A web browsing system based on adaptive presentation of web contents for cellular phones
Yuki Arase, Takuya Maekawa, Takahiro Hara, Toshiaki Uemukai, Shojiro Nishio · 2006 · Proceedings of the 2006 International Cross-Disciplinary Workshop on Web Accessibility (W4A) · doi:10.1145/1133219.1133234
Summary
This paper from Osaka University and KDDI R&D Laboratories presents a web browsing system that adaptively presents web page components on cellular phones based on their content characteristics. The authors identify that browsing desktop-designed web pages on mobile devices involves two burdensome steps: scrolling through the entire page to find a target content block, then scrolling within that block to read the information. Their system addresses both steps through a two-level approach. First, it provides a scaled-down overview of the entire page where components are automatically identified and can be zoomed into by focusing the pointer on them. Second, when a user selects a component, the system presents it using an adaptive strategy matched to the component's content type. The authors classify web page components into six categories based on analysis of fifteen typical websites: Text, Text&Image, Image, Link&Image, Vertical link set, and Vertical&horizontal link set. A preliminary experiment with sixteen subjects tested nine auto-scrolling patterns (varying speed and zoom ratio) to determine the most effective presentation for each category. The system was implemented as a server/client architecture with a PC server (Visual C# and PHP) and a Java client on an NTT Docomo cellular phone (SH900i) with a 240x320 pixel display.
Key findings
The preliminary experiment revealed that auto-scrolling is not suitable for Text and Text&Image components because reading speeds vary significantly among individuals, and users of Text&Image components could not view images when they wanted during auto-scrolling. Auto-scrolling was effective for Image, Link&Image, and link set components. Zigzag scrolling paths were universally disliked. For Image components, users preferred 50 pixels/second auto-scrolling; for link set components, preferences split between slow (20 pix/sec) and medium (50 pix/sec) speeds, corresponding to two distinct link-searching styles — thorough reading versus skimming. The system calculates adaptive scrolling speed using a formula based on information density (characters per pixel area) and component breadth, with a configurable parameter for characters recognizable per second. For Text&Image components, the system allows users to manually scroll text while switching to associated images on demand via key press — a design that respects both the user's reading pace and their desire to view images in context. The system also avoids zigzag scrolling by zooming out oversized components to fit one display dimension.
Relevance
This paper contributes to understanding how content structure affects browsing behavior on constrained devices — an insight directly applicable to accessibility. The finding that different content types require different interaction models (manual scrolling for text, auto-scrolling for image galleries and link lists) parallels accessibility concerns about providing appropriate interaction modes for different user needs. The component classification approach — analyzing what a content block contains and adapting presentation accordingly — anticipates modern responsive design thinking where content type drives layout decisions. For practitioners, the key lesson is that a one-size-fits-all approach to content presentation fails both on small screens and for users with diverse abilities. The research also demonstrates that user control matters: auto-scrolling worked only for content types where users scan rather than read carefully, reinforcing the WCAG principle that users should control the pace of content consumption.
Tags: mobile accessibility · content adaptation · adaptive presentation · small screen devices · auto-scrolling · web browsing