Upload README.md with huggingface_hub

README.md

@@ -16,5 +16,5 @@ Each row carries the original item metadata plus:
 - `transcript_text`: full transcript
 - `language_code` / `language_probability`
 - `num_words`, `num_speakers`, `num_audio_events`
-- `words_json`: full Scribe v2 word array (with `start`, `end`, `speaker_id`,
-  `type`, `logprob`) as a JSON string — `json.loads()` to expand
+- `words`: full Scribe v2 word array — list of `{text, start, end, type,
+  speaker_id, logprob}` structs