Dataset on multiregional variations of Bangla language (BD-Dialect)

Area of Science:

• Computational Linguistics
• Dialectology
• Natural Language Processing (NLP)

Background:

• Limited availability of comprehensive linguistic resources for Bangla dialects.
• Need for aligned data to support cross-dialectal research and NLP model development.
• Existing resources often lack regional authenticity or broad dialectal coverage.

Purpose of the Study:

• To introduce the BD-Dialect dataset, a novel multiregional linguistic resource for Bangla.
• To provide aligned translations between Standard Bangla and five major regional dialects (Noakhali, Sylhet, Chittagong, Rajshahi, Mymensingh) with English.
• To facilitate research in computational linguistics, dialectology, and NLP applications for Bangla.

Main Methods:

• Data collection through native speaker interviews and regional literature.
• Creation of a dataset comprising two CSV files (words and clauses) with 950 aligned entries each.
• Preprocessing and validation using Python in Google Colab and review by native speakers.

Main Results:

• A comprehensive dataset of aligned Bangla dialectal data is now available.
• The dataset includes Standard Bangla, five dialects, and English translations across 950 entries.
• Preprocessing scripts and the dataset are publicly accessible via Mendeley Data.

Conclusions:

• The BD-Dialect dataset serves as a foundational resource for Bangla NLP and linguistic studies.
• It enables the development of dialect recognition models and comparative linguistic research.
• The open-access nature promotes further research and innovation in Bangla language technology.