The Speech Emotion Intensity Recognition Database (SEIR-DB) project aims to facilitate tasks related to speech emotion recognition and emotion intensity estimation. The database is comprehensive and multilingual, offering over 600,000 instances collected from various sources. It features languages like English, Russian, Mandarin, Greek, Italian, and French, making it highly diverse and versatile.

SEIR-DB provides an excellent resource for speech emotion recognition and emotion intensity estimation tasks. Each instance in the database is represented by crucial data fields like ID, WAV, EMOTION, INTENSITY, and LENGTH. The database is divided into training, testing, and validation sets, offering flexibility for various machine learning applications.

This project was an attempt to address the challenge of insufficient emotion data in speech emotion recognition (SER) experimentation. With SEIR-DB, researchers and developers have access to a large volume of cleanly formatted, emotion-annotated data for use in their work.

The creation of SEIR-DB involved meticulous data curation and processing from multiple sources. Each dataset was processed individually, with a focus on maintaining a balance in terms of the number of samples, emotion distribution, and language distribution. However, potential biases may still exist.

This project was curated by Gabriel Giangi from Concordia University. The SEIR-DB promises to significantly advance the research and development of speech emotion recognition technologies. Applications for this technology are vast, including mental health monitoring, virtual assistant enhancement, customer support, and communication aids for individuals with disabilities.

Source: https://huggingface.co/datasets/GDGiangi/SEIRDB

For more information and support about loading datasets from the HuggingFace API, please refer to the documentation.