Year: 2025
Type: article
Abstract: With the rapid growth of internet access, Bangladeshi social media users have been observed to become increasingly active, with Facebook having been established as the primary platform for communication and content sharing.Among various content types, food reviews have been recognized as gaining significant popularity, particularly through video-based posts created by influencers.In this study, the sentiment expressed in Bengali food reviews on Facebook has been investigated through a dual approach: interactionbased sentiment analysis has been performed using Facebook reactions, and sequence classification has been carried out using BanglaBERT, a state-of-the-art Bengali language model.Data have been collected from 905 food review videos and 26,004 associated comments that have been posted between 2020 and 2022.The BanglaBERT model has been fine-tuned on multiple Bengali sentiment datasets, and a prediction accuracy of 83.76% has been achieved, demonstrating its ability to capture nuanced patterns in user opinions.It has been found that the majority of viewers have expressed positive sentiment towards food-related content, although a small subset of comments has reflected contrasting negative opinions.These insights have highlighted the strong engagement value of food reviews in Bangladeshi social media, while the presence of subtle critical perspectives has also been emphasized.A deeper understanding of user perception in this domain has been contributed by the study, and practical implications for content creators, businesses, and researchers in natural language processing have been offered.
Authors: Ataf Fazledin Ahamed, Sadia Sharmin
Institution: Bangladesh University of Engineering and Technology
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4413888106