Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
video: string
duration_sec: double
mode: string
total_utterances: int64
speakers: list<item: string>
  child 0, item: string
person_id_database: struct<user: string, matthew: string, jason: string>
  child 0, user: string
  child 1, matthew: string
  child 2, jason: string
num_clips_processed: int64
utterances_by_speaker: struct<User: list<item: struct<person_id: string, name: string, text: string, start_sec: double, end (... 435 chars omitted)
  child 0, User: list<item: struct<person_id: string, name: string, text: string, start_sec: double, end_sec: double, (... 67 chars omitted)
      child 0, item: struct<person_id: string, name: string, text: string, start_sec: double, end_sec: double, confidence (... 55 chars omitted)
          child 0, person_id: string
          child 1, name: string
          child 2, text: string
          child 3, start_sec: double
          child 4, end_sec: double
          child 5, confidence: double
          child 6, source: string
          child 7, gemini_speaker_label: string
  child 1, Matthew: list<item: struct<person_id: string, name: string, text: string, start_sec: double, end_sec: double, (... 67 chars omitted)
      child 0, item: struct<person_id: string, name: string, text: string, start_sec: double, end_sec: double, confidence (... 55 chars omitted)
          child 0, person_id: string
          child 1, name: string
          child 2, text: string
          child 3, start_sec: double
          child 4, end_sec: double
          
...
ing, text: string, start_sec: double, end_sec: double, confidence (... 55 chars omitted)
      child 0, person_id: string
      child 1, name: string
      child 2, text: string
      child 3, start_sec: double
      child 4, end_sec: double
      child 5, confidence: double
      child 6, source: string
      child 7, gemini_speaker_label: string
latency_gemini_call_ms: double
diarized_utterances: list<item: struct<speaker_label: string, text: string, start_sec: double, end_sec: double, confidenc (... 11 chars omitted)
  child 0, item: struct<speaker_label: string, text: string, start_sec: double, end_sec: double, confidence: double>
      child 0, speaker_label: string
      child 1, text: string
      child 2, start_sec: double
      child 3, end_sec: double
      child 4, confidence: double
clip_start_sec: double
clip_end_sec: double
attributed_utterances: list<item: struct<person_id: string, name: string, text: string, start_sec: double, end_sec: double, (... 67 chars omitted)
  child 0, item: struct<person_id: string, name: string, text: string, start_sec: double, end_sec: double, confidence (... 55 chars omitted)
      child 0, person_id: string
      child 1, name: string
      child 2, text: string
      child 3, start_sec: double
      child 4, end_sec: double
      child 5, confidence: double
      child 6, source: string
      child 7, gemini_speaker_label: string
latency_total_ms: double
latency_fuse_ms: double
clip_index: int64
latency_name_extraction_ms: double
to
{'clip_index': Value('int64'), 'clip_start_sec': Value('float64'), 'clip_end_sec': Value('float64'), 'diarized_utterances': List({'speaker_label': Value('string'), 'text': Value('string'), 'start_sec': Value('float64'), 'end_sec': Value('float64'), 'confidence': Value('float64')}), 'attributed_utterances': List({'person_id': Value('string'), 'name': Value('string'), 'text': Value('string'), 'start_sec': Value('float64'), 'end_sec': Value('float64'), 'confidence': Value('float64'), 'source': Value('string'), 'gemini_speaker_label': Value('string')}), 'latency_gemini_call_ms': Value('float64'), 'latency_fuse_ms': Value('float64'), 'latency_name_extraction_ms': Value('float64'), 'latency_total_ms': Value('float64')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 99, in get_rows_or_raise
                  return get_rows(
                         ^^^^^^^^^
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                         ^^^^^^^^^^^^^^^^^^^^^
                File "/src/services/worker/src/worker/utils.py", line 77, in get_rows
                  rows_plus_one = list(itertools.islice(ds, rows_max_number + 1))
                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2690, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2227, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2251, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 494, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 384, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 289, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 124, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2272, in table_cast
                  return cast_table_to_schema(table, schema)
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2218, in cast_table_to_schema
                  raise CastError(
              datasets.table.CastError: Couldn't cast
              video: string
              duration_sec: double
              mode: string
              total_utterances: int64
              speakers: list<item: string>
                child 0, item: string
              person_id_database: struct<user: string, matthew: string, jason: string>
                child 0, user: string
                child 1, matthew: string
                child 2, jason: string
              num_clips_processed: int64
              utterances_by_speaker: struct<User: list<item: struct<person_id: string, name: string, text: string, start_sec: double, end (... 435 chars omitted)
                child 0, User: list<item: struct<person_id: string, name: string, text: string, start_sec: double, end_sec: double, (... 67 chars omitted)
                    child 0, item: struct<person_id: string, name: string, text: string, start_sec: double, end_sec: double, confidence (... 55 chars omitted)
                        child 0, person_id: string
                        child 1, name: string
                        child 2, text: string
                        child 3, start_sec: double
                        child 4, end_sec: double
                        child 5, confidence: double
                        child 6, source: string
                        child 7, gemini_speaker_label: string
                child 1, Matthew: list<item: struct<person_id: string, name: string, text: string, start_sec: double, end_sec: double, (... 67 chars omitted)
                    child 0, item: struct<person_id: string, name: string, text: string, start_sec: double, end_sec: double, confidence (... 55 chars omitted)
                        child 0, person_id: string
                        child 1, name: string
                        child 2, text: string
                        child 3, start_sec: double
                        child 4, end_sec: double
                        
              ...
              ing, text: string, start_sec: double, end_sec: double, confidence (... 55 chars omitted)
                    child 0, person_id: string
                    child 1, name: string
                    child 2, text: string
                    child 3, start_sec: double
                    child 4, end_sec: double
                    child 5, confidence: double
                    child 6, source: string
                    child 7, gemini_speaker_label: string
              latency_gemini_call_ms: double
              diarized_utterances: list<item: struct<speaker_label: string, text: string, start_sec: double, end_sec: double, confidenc (... 11 chars omitted)
                child 0, item: struct<speaker_label: string, text: string, start_sec: double, end_sec: double, confidence: double>
                    child 0, speaker_label: string
                    child 1, text: string
                    child 2, start_sec: double
                    child 3, end_sec: double
                    child 4, confidence: double
              clip_start_sec: double
              clip_end_sec: double
              attributed_utterances: list<item: struct<person_id: string, name: string, text: string, start_sec: double, end_sec: double, (... 67 chars omitted)
                child 0, item: struct<person_id: string, name: string, text: string, start_sec: double, end_sec: double, confidence (... 55 chars omitted)
                    child 0, person_id: string
                    child 1, name: string
                    child 2, text: string
                    child 3, start_sec: double
                    child 4, end_sec: double
                    child 5, confidence: double
                    child 6, source: string
                    child 7, gemini_speaker_label: string
              latency_total_ms: double
              latency_fuse_ms: double
              clip_index: int64
              latency_name_extraction_ms: double
              to
              {'clip_index': Value('int64'), 'clip_start_sec': Value('float64'), 'clip_end_sec': Value('float64'), 'diarized_utterances': List({'speaker_label': Value('string'), 'text': Value('string'), 'start_sec': Value('float64'), 'end_sec': Value('float64'), 'confidence': Value('float64')}), 'attributed_utterances': List({'person_id': Value('string'), 'name': Value('string'), 'text': Value('string'), 'start_sec': Value('float64'), 'end_sec': Value('float64'), 'confidence': Value('float64'), 'source': Value('string'), 'gemini_speaker_label': Value('string')}), 'latency_gemini_call_ms': Value('float64'), 'latency_fuse_ms': Value('float64'), 'latency_name_extraction_ms': Value('float64'), 'latency_total_ms': Value('float64')}
              because column names don't match

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

4_6_2026_wsw_run2

Who-Said-What diarization (ONLINE mode) on jason_matt_h264.mp4.

Online Mode Features

  1. Egocentric speaker detection — Gemini prompt explicitly handles AR glasses perspective. The wearer is labeled as "User" and is never visible in keyframes.
  2. Name extraction from context — When someone says "my name is X", the system learns their name and retroactively updates all previous utterances from the same speaker.
  3. Person ID database — Learned names persist across clips. Database cleared between evaluations.

Summary

  • Video: jason_matt_h264.mp4 (115.4 sec)
  • Clips processed: 9
  • Total utterances: 37
  • Speakers detected: User, Matthew, Jason
  • Person ID database: {"user": "User", "matthew": "Matthew", "jason": "Jason"}
  • Avg Gemini latency: 3358 ms/clip

Full Transcript

  [0.0s - 0.8s] User: "And then begin the conversation."
  [1.2s - 1.6s] Matthew: "Okay."
  [1.9s - 4.4s] Matthew: "My name is Matthew."
  [4.9s - 6.6s] Matthew: "I am a jig student."
  [7.2s - 9.6s] Matthew: "And I'm going to graduate this March."
  [10.5s - 10.9s] Jason: "Uh my name is Jason."
  [11.6s - 12.0s] Jason: "And then."
  [12.3s - 13.8s] Jason: "I'm a student of"
  [26.0s - 32.5s] Matthew: "big, um, they figured out a lot to be in the military."
  [33.1s - 34.9s] Matthew: "I had an open eye, I signed it the next day."
  [35.4s - 36.3s] Matthew: "So what is your opinion on this?"
  [37.0s - 38.1s] Jason: "We probably want to get closer."
  [38.1s - 39.6s] Jason: "It's too wide for now."
  [52.3s - 53.6s] Matthew: "I think that it's normal, like like."
  [53.6s - 55.3s] Matthew: "Like I feel like that's what the government would ask you to do."
  [55.4s - 55.5s] Jason: "Yeah."
  [56.0s - 62.9s] Jason: "Oh, so they would probably prevent or any any private company from getting too much."
  [65.2s - 72.7s] Matthew: "any private company from getting too much power in the military because they would want to have more control over national security and defense."
  [72.8s - 72.9s] Jason: "I."
  [72.9s - 73.5s] Jason: "Yes, that's right."
  [73.8s - 74.4s] Jason: "The US government."
  [74.4s - 74.8s] Matthew: "Yeah, yeah, yeah."
  [76.0s - 79.8s] Matthew: "I mean, I feel like if you are a"
  [79.4s - 84.7s] Matthew: "I feel like if you are allowed to use AI at work, they can just use AI, they don't want to get it."
  [85.5s - 86.1s] Jason: "So do you think any government officials use AI currently?"
  [86.8s - 87.7s] Matthew: "Yeah, yeah."
  [88.2s - 92.6s] Matthew: "It's it's used by I think they use quantum."
  [91.2s - 92.4s] Matthew: "I think they use graph at all right, that's what I said."
  [92.5s - 92.9s] Jason: "Oh, they use graph."
  [93.1s - 95.3s] Matthew: "So on using graph in the workplace, yeah."
  [95.4s - 98.5s] Jason: "So graph's able to see all the confidential public information."
  [98.5s - 100.5s] Matthew: "Yeah, yeah, I'll probably provision."
  [100.7s - 104.9s] Matthew: "Because they're the partner and then they and then now."
  [104.2s - 111.6s] Matthew: "the partner and then they and then now with the that they don't want to continue."
  [111.6s - 112.9s] Matthew: "Like cuz the government workers are probably using."
  [112.9s - 115.5s] Jason: "Oh, so they they probably they probably used anthropic."
  [115.5s - 116.7s] Jason: "there's probably there's probably."

Files

File Description
diarization_results.json Full results with named speakers + person ID database
clips_detail.jsonl Per-clip diarized + attributed utterances
latency_stats.json Per-clip latency (Gemini, fusion, name extraction)
Downloads last month
22