E-commerce platforms generate extensive multi-modal data, including product descriptions, images, and customer reviews, which significantly influence consumer purchasing. However, discrepancies between seller claims and buyer experiences often lead to mistrust, dissatisfaction, and financial loss. Traditional e-commerce analytics approaches, including text-based sentiment analysis, standalone image classification, and rudimentary summarization, often fail to capture the complex interplay between modalities and therefore overlook nuanced discrepancies across textual and visual inputs. To address these limitations, we introduce ReViewQwen, a novel multimodal discrepancy detection and summarization framework leveraging the advanced Qwen2-VL Vision-Language Model. ReViewQwen integrates textual and visual inputs (product images from both buyer and seller) into a unified embedding space to systematically detect and contextualize discrepancies. Our comprehensive evaluation demonstrates that ReViewQwen outperforms state-of-the-art models such as LLaMA 3.2, Phi-3.5, and PaLiGemma 2, achieving superior precision, recall, and F1-score. Notably, the proposed system achieves accuracy improvement over the best-performing baseline from 51.15% to 88.00%. Additionally, our method promotes fairness in e-commerce review analytics by substantially reducing model biases and hallucinated content, thereby ensuring more trustworthy and balanced explanations. To access source code, data, and Prompts used https://github.com/domsoos/reviewqwen

Perusing web data items such as shopping products is a core online user activity. To prevent information overload, the content associated with data items is typically dispersed across multiple webpage sections over multiple web pages. However, such content distribution manifests an unintended side effect of significantly increasing the interaction burden for blind users, since navigating to-and-fro between different sections in different pages is tedious and cumbersome with their screen readers. While existing works have proposed methods for the context of a single webpage, solutions enabling usable access to content distributed across multiple webpages are few and far between. In this paper, we present InstaFetch, a browser extension that dynamically generates an alternative screen reader-friendly user interface in real-time, which blind users can leverage to almost instantly access different item-related information such as description, full specification, and user reviews, all in one place, without having to tediously navigate to different sections in different webpages. Moreover, InstaFetch also supports natural language queries about any item, a feature blind users can exploit to quickly obtain desired information, thereby avoiding manually trudging through reams of text. In a study with 14 blind users, we observed that the participants needed significantly lesser time to peruse data items with InstaFetch, than with a state-of-the-art solution.

Coordinating viewpoints with another person during a collaborative task can provide informative cues on human behavior. Despite the massive shift of collaborative spaces into virtual environments, versatile setups that enable eye-tracking in an online collaborative environment (distributed eye-tracking) remain unexplored. In this study, we present DisETrac- a versatile setup for eye-tracking in online collaborations. Further, we demonstrate and evaluate the utility of DisETrac through a user study. Finally, we discuss the implications of our results for future improvements. Our results indicate promising avenue for developing versatile setups for distributed eye-tracking.

Web data items such as shopping products, classifieds, and job listings are indispensable components of most e-commerce websites. The information on the data items are typically distributed over two or more webpages, e.g., a ‘Query-Results’ page showing the summaries of the items, and ‘Details’ pages containing full information about the items. While this organization of data mitigates information overload and visual cluttering for sighted users, it however increases the interaction overhead and effort for blind users, as back-and-forth navigation between webpages using screen reader assistive technology is tedious and cumbersome. Existing usability-enhancing solutions are unable to provide adequate support in this regard as they predominantly focus on enabling efficient content access within a single webpage, and as such are not tailored for content distributed across multiple webpages. As an initial step towards addressing this issue, we developed AutoDesc, a browser extension that leverages a custom extraction model to automatically detect and pull out additional item descriptions from the ‘details’ pages, and then proactively inject the extracted information into the ‘Query-Results’ page, thereby reducing the amount of back-and-forth screen reader navigation between the two webpages. In a study with 16 blind users, we observed that within the same time duration, the participants were able to peruse significantly more data items on average with AutoDesc, compared to that with their preferred screen readers as well as with a state-of-the-art solution.

Online discussion forums have become an integral component of news, entertainment, information, and video-streaming websites, where people all over the world actively engage in discussions on a wide range of topics including politics, sports, music, business, health, and world affairs. Yet, little is known about their usability for blind users, who aurally interact with the forum conversations using screen reader assistive technology. In an interview study, blind users stated that they often had an arduous and frustrating interaction experience while consuming conversation threads, mainly due to the highly redundant content and the absence of customization options to selectively view portions of the conversations. As an initial step towards addressing these usability concerns, we designed PView - a browser extension that enables blind users to customize the content of forum threads in real time as they interact with these threads. Specifically, PView allows the blind users to explicitly hide any post that is irrelevant to them, and then PView automatically detects and filters out all subsequent posts that are substantially similar to the hidden post in real time, before the users navigate to those portions of the thread. In a user study with blind participants, we observed that compared to the status quo, PView significantly improved the usability, workload, and satisfaction of the participants while interacting with the forums.