feat(keyword-matcher): jieba分词优化 + 评分归一化 (Phase 1&2)

Phase 1 - 关键词提取优化：
- 引入jieba分词替代暴力替换，解决语义丢失问题
- 精简修饰词列表，添加项目词汇白名单
- extract_value支持中文值提取

Phase 2 - 评分体系归一化：
- 精确匹配(1.0)高于包含匹配(0.3)
- click对INPUT减分更激进(0.1)，fill优先匹配INPUT
- 归一化评分到[0,1]区间

新增测试脚本验证改动
Co-Authored-By: Claude <noreply@anthropic.com>

backend/requirements.txt

@@ -11,3 +11,4 @@ websockets==12.0
 requests==2.31.0
 paho-mqtt==2.1.0
 openpyxl==3.1.2
+jieba>=0.42.1
\ No newline at end of file

backend/scripts/test_phase1_keywords.py

+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+测试脚本：验证 Phase 1 关键词提取优化
+
+测试项：
+1. jieba分词替代暴力替换
+2. 修饰词精简后不再空关键词
+3. 中文值提取
+"""
+
+import sys
+import os
+import io
+
+# 设置标准输出编码为UTF-8（解决Windows GBK编码问题）
+sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf-8')
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+from app.services.keyword_matcher import extract_keywords, extract_value_from_description
+
+
+def test_extract_keywords():
+    """测试关键词提取"""
+    print("=" * 60)
+    print("测试关键词提取（jieba分词版本）")
+    print("=" * 60)
+
+    test_cases = [
+        # (输入, 期望包含的关键词)
+        ("点击【新建会议】按钮", ["新建会议"]),
+        ("点击展开", ["展开"]),  # 之前返回空列表
+        ("点击系统设置", ["系统设置"]),  # 之前返回空列表
+        ("输入会议名称：自动化新建会议", ["会议名称", "自动化", "新建会议"]),
+        ("点击信息发布", ["信息发布"]),
+        ("点击功能中心图标", ["功能中心"]),
+        ("输入备注：这是一个测试备注", ["备注"]),
+        ("点击确定按钮", ["确定"]),
+        ("点击取消按钮", ["取消"]),
+        ("输入用户名：admin@xty", ["用户名"]),
+    ]
+
+    passed = 0
+    failed = 0
+
+    for desc, expected_keywords in test_cases:
+        result = extract_keywords(desc)
+
+        # 检查期望的关键词是否在结果中
+        missing = [kw for kw in expected_keywords if kw not in result]
+
+        if missing:
+            print(f"[FAIL] '{desc}'")
+            print(f"   期望包含: {expected_keywords}")
+            print(f"   实际结果: {result}")
+            print(f"   缺失: {missing}")
+            failed += 1
+        else:
+            print(f"[PASS] '{desc}' -> {result}")
+            passed += 1
+
+    print()
+    print(f"结果: 通过 {passed}/{passed + failed}")
+    return failed == 0
+
+
+def test_extract_value():
+    """测试值提取"""
+    print()
+    print("=" * 60)
+    print("测试值提取（支持中文值）")
+    print("=" * 60)
+
+    test_cases = [
+        # (输入, 期望值)
+        ("输入用户名：admin@xty", "admin@xty"),
+        ("输入账号：admin@xty", "admin@xty"),
+        ("输入密码：Ubains@13579", "Ubains@13579"),
+        ("输入参考价值：100", "100"),
+        ("输入备注：这是一个测试备注", "这是一个测试备注"),
+        ("输入会议名称：自动化测试会议", "自动化测试会议"),
+        ("输入用户名 admin@xty", "admin@xty"),
+        ("输入值'test_value'", "test_value"),
+        ("输入值\"test_value\"", "test_value"),
+    ]
+
+    passed = 0
+    failed = 0
+
+    for desc, expected_value in test_cases:
+        result = extract_value_from_description(desc)
+
+        if result == expected_value:
+            print(f"[PASS] '{desc}' -> '{result}'")
+            passed += 1
+        else:
+            print(f"[FAIL] '{desc}'")
+            print(f"   期望: '{expected_value}'")
+            print(f"   实际: '{result}'")
+            failed += 1
+
+    print()
+    print(f"结果: 通过 {passed}/{passed + failed}")
+    return failed == 0
+
+
+def main():
+    """运行所有测试"""
+    print("\n" + "=" * 60)
+    print("Phase 1 关键词提取优化 - 验证测试")
+    print("=" * 60)
+
+    all_passed = True
+
+    # 测试关键词提取
+    if not test_extract_keywords():
+        all_passed = False
+
+    # 测试值提取
+    if not test_extract_value():
+        all_passed = False
+
+    print()
+    print("=" * 60)
+    if all_passed:
+        print("[SUCCESS] 所有测试通过！Phase 1 验证成功")
+    else:
+        print("[WARNING] 部分测试失败，请检查")
+    print("=" * 60)
+
+    return 0 if all_passed else 1
+
+
+if __name__ == "__main__":
+    sys.exit(main())
\ No newline at end of file

backend/scripts/test_phase2_scoring.py

+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+测试脚本：验证 Phase 2 评分体系归一化
+
+测试项：
+1. 精确匹配 vs 包含匹配 - 精确匹配应排序更高
+2. 动作类型适配 - click优先匹配BUTTON，fill优先匹配INPUT
+"""
+
+import sys
+import os
+import io
+
+# 设置标准输出编码为UTF-8
+sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf-8')
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+from app.services.keyword_matcher import (
+    _calculate_match_score,
+    _apply_action_type_adjustment,
+    EXACT_MATCH_SCORE,
+    CONTAINS_MATCH_SCORE,
+)
+
+
+def test_match_score():
+    """测试匹配分数计算"""
+    print("=" * 60)
+    print("测试匹配分数计算")
+    print("=" * 60)
+
+    test_cases = [
+        # (关键词, 目标文本, 期望精确匹配)
+        ("确定", "确定", True),
+        ("确定", "确定按钮", False),
+        ("新建", "新建会议", False),
+        ("新建会议", "新建会议", True),
+        ("admin", "admin@xty", False),
+        ("admin@xty", "admin@xty", True),
+    ]
+
+    passed = 0
+    failed = 0
+
+    for keyword, target, expect_exact in test_cases:
+        score = _calculate_match_score(keyword, target)
+        is_exact = (score == EXACT_MATCH_SCORE)
+
+        if is_exact == expect_exact:
+            match_type = "精确" if is_exact else "包含"
+            print(f"[PASS] '{keyword}' vs '{target}' -> {match_type}匹配 (score={score:.2f})")
+            passed += 1
+        else:
+            print(f"[FAIL] '{keyword}' vs '{target}'")
+            print(f"   期望: {'精确' if expect_exact else '包含'}匹配")
+            print(f"   实际: score={score:.2f}")
+            failed += 1
+
+    print()
+    print(f"结果: 通过 {passed}/{passed + failed}")
+    return failed == 0
+
+
+def test_action_type_adjustment():
+    """测试动作类型适配"""
+    print()
+    print("=" * 60)
+    print("测试动作类型适配")
+    print("=" * 60)
+
+    test_cases = [
+        # (原分数, 标签, 动作, 期望结果描述)
+        (0.5, "BUTTON", "click", "应该加分"),
+        (0.5, "INPUT", "click", "应该减分"),
+        (0.5, "INPUT", "fill", "应该加分"),
+        (0.5, "BUTTON", "fill", "应该减分"),
+        (0.5, "SELECT", "select", "应该加分"),
+    ]
+
+    passed = 0
+    failed = 0
+
+    for base_score, tag, action, expect_desc in test_cases:
+        adjusted = _apply_action_type_adjustment(base_score, tag, action, [])
+
+        if expect_desc == "应该加分" and adjusted > base_score:
+            print(f"[PASS] {tag}+{action}: {base_score:.2f} -> {adjusted:.2f} ({expect_desc})")
+            passed += 1
+        elif expect_desc == "应该减分" and adjusted < base_score:
+            print(f"[PASS] {tag}+{action}: {base_score:.2f} -> {adjusted:.2f} ({expect_desc})")
+            passed += 1
+        else:
+            print(f"[FAIL] {tag}+{action}: {base_score:.2f} -> {adjusted:.2f} ({expect_desc})")
+            failed += 1
+
+    print()
+    print(f"结果: 通过 {passed}/{passed + failed}")
+    return failed == 0
+
+
+def test_exact_vs_contains_ranking():
+    """测试精确匹配排名高于包含匹配"""
+    print()
+    print("=" * 60)
+    print("测试精确匹配排名高于包含匹配")
+    print("=" * 60)
+
+    # 模拟场景：关键词"确定"
+    # 元素A: text="确定" (精确匹配)
+    # 元素B: text="确定按钮" (包含匹配)
+    # 期望：A的分数高于B
+
+    score_exact = _calculate_match_score("确定", "确定") * 0.15  # text维度权重
+    score_contains = _calculate_match_score("确定", "确定按钮") * 0.15
+
+    if score_exact > score_contains:
+        print(f"[PASS] 精确匹配({score_exact:.4f}) > 包含匹配({score_contains:.4f})")
+        passed = True
+    else:
+        print(f"[FAIL] 精确匹配({score_exact:.4f}) <= 包含匹配({score_contains:.4f})")
+        passed = False
+
+    # 模拟场景：关键词"新建"
+    # 元素A: text="新建" (精确匹配)
+    # 元素B: text="新建会议" (包含匹配)
+    score_exact2 = _calculate_match_score("新建", "新建") * 0.15
+    score_contains2 = _calculate_match_score("新建", "新建会议") * 0.15
+
+    if score_exact2 > score_contains2:
+        print(f"[PASS] 精确匹配({score_exact2:.4f}) > 包含匹配({score_contains2:.4f})")
+        passed2 = True
+    else:
+        print(f"[FAIL] 精确匹配({score_exact2:.4f}) <= 包含匹配({score_contains2:.4f})")
+        passed2 = False
+
+    return passed and passed2
+
+
+def main():
+    """运行所有测试"""
+    print("\n" + "=" * 60)
+    print("Phase 2 评分体系归一化 - 验证测试")
+    print("=" * 60)
+
+    all_passed = True
+
+    if not test_match_score():
+        all_passed = False
+
+    if not test_action_type_adjustment():
+        all_passed = False
+
+    if not test_exact_vs_contains_ranking():
+        all_passed = False
+
+    print()
+    print("=" * 60)
+    if all_passed:
+        print("[SUCCESS] 所有测试通过！Phase 2 验证成功")
+    else:
+        print("[WARNING] 部分测试失败，请检查")
+    print("=" * 60)
+
+    return 0 if all_passed else 1
+
+
+if __name__ == "__main__":
+    sys.exit(main())
\ No newline at end of file