directtt/polish_presidential_debate

Polish Presidential ASR Dataset

Public domain (2025)
Source: DEBATA PREZYDENCKA TVP | 12.05.2025

Description

Studio-recorded utterances by 13 Polish presidential debate candidates, each providing 15 audio samples (read or spontaneous). Audio is in 16 kHz WAV format. The dataset is designed for automatic speech recognition (ASR) tasks, particularly in the context of Polish language processing in political domain.

File Structure

├── README.md
├── audio/
│   └── pl/
│       └── <NN-name>/
│           ├── <name>_0.wav
│           ├── <name>_1.wav
│           └── … (15 files per speaker)
└── transcript/
    └── pl/
        └── test.tsv

Each speaker folder is named with a two-digit prefix and speaker identifier (e.g. 01-bartoszewicz_artur).

Metadata (test.tsv)

TSV headers:

• path - relative path to the audio file
• sentence - spoken text
• age - speaker's age
• gender - speaker's gender
• locale - locale of the recording (e.g., "studio")
• speech_type - type of speech (e.g., "read" or "spontaneous")
• source - source of the audio (e.g., "youtube")

Each row corresponds to one WAV file under audio/pl/..., with transcript and speaker metadata.

Recordings duration per candidate

Candidate ID	Name	Total Duration
01	Bartoszewicz Artur	0h 5m 19s
02	Biejat Magdalena	0h 5m 4s
03	Braun Grzegorz	0h 4m 55s
04	Hołownia Szymon	0h 3m 0s
05	Jakubiak Marek	0h 3m 4s
06	Maciak Maciej	0h 3m 0s
07	Mentzen Sławomir	0h 3m 21s
08	Nawrocki Karol	0h 3m 10s
09	Senyszyn Joanna	0h 3m 10s
10	Stanowski Krzysztof	0h 3m 33s
11	Trzaskowski Rafał	0h 3m 39s
12	Woch Marek	0h 3m 19s
13	Zandberg Adrian	0h 2m 6s

Usage

You can load the dataset using the huggingface datasets library. To download the entire dataset, use the following code:

from datasets import load_dataset

ds = load_dataset("directtt/polish_presidential_debate", split="test")
print(ds[0])
# {
#   'path': 'audio/pl/01-bartoszewicz_artur/bartoszewicz_artur_0.wav',
#   'audio': {...},
#   'sentence': '...',
#   'age': 51,
#   'gender': 'male',
#   ...
# }

Using the datasets library, you can also stream the dataset on-the-fly by adding a streaming=True argument to the load_dataset function call. Loading a dataset in streaming mode loads individual samples of the dataset at a time, rather than downloading the entire dataset to disk.

from datasets import load_dataset

ds = load_dataset("directtt/polish_presidential_debate", split="test", streaming=True)

print(next(iter(ds)))

License

Public domain, Apache-2.0 license.

Citation

If you use this dataset in your research, please cite it as follows:

@misc{polish_presidential_debate,
  title = {Polish Presidential Debate ASR Dataset},
  author = {Dudek Marcel, Jerzykiewicz Sebastian, Rybczyński Jędrzej, Solarski Antoni},
  year = {2025}
}