Chatga-Boli: Bangla–Chittagonian Parallel Corpus

Overview

Chatga-Boli is a manually curated parallel corpus for machine translation between Standard Bangla and the Chittagonian dialect. The dataset is designed to support research on low-resource dialect translation and intra-language variation.

Languages

• Source: Standard Bangla
• Target: Chittagonian dialect

Dataset Size

Total sentence pairs: 2163

Train: 1730
Validation: 216
Test: 217

Data Sources

Sentences were collected from publicly available materials, including:

• Social media posts
• Drama scripts
• Song lyrics
• Short stories
• Informal written communication

Annotation Process

Chittagonian sentences were manually translated into Standard Bangla by two native speakers fluent in both varieties. Independent validation was performed by two additional native speakers to ensure linguistic accuracy and semantic fidelity.

Quality Control

• Manual translation
• Cross-checking by independent annotators
• Inter-annotator validation
• Removal of noisy or ambiguous samples

Intended Use

• Machine translation
• Dialect modeling
• Low-resource NLP research
• Sociolinguistic studies

Ethical Considerations

All content originates from publicly accessible sources. No personally identifiable information is included.

Availability

This dataset is currently private and will be publicly released upon publication of the associated research paper.

Contact

Mohammed Parves
Department of Computer Science and Engineering
Port City International University
Email: [mdperrfan@gmail.com]