Accessible Structured Overview and Preview to Support Visually Impaired Users Exploration of Search Results

Abstract

Attempts to address visually impaired (VI) users’ challenges when viewing search results using a screen reader have been scarce. These challenges include navigating the search results to webpage accessibility and extracting useful information. This study investigates the effect of providing an overview of search results to VI users. It then seeks to boost this approach to form an overview and extract preview for each individual search result. This research presents a novel interactive search engine interface called InteractSE to support VI users during the results exploration stage and improve their interaction experience and web search efficiency. InteractSE aims to reduce the time and effort required by the VI users to browse or explore the search results. The proposed approach of InteractSE aims to generate high-level and multi-level summaries of the search results. Furthermore, the interface is designed to enable easy and interactive navigation through the search results. That interface allows users to explore the clustered set of search results. The search results summary is generated using an unsupervised machine learning approach to present the discovered concepts via a Formal Concept Analysis (FCA) approach, which is a domain-independent approach. These concepts are arranged in an accessible multi-level tree structure by following a hierarchical order and covering all retrieved documents that share specific features. FCA is employed in many information retrieval systems to support browsing search results. It has a high-performance score compared to other techniques. This approach highlights the hierarchized scenario to cluster the retrieved search results into a multi-level tree of overviews representing the main discovered ideas as concepts. According to the user selection for the clustered discovered concept as the main idea related to his/her query, this approach continues with the same scenario to generate an overview and preview summaries for each search result belonging to the select concept. Preliminary evaluation was conducted of the InteractSE interface with five sighted expert users. The experts reported the usability problems with their severity ratings against the heuristic of Nielsen’s set and Web Content Accessibility Guidelines (WCAG) 2.0. System Usability Scale (SUS) was used in this experiment, which is considered the most widely used standardized post-study usability questionnaire. The results show that the most frequently violated heuristics from Nielsen’s set are consistency and documentation, and the average severity rating of all the problems is minor. The results also show that the most frequently violated WCAG 2.0 guidelines are distinguishable, followed by navigable and affordable. The average severity rating of all the problems found using WCAG 2.0 guidelines is also minor. The average SUS post-study questionnaire score for InteractSE was 80 out of 100. Considering a benchmark of 68 defining a categorization of average and a threshold of 72 required for a good usability rating. The result for this study obtained a good usability rating score. Following the preliminary study, the InteractSE interface was evaluated by 16 VI users and compared with the Google search engine interface for complex search tasks. The evaluation results were obtained based on both quantitative (as task completion time) and qualitative (as participants’ feedback) measures. The results are promising and indicate that InteractSE enhances the search efficiency and consequently advances user experience. The observations and analysis obtained from the user interactions and feedback yielded design suggestions to support VI users when exploring and interacting with search results. Based on the feedback received from both studies, the summary approach was enhanced from providing the user with a high-level overview tree of the search results to providing the user with a webpage overview. That covers discovering the related items in that webpage and generating a webpage preview. The preview offers a quick review of specific discovered elements related to the search query on the website. InteractSE interface is designed to scrape each search result webpage and create a list of the discovered concepts. Each concept had its own summary based on the webpage content. The user would have different options to expand and navigate between other discovered concepts (overviews) or drill down for the same concept-generated summary (preview). In this dissertation, the second experiment was conducted with 35 VI users using the enhanced InteractSE interface. That experiment showed promising results when using the auto-generated clustered overview and preview summaries of the target web pages. These results support our findings of the first experiment and help expand the design suggestions. The results also guided us in outlining future improvements that target the summary integration with other automated technologies as conversation assistant and Chatbot.

Date of Award	2021
Original language	American English
Awarding Institution	HBKU College of Science and Engineering