Turkish Speech Dataset with 41,427 Audio Segments and Emotion Labels

A merged speech dataset containing 41,427 audio segments from 88 original source datasets. The collection includes 222 speakers and features transcriptions and emotion labels for neutral, angry, sad, and happy speech. It was created by umutkkgz and last updated on Hugging Face in December 2025.

Use Cases

• Training Turkish automatic speech recognition (ASR) models based on the provided audio-text pairs.
• Developing speech emotion recognition systems based on the labeled neutral, angry, sad, and happy audio segments.
• Building speaker identification or verification models based on the unique speaker identifiers across 222 speakers.
• Creating or fine-tuning text-to-speech (TTS) models using the aligned audio and transcription data.

Strengths

• Contains 41,427 audio segments, providing a substantial volume of speech data.
• Merges content from 88 source datasets, potentially increasing diversity.
• Includes emotion labels (neutral, angry, sad, happy) for each audio segment.
• Covers 222 unique speakers, which may support speaker-diverse modeling.

Limitations

• Description metadata is limited; actual data quality, audio characteristics, and source dataset details require manual inspection after download.
• Column-level documentation is absent; field semantics and data structure must be inferred after download.

Provenance

Source

Hugging Face dataset created by author umutkkgz.

Collection Method

Merged from 88 source speech datasets.

Time Range

null

Freshness

Last updated 2025-12-26 22:25:19; freshness should be verified.

Geography

null