ChartMaster: A Tool for Interacting with Stock Market Charts using a Screen Reader
Hong Zou, Jutta Treviranus · 2015 · ASSETS '15: Proceedings of the 17th International ACM SIGACCESS Conference on Computers & Accessibility · doi:10.1145/2700648.2809862
Summary
This paper presents ChartMaster, a tool enabling screen reader users to query interactive stock market charts, developed through an inclusive co-design process with visually impaired participants. The authors frame the problem in economic terms: 1.3 billion people with disabilities globally control $8 trillion in annual disposable income, yet are disadvantaged in using online investment instruments due to inaccessible interfaces. Stock market charts have evolved from static images to interactive, real-time visualizations—powerful for sighted users but largely inaccessible via screen readers. A manual usability inspection of ten major financial websites (Yahoo Finance, Google Finance, Bloomberg, Wall Street Journal, and others) against WCAG 2.0 principles found comprehensive failures. Charts were not perceivable (screen readers detected only 4 of 10 charts; 9 of 10 lacked meaningful text alternatives). They were not operable (hover interactions impossible without mouse). They were not understandable (abbreviations read as unrelated words; data sequences without pauses were incomprehensible). They were not robust (none worked reliably with screen readers). Even numeric data tables—WCAG's recommended alternative—proved "very time consuming" with "enormous cognitive effort," especially for comparison tasks. The study employed four iterative design rounds with 18 participants (16 with visual impairments), most using screen readers, testing four interface approaches: numeric tables, audio input/output (voice queries like Siri), text input/output (typed queries like IBM Watson), and dropdown menus. The Wizard of Oz method allowed rapid evaluation before full implementation. Sessions occurred primarily at participants' homes, capturing real-world usage contexts.
Key findings
The dropdown menu solution received overwhelmingly positive responses. Participants were "thrilled"—"This is so cool! I can't believe that nobody has developed something like this before!" Dropdown menus solved three critical problems with natural language approaches: cognitive load (users don't need to memorize and calculate from raw data), accuracy (structured input eliminates speech recognition and parsing errors), and predictability (users know exactly what questions the system can answer). The menus are hierarchically organized—selecting "Specific Data" determines time frame options, which in turn determine available price metrics—guiding users through valid query construction. Audio and text input approaches, while valued for flexibility, raised significant concerns. Users with speech disabilities or non-native English worried about recognition accuracy. The auditory approach raised privacy concerns for public spaces. When systems failed, users couldn't determine whether they phrased questions incorrectly or the information wasn't available. Numeric tables, despite being WCAG-recommended, proved inadequate: even with proper headers, navigating dense data while mentally tracking comparisons overwhelmed working memory. A critical finding was that users rejected one-size-fits-all design. They wanted all four options available, selecting based on task context and personal preference. One participant summarized: "Accessibility is for everyone, not just for people with disabilities." The study also discovered unexpected educational value: financial novices reported that the dropdown structure taught them what questions to ask about stock charts—information they "might have overlooked." This exemplifies the curb-cut effect, where accessible design benefits broader populations including sighted users with low financial literacy.
Relevance
This research provides a model for making complex data visualizations accessible beyond simple text alternatives. The finding that WCAG-recommended numeric tables fail in practice for complex, time-series financial data is significant—it demonstrates that technical compliance doesn't guarantee usability. ChartMaster's approach of working with underlying data (Java ArrayList) rather than the visual chart itself, generating natural language summaries with computed statistics (averages, highs, lows), offers a pattern applicable to other chart types. The co-design methodology proved exceptionally productive: two of three design concepts (text input and dropdown menu) came directly from participant suggestions. This validates involving users throughout the design process rather than only for final evaluation. The discovery that dropdown menus have educational value—teaching novices what questions to ask—suggests accessible interfaces can scaffold domain knowledge, not just present information. For practitioners, the key lesson is that flexible, multi-modal interfaces outperform single solutions, and that accessibility for complex professional tools like financial charts requires domain-specific design, not just technical remediation.
Tags: blindness · screen reader · data visualization · financial literacy · inclusive design · co-design · charts and graphs · multimodal interface
Standards referenced: WCAG 2.0