Speech2Latex: 66,000 Audio Samples of Mathematical Expressions

66,000 human-annotated audio samples of spoken mathematical equations and sentences in English and Russian form the Speech2LaTeX dataset. It is the first fully open-source large-scale dataset for converting spoken math to LaTeX, drawn from diverse scientific domains. The dataset was created by marsianin500 and last updated on November 16, 2025.

Use Cases

• Training speech recognition models for mathematical notation based on annotated audio samples.
• Developing multimodal AI systems for scientific content understanding based on the dataset's focus.
• Benchmarking models for converting spoken English and Russian math into LaTeX code.
• Researching cross-lingual audio processing for technical domains based on the bilingual content.

Strengths

• Over 66,000 audio samples provide a substantial corpus for training.
• Human-annotated labels ensure high-quality ground truth for LaTeX conversion.
• Bilingual content in English and Russian supports multilingual model development.
• Drawn from diverse scientific domains, offering varied mathematical contexts.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last updated 2025-11-16 21:58:46; freshness should be verified.

Provenance

Source

huggingface

Collection Method

Human-annotated audio samples drawn from diverse scientific domains.

Freshness

Last updated 2025-11-16 21:58:46.