fix(aihr): clean video asr snippets

backend/ruoyi-modules/ruoyi-aihr/src/main/java/org/dromara/aihr/service/AihrSopSeedService.java

@@ -2214,7 +2214,7 @@ public class AihrSopSeedService {
         KnowledgeHit first = evidenceHits.get(0);
         String responseCategory = isBlank(category) ? first.category() : category;
         List<SnippetResponse> snippets = evidenceHits.stream()
-            .map(hit -> new SnippetResponse("第 " + hit.idx() + " 段：" + hit.title(), hit.content(), hit.fragmentId()))
+            .map(hit -> new SnippetResponse("第 " + hit.idx() + " 段：" + hit.title(), displaySnippetText(hit.content()), hit.fragmentId()))
             .toList();
         Map<String, DocResponse> docs = new LinkedHashMap<>();
         for (KnowledgeHit hit : evidenceHits) {
@@ -2272,6 +2272,20 @@ public class AihrSopSeedService {
             .replaceAll("[^\\p{IsHan}\\p{Alnum}]+", "");
     }
 
+    static String displaySnippetText(String text) {
+        String value = Optional.ofNullable(text).orElse("").strip();
+        if (!value.contains("【语音转写】")) {
+            return value;
+        }
+        return value
+            .replaceAll("[🎼♪♫♬]+", " ")
+            .replaceAll("(?iu)\\b(?:yeah|the|uh|um)\\b[.!。！？，,\\s]*", " ")
+            .replaceAll("[ぁ-ゟァ-ヿ]+[.!。！？，,\\s]*", " ")
+            .replaceAll("\\s+", " ")
+            .trim()
+            .replace("【语音转写】 ", "【语音转写】\n");
+    }
+
     static SearchResponse rawHitResponse(String category, String queryText, List<DocResponse> docs, List<SnippetResponse> snippets, List<String> extraCautions) {
         if (snippets == null || snippets.isEmpty()) {
             return noEvidenceResponse(queryText, category, extraCautions);

backend/ruoyi-modules/ruoyi-aihr/src/test/java/org/dromara/aihr/service/AihrSopSeedServiceTest.java

@@ -65,6 +65,21 @@ public class AihrSopSeedServiceTest {
         assertEquals(3L, evidence.get(2).fragmentId());
     }
 
+    @Test
+    @Tag("dev")
+    public void displaySnippetTextCleansVideoAsrNoise() {
+        String text = "【语音转写】\n[00:00] 🎼加油。🎼 The。Yeah.う。今天的汇报结束。";
+
+        String cleaned = AihrSopSeedService.displaySnippetText(text);
+
+        assertTrue(cleaned.contains("【语音转写】\n[00:00] 加油。"));
+        assertTrue(cleaned.contains("今天的汇报结束。"));
+        assertFalse(cleaned.contains("🎼"));
+        assertFalse(cleaned.contains("The"));
+        assertFalse(cleaned.contains("Yeah"));
+        assertFalse(cleaned.contains("う"));
+    }
+
     private static AihrSopSeedService.KnowledgeHit hit(Long fragmentId, String title) {
         return new AihrSopSeedService.KnowledgeHit(fragmentId, title, "sop", "", "doc-" + fragmentId, "片段内容", 1, 1.0);
     }