ajd12342/paraspeechcaps-situational-train

ParaSpeechCaps Situational Training Dataset

Training dataset for the ParaSpeechCLAP-Situational and ParaSpeechCLAP-Combined models, from the paper:

ParaSpeechCLAP: A Dual-Encoder Speech-Text Model for Rich Stylistic Language-Audio Pretraining Anuj Diwan, Eunsol Choi, David Harwath Under review

This dataset contains the situational-tag subset of ParaSpeechCaps, filtered to include only examples annotated with situational (utterance-level) style tags. It is used to train the ParaSpeechCLAP-Situational and ParaSpeechCLAP-Combined models with a contrastive loss.

Installation

Install the datasets package to load the dataset:

pip install datasets

To train ParaSpeechCLAP models using this dataset, install the ParaSpeechCLAP GitHub repository:

git clone https://github.com/ajd12342/paraspeechclap.git
cd paraspeechclap
pip install -r requirements.txt

Setting up audio files

The dataset contains a relative_audio_path column but not the audio files themselves. Resolving audio paths requires specifying data.audio_root, a common root directory organized as ${audio_root}/{source}/, where {source} matches the value of the source column in the dataset.

This dataset uses the Expresso and EARS sources. Follow the ParaSpeechCaps audio setup instructions for those sources, with the following adjustment: instead of placing each source at its own root directory, place them under a common root:

• ${audio_root}/expresso/ (instead of ${expresso_root})
• ${audio_root}/ears/ (instead of ${ears_root})

Then pass data.audio_root=${audio_root} when running any ParaSpeechCLAP script.

Usage with ParaSpeechCLAP

Training

torchrun --nproc_per_node=4 scripts/train.py \
  --config-name train/situational \
  data.audio_root=/path/to/audio_root \
  meta.results=./experiments

Loading the dataset

from datasets import load_dataset

dataset = load_dataset("ajd12342/paraspeechcaps-situational-train", split="train")
print(f"Number of examples: {len(dataset)}")
print(dataset[0])

Related Resources

• GitHub Repository: https://github.com/ajd12342/paraspeechclap
• Models: ajd12342/paraspeechclap-intrinsic, ajd12342/paraspeechclap-situational and ajd12342/paraspeechclap-combined
• Parent Dataset: https://huggingface.co/datasets/ajd12342/paraspeechcaps
• Training Datasets: https://huggingface.co/datasets/ajd12342/paraspeechcaps-intrinsic-train and https://huggingface.co/datasets/ajd12342/paraspeechcaps-situational-train
• Evaluation Datasets: https://huggingface.co/datasets/ajd12342/paraspeechclap-eval-intrinsic, https://huggingface.co/datasets/ajd12342/paraspeechclap-eval-situational and https://huggingface.co/datasets/ajd12342/paraspeechclap-eval-combined

Citation

@misc{diwan2026paraspeechclapdualencoderspeechtextmodel,
      title={ParaSpeechCLAP: A Dual-Encoder Speech-Text Model for Rich Stylistic Language-Audio Pretraining},
      author={Anuj Diwan and Eunsol Choi and David Harwath},
      year={2026},
      eprint={2603.28737},
      archivePrefix={arXiv},
      primaryClass={eess.AS},
      url={https://arxiv.org/abs/2603.28737},
}