Speech2Latex Dataset

The Speech2LaTeX dataset is the first fully open-source large-scale dataset for converting spoken mathematical expressions and sentences into LaTeX. It comprises over 66,000 human-annotated audio samples of mathematical equations and sentences in both English and Russian, drawn from diverse scientific domains. This work lays the groundwork for future advances in multimodal AI, with a particular focus on mathematical content recognition.

The dataset was presented in the paper: Speech-to-LaTeX: New Models and Datasets for Converting Spoken Equations and Sentences.

Installation

pip install -U datasets==3.0.2 huggingface_hub==0.34.3

Optionally install librosa

ffmpeg==1.4
librosa==0.11.0

To convert audio to tensors, the preferred requirements are:

torch==2.5.1
torchaudio==2.5.1

to avoid issues with a torchcodec using newer versions of libraries.

Run in Python

from datasets import load_dataset
ds = load_dataset("marsianin500/Speech2Latex")
ds

Additional data is presented at https://huggingface.co/datasets/marsianin500/Speech2LatexMathBridge.