Enhancing Accessibility in Online Shopping: A Dataset and Summarization Method for Visually Impaired Individuals

Abstract

A visually impaired individual (VI) encounters numerous challenges in their daily activities, particularly in tasks reliant on visual systems such as navigation, educational pursuits, and shopping. Online shopping poses a heightened difficulty due to its reliance on visual representations of products in digital formats. The impact of visual impairment on product selection based on reviews remains inadequately investigated. This study endeavors to address two primary objectives. Firstly, we propose the creation of a dataset comprising product review videos (referred to as PVS10) tailored for visually impaired individuals. Secondly, we present a foundational summarization methodology designed to facilitate access to pertinent and informative content within extensive video collections for visually impaired individuals. Our dataset, gathered from YouTube, encompasses 10 distinct products, each associated with the top 10 review videos, totaling 100 videos of varying lengths. Utilizing the search term "review videos of PRODUCT NAME", we assembled the dataset to facilitate automated summarization processes aimed at maximizing salient information, minimizing redundant content, and preserving the overarching sentiment conveyed in the reviews. This research focuses on the challenges faced by visually impaired people in online shopping, particularly when selecting products based on customer reviews. Our study demonstrates that people with visual impairments may actively explore product reviews and only acquire the information they require.