update

knowledge_search_traverse.py

@@ -231,12 +231,13 @@ semantic_segmentation_instructions = """
 5. **逻辑连接**：并且、或者、以及等连接词（较少出现）
 
 ## 分段原则
-1. **语义完整性**：每个片段应该是一个完整的语义单元
+1. **语义完整性**：每个片段应该是一个完整的语义单元，如"如何制作猫咪表情包梗图"可以拆分为"如何"、"制作"、"猫咪表情包梗图"三个片段，每一段有完整语义
 2. **类型互斥**：每个片段只能属于一种类型
 3. **保留原文**：片段文本必须保留原query中的字符，不得改写
 4. **顺序保持**：片段顺序应与原query一致
 
 
+
 ## 输出要求
 - segments: 片段列表
   - segment_text: 片段文本（必须来自原query）

test_evaluation_v3.py

@@ -88,11 +88,12 @@ async def test_evaluation_v3(run_context_path: str, max_posts: int = 10):
 
     # 3. 处理结果
     results = []
+    detailed_reports = []  # 收集详细评估报告
     print(f"📊 处理评估结果...\n")
     for i, ((round_idx, search_idx, post_id, post, _), eval_result) in enumerate(zip(tasks, all_eval_results), 1):
         knowledge_eval, content_eval, purpose_eval, category_eval, final_score, match_level = eval_result
 
-        print(f"  [{i}/{len(tasks)}] {post.note_id}", end="")
+        print(f"  [{i}/{len(tasks)}] {post.note_id} - {post.title[:40]}", end="")
         if knowledge_eval:
             if final_score is not None:
                 print(f" → {match_level} ({final_score:.1f}分)")
@@ -103,6 +104,52 @@ async def test_evaluation_v3(run_context_path: str, max_posts: int = 10):
             else:
                 print(f" → 评估未完成")
 
+            # 打印详细判断原因
+            print(f"      📝 知识评估: {knowledge_eval.conclusion if knowledge_eval.conclusion else '无'}")
+            if content_eval and content_eval.is_content_knowledge:
+                print(f"      📚 内容知识: {content_eval.summary[:80] if content_eval.summary else '无'}...")
+            if purpose_eval:
+                print(f"      🎯 目的匹配: {purpose_eval.core_basis[:80] if purpose_eval.core_basis else '无'}...")
+            if category_eval:
+                print(f"      🏷️  品类匹配: {category_eval.core_basis[:80] if category_eval.core_basis else '无'}...")
+            print()
+
+            # 收集详细报告
+            detailed_report = {
+                'post_index': i,
+                'note_id': post.note_id,
+                'title': post.title,
+                'final_score': final_score,
+                'match_level': match_level,
+                'is_knowledge': knowledge_eval.is_knowledge if knowledge_eval else None,
+                'is_content_knowledge': content_eval.is_content_knowledge if content_eval else None,
+                'knowledge_score': content_eval.final_score if content_eval else None,
+                'evaluations': {
+                    'knowledge': {
+                        'conclusion': knowledge_eval.conclusion if knowledge_eval else None,
+                        'core_evidence': knowledge_eval.core_evidence if knowledge_eval and hasattr(knowledge_eval, 'core_evidence') else None,
+                        'issues': knowledge_eval.issues if knowledge_eval and hasattr(knowledge_eval, 'issues') else None
+                    },
+                    'content_knowledge': {
+                        'summary': content_eval.summary if content_eval else None,
+                        'final_score': content_eval.final_score if content_eval else None,
+                        'level': content_eval.level if content_eval else None
+                    } if content_eval and content_eval.is_content_knowledge else None,
+                    'purpose': {
+                        'score': purpose_eval.purpose_score if purpose_eval else None,
+                        'core_motivation': purpose_eval.core_motivation if purpose_eval else None,
+                        'core_basis': purpose_eval.core_basis if purpose_eval else None,
+                        'match_level': purpose_eval.match_level if purpose_eval else None
+                    } if purpose_eval else None,
+                    'category': {
+                        'score': category_eval.category_score if category_eval else None,
+                        'core_basis': category_eval.core_basis if category_eval else None,
+                        'match_level': category_eval.match_level if category_eval else None
+                    } if category_eval else None
+                }
+            }
+            detailed_reports.append(detailed_report)
+
             # 应用评估结果
             apply_evaluation_v3_to_post(
                 post,
@@ -115,7 +162,7 @@ async def test_evaluation_v3(run_context_path: str, max_posts: int = 10):
             )
             results.append((round_idx, search_idx, post_id, post))
         else:
-            print(f" → ❌ 评估失败")
+            print(f" → ❌ 评估失败\n")
 
     print(f"\n✅ 评估完成: {len(results)}/{len(posts)} 成功\n")
 
@@ -155,7 +202,22 @@ async def test_evaluation_v3(run_context_path: str, max_posts: int = 10):
     output_path = run_context_path.replace('.json', '_v3.json')
     with open(output_path, 'w', encoding='utf-8') as f:
         json.dump(run_context, f, ensure_ascii=False, indent=2)
-    print(f"✅ 已保存: {output_path}\n")
+    print(f"✅ 已保存: {output_path}")
+
+    # 保存详细评估报告
+    report_path = run_context_path.replace('.json', '_evaluation_report.json')
+    evaluation_report = {
+        'metadata': {
+            'original_query': original_query,
+            'total_posts': len(results),
+            'evaluation_time': datetime.now().strftime('%Y-%m-%d %H:%M:%S'),
+            'evaluator_version': 'v3.0'
+        },
+        'detailed_reports': detailed_reports
+    }
+    with open(report_path, 'w', encoding='utf-8') as f:
+        json.dump(evaluation_report, f, ensure_ascii=False, indent=2)
+    print(f"📄 已保存详细评估报告: {report_path}\n")
 
     # 生成统计报告
     print(f"\n{'='*80}")

visualization/knowledge_search_traverse/index.js

@@ -678,7 +678,7 @@ function NoteNode({ id, data, sourcePosition, targetPosition }) {
                     {data.purpose_evaluation && (
                       <div style={{ fontSize: '24px', color: '#9f1239', lineHeight: '1.4', marginBottom: '12px' }}>
                         <div style={{ fontWeight: '600', marginBottom: '6px' }}>
-                          目的性匹配（{data.purpose_score}分，占比70%）
+                          目的性匹配（{data.purpose_score}分）
                         </div>
                         <div>{data.purpose_evaluation.core_basis || '无评估信息'}</div>
                       </div>
@@ -686,7 +686,7 @@ function NoteNode({ id, data, sourcePosition, targetPosition }) {
                     {data.category_evaluation && (
                       <div style={{ fontSize: '24px', color: '#9f1239', lineHeight: '1.4' }}>
                         <div style={{ fontWeight: '600', marginBottom: '6px' }}>
-                          品类匹配（{data.category_score}分，占比30%）
+                          品类匹配（{data.category_score}分）
                         </div>
                         <div>{data.category_evaluation.core_basis || '无评估信息'}</div>
                       </div>