Datasets:

IParraMartin
/

TIMITPhones

dataset_info:
  features:
    - name: audio
      dtype: audio
    - name: phoneme
      dtype: string
    - name: phoneme_label
      dtype: int64
    - name: category
      dtype: string
    - name: category_label
      dtype: int64
  splits:
    - name: train
      num_bytes: 482969101.786
      num_examples: 59527
  download_size: 205608801
  dataset_size: 482969101.786
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

TIMITPhones: TIMIT Phoneme Dataset

This corpus is a phoneme‑level derivative of the original TIMIT Acoustic‑Phonetic Continuous Speech Corpus.
Each entry pairs a 1‑second waveform excerpt with a single phoneme label taken from the 61‑phone TIMIT inventory (the mapping to 39‑phone and broad‑class sets is also provided).

This version is designed for quick prototyping of phoneme classifiers or probing acoustic representations.

Supported Tasks and Leaderboards

Automatic Speech Recognition (phoneme classification)
Representation Probing (e.g. evaluate SSL acoustic models)

Languages

American English (predominantly GenAm, multiple dialects represented).

Source Data

The source material is the licensed TIMIT corpus (LDC93S1). Waveforms were segmented to align with individual phone boundaries using the provided .PHN time‑marks.

Annotations

Phoneme boundaries originate from the manual TIMIT annotations. No additional human annotation was performed.

Authors

@IParraMartin

Citation

@misc{parra2025timitphones,
  title   = {TIMIT Phones (Trimmed): A Phoneme‑level Corpus for Speech Representation Probing},
  author  = {Parra, Iñigo},
  year    = 2025,
  url     = {https://huggingface.co/datasets/<YOUR_USERNAME>/<REPO_NAME>},
  note    = {Version 1.0.0}
}

References

[1] Garofolo, John S., et al. TIMIT Acoustic-Phonetic Continuous Speech Corpus LDC93S1. Web Download. Philadelphia: Linguistic Data Consortium, 1993.