Dataset for the recognition of Kurdish sound dialects

Area of Science:

• Speech Analysis
• Computational Linguistics
• Phonetics

Background:

• Speech recognition systems are negatively affected by speaker variations including age, gender, and dialect.
• Dialect recognition systems (DRS) can be integrated into speech recognition to mitigate performance issues caused by dialectal differences.
• The Kurdish language, with its distinct dialects, lacks dedicated datasets for automatic dialect recognition.

Purpose of the Study:

• To address the scarcity of resources for Kurdish dialect recognition.
• To introduce and evaluate a novel dataset for automatic dialect recognition of Kurdish.
• To facilitate the development of improved speech recognition models for the Kurdish language.

Main Methods:

• A dataset was compiled using experimental data collected by the Computer Science Department at the University of Halabja.
• The dataset includes samples from three main Kurdish dialects: Northern Kurdish (Badini), Central Kurdish (Sorani), and Hawrami.
• The dataset's suitability for developing automatic dialect recognition systems was assessed.

Main Results:

• A new dataset for Kurdish dialect recognition has been created and assessed.
• The dataset encompasses three primary dialects of the Kurdish language, providing a foundation for research.
• The availability of this dataset is a significant step towards enhancing Kurdish speech recognition technology.

Conclusions:

• The developed dataset is essential for advancing automatic dialect recognition for the Kurdish language.
• This resource will enable the creation of more accurate and robust speech recognition systems tailored to Kurdish speakers.
• Further research and development in Kurdish speech processing can now be pursued with this foundational dataset.