Upload README.md with huggingface_hub

README.md

@@ -28,15 +28,16 @@ retrieved directly from YouTube.
 
 ## Dataset structure
 
-One row = one summary message for one video.
+One row = one summary message annotated by one annotator.
 
 | column | type | description |
 |---|---|---|
 | `video_id` | string | YouTube video ID |
 | `message` | string | Summary message text (English) |
-| `segments` | list of list of `[start, end]` | The cuts annotators marked as supporting this message. Outer list = one entry per annotator (3 annotators per message); inner list = that annotator's `[start, end]` timecode pairs (`HH:MM:SS:FF`, frame number — not converted to seconds since source frame rate varies). |
+| `annotator_index` | int | Index (0-based) of the annotator — 3 annotators per message |
+| `segments` | list of `[start, end]` | The cuts this annotator marked as supporting the message. `HH:MM:SS:FF` timecode (frame number — not converted to seconds since source frame rate varies). |
 
-- 60 videos, 180 messages (3 per video).
+- 60 videos, 180 messages (3 per video), 540 rows (3 annotators per message).
 
 ## License