An Interactive Method for Accessing Tables in HTML
Toshiya Oogane, Chieko Asakawa · 1998 · Proceedings of the Third International ACM Conference on Assistive Technologies (Assets '98) · doi:10.1145/274497.274521
Summary
This paper from IBM Japan proposes a method for converting HTML tables into non-visual, navigable representations for screen reader users. At the time, Screen Reader/2 (the Japanese version of OS/2's screen reader) could read displayed text information line by line and word by word, but could not interpret the two-dimensional structure of HTML tables — it simply linearized the content top-to-bottom, left-to-right, destroying the spatial relationships that give tables meaning. The authors developed a prototype system with two components: an HTML analyzer running alongside Netscape Navigator that parses table tags and converts them into individual HTML files (one per cell), and a linking structure that enables directional navigation between cells. Each cell's HTML file contains 8 navigation links: to the cell above, below, left, right, top cell, bottom cell, left-edge cell, and right-edge cell. The system also handles "unsquared" tables (where cells span multiple rows or columns) by splitting merged cells into individual linked cells. Additionally, a table index page provides an overview listing the table dimensions and offering direct links to column header sections for quick access.
Key findings
The prototype successfully enabled visually impaired users who previously could not understand HTML tables to navigate and comprehend tabular data. Users could move between cells directionally using the 8-link navigation structure, and the table index provided a useful overview for quickly accessing specific parts of a table. The evaluation showed that the table index combined with navigation links was most effective — users could access specific cells directly when they already knew the table's rough structure, while the navigation links supported systematic exploration. For unsquared tables with merged cells, the system's approach of splitting into regular cells and linking them preserved usability but lost the original merged structure. The system was evaluated with visually impaired users who were already advanced computer users familiar with Screen Reader/2. When using the original screen reader functions, they could not understand tables correctly and could not extract tabular information from the web. With the new conversion method, they were able to understand tables by navigating them using the table index and cell navigation links, including accessing structured content like television program schedules.
Relevance
This paper documents early work on a problem that would become central to web accessibility: making HTML tables navigable for screen reader users. Chieko Asakawa, one of the authors, went on to become one of the most influential figures in web accessibility research at IBM. The cell-by-cell navigation approach with directional movement (up, down, left, right) directly anticipated the table navigation features now built into modern screen readers like JAWS, NVDA, and VoiceOver, which allow users to move between cells using keyboard shortcuts. The challenge of "unsquared" tables with merged cells (colspan/rowspan) remains a pain point in web accessibility today. For current practitioners, the paper reinforces the importance of using proper HTML table markup (th elements, scope attributes, captions) so that screen readers can present the same navigable structure the authors had to construct manually. The work also illustrates how early web accessibility research at IBM Japan laid groundwork for the accessibility features we now take for granted.
Tags: web accessibility · screen reader · HTML tables · table accessibility · blind and low vision · non-visual interaction · web development
Standards referenced: HTML