fix transcribe settings

app/services/processor.py

@@ -284,6 +284,7 @@ class Processor:
                 )
             ]
 
+        processed_segments = cls._filter_segments_with_context(processed_segments)
         processing_time = time.time() - t0
         
         txt_content = cls._generate_txt(
@@ -306,6 +307,53 @@ class Processor:
             txt_content=txt_content,
             csv_content=csv_content,
         )
+        
+        
+    @staticmethod
+    def _is_meaningful_segment(
+        seg: TranscriptSegment,
+        min_duration_s: float = 0.6,
+        min_words: int = 3
+    ) -> bool:
+        duration = seg.end - seg.start
+        word_count = len(seg.text.split())
+
+        if duration >= min_duration_s:
+            return True
+        if word_count >= min_words:
+            return True
+        if seg.role == "KH":
+            return True
+
+        return False
+
+    @classmethod
+    def _filter_segments_with_context(
+        cls,
+        segments: List[TranscriptSegment]
+    ) -> List[TranscriptSegment]:
+        if not segments:
+            return segments
+        
+        segments = sorted(segments, key=lambda s: s.start)
+        result = []
+        n = len(segments)
+
+        for i, seg in enumerate(segments):
+            prev_seg = segments[i - 1] if i > 0 else None
+            next_seg = segments[i + 1] if i < n - 1 else None
+
+            if cls._is_meaningful_segment(seg):
+                result.append(seg)
+                continue
+
+            if prev_seg and next_seg:
+                if prev_seg.speaker == seg.speaker == next_seg.speaker:
+                    result.append(seg)
+
+        return result
+    
+        
     @classmethod
     def _generate_txt(
             cls,