Datasets:

aihpi
/

FrWhisper-dataset

Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
client_id: string
path: string
audio: struct<path: string, array: list<element: float>, sampling_rate: int64>
  child 0, path: string
  child 1, array: list<element: float>
      child 0, element: float
  child 2, sampling_rate: int64
sentence: string
up_votes: int64
down_votes: int64
age: string
gender: string
accent: string
locale: string
segment: string
-- schema metadata --
huggingface: '{"info": {"features": {"client_id": {"dtype": "string", "_t' + 652
to
{'client_id': Value('string'), 'path': Value('string'), 'audio': {'path': Value('string'), 'array': List(Value('float32')), 'sampling_rate': Value('int64')}, 'sentence': Value('string'), 'age': Value('string'), 'gender': Value('string'), 'accent': Value('string'), 'locale': Value('string'), 'segment': Value('string')}
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 2431, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 1952, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 1975, 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 503, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 350, 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/parquet/parquet.py", line 106, in _generate_tables
                  yield f"{file_idx}_{batch_idx}", self._cast_table(pa_table)
                                                   ^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/parquet/parquet.py", line 73, 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
              client_id: string
              path: string
              audio: struct<path: string, array: list<element: float>, sampling_rate: int64>
                child 0, path: string
                child 1, array: list<element: float>
                    child 0, element: float
                child 2, sampling_rate: int64
              sentence: string
              up_votes: int64
              down_votes: int64
              age: string
              gender: string
              accent: string
              locale: string
              segment: string
              -- schema metadata --
              huggingface: '{"info": {"features": {"client_id": {"dtype": "string", "_t' + 652
              to
              {'client_id': Value('string'), 'path': Value('string'), 'audio': {'path': Value('string'), 'array': List(Value('float32')), 'sampling_rate': Value('int64')}, 'sentence': Value('string'), 'age': Value('string'), 'gender': Value('string'), 'accent': Value('string'), 'locale': Value('string'), 'segment': Value('string')}
              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.

FrWhisper Training Dataset

Dataset Description

This dataset contains the training data used to fine-tune the FrWhisper model, a specialized French speech recognition model optimized for conversational speech with interjections and hesitations.

The dataset combines two major French speech corpora:

LangAge Corpus: Demographically-structured conversational French data
ESLO (Enquêtes Sociolinguistiques à Orléans): French conversational data from specific age groups

Dataset Statistics

Total samples: 114,217
Training samples: 91,374
Test samples: 22,843
Language: French (fr)
Domain: Conversational speech
Audio format: 16kHz WAV files
Transcript format: Text with interjections and hesitations preserved

Features

The dataset includes the following fields:

client_id (anonymised)
path
audio
sentence
age
gender
accent
locale
segment

Special Characteristics

This dataset is specifically designed to capture:

Interjections: ah, euh, hé, hein, etc.
Hesitations: Natural speech disfluencies
Word repetitions: Stutters and false starts
Interrupted words: Partial utterances
Conversational markers: Discourse particles

Data Processing

The dataset has been processed with:

Audio resampling to 16kHz (Whisper requirement)
Removal of segments < 100ms
Filtering of silent audio segments
Exclusion of noise patterns like "(buzz)", "XXX"
Speaker filtering (excluding interviewers)

Intended Use

This dataset is intended for:

Training speech recognition models for conversational French
Research on interjections and hesitations in speech
Linguistic analysis of conversational patterns

Model Performance

When used to fine-tune Whisper Large V3, this dataset achieved:

14.18 percentage points improvement in WER over base model
67.21% WER on LangAge data (vs 78.43% for Whisper Large V3)
94.47% WER on ESLO data (vs 110.25% for Whisper Large V3)

Limitations

Primarily French from France (Orléans region and national data)
Conversational domain - may not generalize to formal speech
Contains natural speech disfluencies (intended feature)

Citation

If you use this dataset, please cite:

@misc{frwhisper-dataset2025,
  title={FrWhisper Training Dataset: French Conversational Speech with Interjections},
  author={Hanno Müller, Annette Gerstenberg},
  year={2025},
  note={Processed from LangAge and ESLO corpora}
}

Acknowledgments

The authors acknowledge the financial support by the German Federal Ministry of Research, Technology and Space (BMFTR) through the project «KI-Servicezentrum Berlin Brandenburg» (16IS22092).

Model Card Contact

For questions about this model, please open an issue in the model repository or contact [[email protected]].

Downloads last month: 29

Size of downloaded dataset files:

7.1 GB

Size of the auto-converted Parquet files:

7.1 GB

Number of rows:

114,217