Bangla-Bayanno: A 52K-Pair Bengali Visual Question Answering Dataset with LLM-Assisted Translation Refinement

2025-08-27

AI Summary

Existing Bengali Visual Question Answering (VQA) datasets suffer from limited human annotation, monolithic answer formats, poor translation quality, and the absence of high-quality open benchmarks—severely hindering low-resource multimodal research. To address these limitations, we introduce BanglaVQA, the first high-quality, open-source Bengali VQA benchmark, comprising 4,750 images and 52,650 question-answer pairs spanning nominal, quantitative, and yes/no question types. Our methodology employs a multilingual large language model–assisted translation and refinement pipeline, rigorously validated by human experts to ensure semantic fidelity and linguistic naturalness. We further propose a fine-grained answer-type taxonomy to support diverse evaluation protocols. BanglaVQA fills a critical gap in low-resource language VQA benchmarks and establishes a robust foundation for inclusive, multilingual multimodal AI research.

Technology Category

• Natural Language Processing: Question Answering
• Computer Vision: Language and Vision
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Community question answering

Abstract

In this paper, we introduce Bangla-Bayanno, an open-ended Visual Question Answering (VQA) Dataset in Bangla, a widely used, low-resource language in multimodal AI research. The majority of existing datasets are either manually annotated with an emphasis on a specific domain, query type, or answer type or are constrained by niche answer formats. In order to mitigate human-induced errors and guarantee lucidity, we implemented a multilingual LLM-assisted translation refinement pipeline. This dataset overcomes the issues of low-quality translations from multilingual sources. The dataset comprises 52,650 question-answer pairs across 4750+ images. Questions are classified into three distinct answer types: nominal (short descriptive), quantitative (numeric), and polar (yes/no). Bangla-Bayanno provides the most comprehensive open-source, high-quality VQA benchmark in Bangla, aiming to advance research in low-resource multimodal learning and facilitate the development of more inclusive AI systems.

Problem

• Creating a high-quality Bengali VQA dataset
• Overcoming low-quality multilingual translation issues
• Advancing low-resource multimodal AI research

Innovation

• LLM-assisted translation refinement pipeline
• Multilingual sources for dataset creation
• Three distinct answer type classifications