{"between": ["10:19", "10:20"]}],
+        "rows": [
+            {"divisional_ascendants": {"D9": {"sign": "Aries", "degree": 1}, "D10": {"sign": "Leo"}}},
+            {"divisional_ascendants": {"D9": {"sign": "Taurus", "degree": 2}, "D10": {"sign": "Leo"}}},
+            {"divisional_ascendants": {"D9": {"sign": "Gemini", "degree": 3}, "D10": {"sign": "Leo"}}},
+        ],
+    }
+    career_window = {
+        "candidate_count": 3,
+        "transitions": [{"between": ["10:20", "10:21"]}],
+        "rows": [
+            {"divisional_ascendants": {"D9": {"sign": "Aries"}, "D10": {"sign": "Leo", "degree": 1}}},
+            {"divisional_ascendants": {"D9": {"sign": "Aries"}, "D10": {"sign": "Virgo", "degree": 2}}},
+            {"divisional_ascendants": {"D9": {"sign": "Aries"}, "D10": {"sign": "Libra", "degree": 3}}},
+        ],
+    }
+
+    first = select_next_questions(
+        {"question_bank": questions, "candidate_scan": {"candidate_count": 3, "minute_scan": relationship_window}},
+        {},
+    )
+    second = select_next_questions(
+        {"question_bank": questions, "candidate_scan": {"candidate_count": 3, "minute_scan": career_window}},
+        {},
+    )
+
+    assert first["selected_questions"][0]["id"] == "relationship_split"
+    assert second["selected_questions"][0]["id"] == "career_split"
+    assert first["selected_questions"][0]["minute_relevance"] > 0
+    assert second["selected_questions"][0]["minute_relevance"] > 0