Producing Accessible Statistics Diagrams in R
Donal Fitzpatrick, A. Jonathan R. Godfrey, Volker Sorge · 2017 · Proceedings of the 14th International Web for All Conference (W4A) · doi:10.1145/3058555.3058564
Summary
This paper presents a system for making statistical diagrams accessible to blind users by generating semantically enriched SVG graphics from the R statistical computing environment. The core problem is that statistical graphics — bar charts, histograms, box plots, scatter plots, time series — are powerful tools for conveying data patterns to sighted audiences but are practically useless to blind users. Simple text descriptions are insufficient because they fail to convey the meta-information that sighted users extract at a glance: patterns, outliers, gaps, distribution shapes, and relationships between data points. The authors' approach works by extending R's BrailleR package (originally designed to produce text summaries of graphs) to generate rich semantic XML alongside SVG output using R's gridSVG package. This semantic information is then consumed by DIAGcess, a JavaScript library that enables screen reading and interactive exploration of annotated SVG diagrams in any web browser. The system was designed with two core principles: blind users should not need additional hardware to explore diagrams, and they should be able to obtain the same information as sighted peers — not just a textual summary. Two of the three authors are themselves blind, directly informing the design.
Key findings
The system implements a three-layer hierarchical navigation model for exploring statistical diagrams via screen reader: (1) a top-level summary providing the title and range information, (2) a major component layer with four elements (title, X axis, Y axis, and data content), and (3) single component exploration for drilling into individual data points. Navigation uses arrow keys — up/down to switch hierarchy layers, left/right to move between components on the same layer. Components are screen-read with short or detailed descriptions based on user preference, accompanied by synchronized visual highlighting and optional magnification of the selected element. For discrete data graphs (histograms, bar charts, box plots), individual bars or points can be explored step-by-step. For continuous data graphs (scatter plots, time series), the data is either broken at natural gaps or chunked into segments, with each segment described by its values and relationship to neighboring segments (increasing/decreasing). The system is implemented as a CRAN package (R's main repository), ensuring broad availability. Initial feedback from four power users was generally positive, though integration issues with dyslexia tools were noted. The approach currently works best with gridSVG output but the goal is to extend to other R renderers.
Relevance
This research addresses a critical barrier to STEM education and employment for blind people: the inaccessibility of data visualizations. Statistical literacy increasingly requires the ability to create and interpret graphics, and blind students and professionals who cannot independently produce or explore charts are at a significant disadvantage. By integrating accessibility directly into R — the tool statisticians already use — the authors lower the adoption barrier compared to requiring specialized software or hardware. The hierarchical navigation model provides a thoughtful balance between overview and detail that mirrors how sighted users naturally explore graphs (scanning for patterns first, then examining specifics). The principle that blind users should obtain the same information, not just a summary, is an important design philosophy that goes beyond typical alt-text approaches to data visualization accessibility. The involvement of blind researchers as both designers and users strengthens the work's credibility. Limitations include the current restriction to gridSVG output, the need for more formal user evaluation, and the challenge of representing inherently visual concepts like Venn diagrams or network graphs in this hierarchical model.
Tags: data visualization accessibility · STEM accessibility · visual impairment · blindness · screen readers · SVG · statistics · R programming · accessible diagrams · hierarchical navigation · BrailleR · interactive exploration · education accessibility · charts · graphs