update_pause_between_segments

.gitignore

@@ -52,6 +52,7 @@ ENNI_transcript.py
 input/
 performance_eval/
 session_data/
-
+eval_for_SATE_v1_paper_step1.py
+eval_for_SATE_v1_paper_step2.py
 # Model
 CrisperWhisper_local/

pause.py

@@ -35,50 +35,35 @@ def annotate_pauses(session_id, threshold, base_dir="session_data"):
                     pauses.append(pause_info)
         segment["pauses"] = pauses
     
-    new_segments = []
-    
-    for i, segment in enumerate(segments):
-
-        new_segments.append(segment)
+    inter_segment_pauses = 0
+    for i in range(len(segments) - 1):
+        current_segment = segments[i]
+        next_segment = segments[i + 1]
         
-        if i < len(segments) - 1:
-            next_segment = segments[i + 1]
-            
-            current_words = segment.get("words", [])
-            next_words = next_segment.get("words", [])
+        current_words = current_segment.get("words", [])
+        next_words = next_segment.get("words", [])
+        
+        if current_words and next_words:
+            last_word_end = current_words[-1]["end"]
+            next_word_start = next_words[0]["start"]
+            gap = next_word_start - last_word_end
             
-            if current_words and next_words:
-                last_word_end = current_words[-1]["end"]
-                next_word_start = next_words[0]["start"]
-                gap = next_word_start - last_word_end
+            if gap > threshold:
+                inter_pause = {
+                    "start": round(last_word_end, 3),
+                    "end": round(next_word_start, 3),
+                    "duration": round(gap, 3)
+                }
                 
-                if gap > threshold:
-                    pause_segment = {
-                        "start": round(last_word_end, 3),
-                        "end": round(next_word_start, 3),
-                        "text": "",
-                        "speaker": "PAUSE",
-                        "words": [],
-                        "pauses": [
-                            {
-                                "start": round(last_word_end, 3),
-                                "end": round(next_word_start, 3),
-                                "duration": round(next_word_start - last_word_end, 3)
-                            }
-                        ]
-                    }
-                    new_segments.append(pause_segment)
-    
-    data["segments"] = new_segments
+                if "pauses" not in next_segment:
+                    next_segment["pauses"] = []
+                next_segment["pauses"].insert(0, inter_pause)
+                inter_segment_pauses += 1
     
     with open(json_file, "w", encoding="utf-8") as f:
         json.dump(data, f, ensure_ascii=False, indent=4)
     
     print(f"Session {session_id} pause annotation done: {json_file}")
-    print(f"Total segments after processing: {len(new_segments)}")
-    
-    pause_segments = [seg for seg in new_segments if seg.get("speaker") == "PAUSE"]
-    print(f"Added {len(pause_segments)} inter-sentence pause segments")
     
     return data