BanGRev (Bengali Gadget Review) Dataset

Bengali ranks seventh globally among spoken languages yet faces severe constraints in computational text analysis resources. Bangladesh's e-commerce sector has expanded rapidly, with internet users reaching 121.75 million by December 2023, making online consumer reviews increasingly influential in purchasing decisions. Studies indicate that 91% of consumers rely on such reviews when making product choices. The primary obstacle to automated review analysis in Bengali has been dataset scarcity. Most deep learning approaches require at least 8,000 samples for adequate performance. Beyond scale limitations, previous datasets relied on single-platform collection, missing the linguistic variation across different user demographics and communication contexts. Bengali presents particular challenges for computational processing. The language's complex morphology and regional dialectal variations complicate automated analysis. Tech product reviews add another layer of complexity through specialized vocabulary, technical specifications, and product-specific evaluation criteria. BanGRev addresses these gaps through systematic multi-source collection from e-commerce platforms, social media, and video comments, combined with rigorous three-class annotation protocols validated at Fleiss' Kappa of 0.73.