dataset_info:
features:
- name: text
dtype: string
- name: audio
dtype:
audio:
sampling_rate: 16000
- name: speaker_embeddings
sequence: float32
splits:
- name: train
num_bytes: 1726884990.125
num_examples: 11247
download_size: 1723089571
dataset_size: 1726884990.125
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
English Technical Speech Dataset
Overview
The English Technical Speech Dataset is a curated collection of English technical vocabulary recordings, designed for applications like Text-to-Speech (TTS), Automatic Speech Recognition (ASR), and Audio Classification. The dataset includes 11,247 entries and provides audio files, transcriptions, and speaker embeddings to support the development of robust technical language models.
- Language: English (technical focus)
- Total Entries: 11,247
- File Format: Parquet
- Sampling Rate: 16 kHz
Domain and Use Cases
Primary Domain: Technical Speech Processing
This dataset is ideal for use in:
- Text-to-Speech (TTS) Systems: Facilitating the generation of technical language audio.
- Automatic Speech Recognition (ASR): Improving transcription accuracy on technical vocabulary.
- Customer Support AI: Enhancing systems that recognize and respond to complex terminology.
Use Cases
- ASR for Technical Support: Optimized for recognizing industry-specific vocabulary in customer service.
- Educational Transcriptions: Useful for e-learning platforms focusing on technical material.
- Technical Support Tools: Enhances AI tools in areas such as IT help desks.
Data Structure
The dataset is stored in a Parquet file and has three main columns:
- audio: Contains the audio data recorded at a 16 kHz sampling rate.
- text: Transcriptions of the audio content.
- speaker_embeddings: Speaker embeddings generated with SpeechBrain's x-vector model for each audio file, providing vector representations of speaker characteristics.
Sample Data Structure
| Column | Description |
|---|---|
audio |
Audio file in 16 kHz WAV format |
text |
Text transcription of the corresponding audio |
speaker_embeddings |
Vectorized embeddings representing speaker identity |
Speaker Embeddings
Speaker embeddings were generated using the SpeechBrain x-vector model to capture speaker characteristics. This vector data is provided in the speaker_embeddings column and can be used for speaker identification or verification.
Getting Started
To load and work with this dataset, you can use the datasets library from Hugging Face:
from datasets import load_dataset
ds = load_dataset("Tejasva-Maurya/English-Technical-Speech-Dataset", split = "train")
Example Data
Each row in the dataset includes:
- Audio: WAV audio data with a 16 kHz sampling rate
- Text: Corresponding transcription for each audio sample
- Speaker Embedding: Vectorized representation of speaker identity
Dataset Composition and Sources
This dataset combines:
- Custom Audio Recordings: Self-recorded technical vocabulary, with the assistance of Saurabh Kumar for additional recordings.
- Open-Source Data from GitHub:Sample set (Pure-set) from
- Technical Speech Data from Hugging Face:
These sources contribute to the dataset’s focus on high-quality technical language audio and transcription accuracy.
License and Citation
This dataset is licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). If you use this dataset, please cite it.
Acknowledgments
Special thanks to Saurabh Kumar for assisting with custom audio recordings, and to AI4Bharat, Yassmen, and other contributors for their open-source datasets. This dataset is part of a larger effort to improve technical language understanding and processing in AI.