1 INTRODUCTION
With the rapid development of Internet techniques, especially, Web2.0 techniques, the Web becomes
the most popular platform for users to share and disseminate the news information. This is extremely
accelerated by the popularization of mobile devices, which allows us to access news easily and
flexibly. Such mobile accessibility is extremely useful for the visually impaired people who have
always been eager to mingle with the normal people’s life by achieving Web information at any time
and any place.
To access digital information, the visually impaired person typically resorts to the screen reader
programs including Dolphin, JAWS and Window-Eyes in PC, which is the third-party software to
convert textual information into audio streams. Instead of reading information, such special users are
listening to the information with a screen reader. In particular, visually impaired people use keyboard
shortcuts to control a screen reader to transfer the textual information in a PC screen into a linear audio
stream line by line (Wright 1991). However, such powerful screen readers are quite inefficient in
mobile devices because of the limited UI channels, i.e., the lack of physical keyboards. In fact, our
participants without sights emphasized that, although the screen reader software enables them to
access the electronic materials, they still have to struggle when accessing complex information spaces.
The challenge is extremely whelming when the reading information is rich, dynamically growing, and
scattering through the entire Web information space.
Apparently, there is a strong needs to intelligently process the digital information to improve the
accessibility of the visually impaired people. Several efforts have been made in Web accessibility. For
example, (Takagi et al. 2002) proposed a method to transcode the Web page, linearizing the content to
make it more readable for the screen reader. (Rosmaita 1997) developed a text-only Web browser to
transfer the diverse Web layout into linear structure. (Andronico et al. 2006) designed a prototype to
alleviate the searching problem for the visually impaired people by simplifying the result page
returned by search engine. In addition, (Kopecek 1998) presented the AUDIS system, which is a
library system tagged books with labels such as “title” and “chapter” to automatically segment books
for users. Such researches provided an effective way to intelligently process information for the
visually impaired users. However, most of them only deal with information within individual pages
while overlooking the underlying structures of relevant information across pages. Furthermore, most
intelligent techniques only concern about the improvement of browsing functions (navigation and
searching) rather than the other functions including summarization and recommendation.
In this article, we propose an innovative mobile newsreader to help visually impaired people to access
news information. It provides unique functions of summarizing content, and recommending relevant
information to the visually impaired users. Our experimental evaluation with the participation of the
visually impaired users shows that this newsreader reduces the navigational overhead significantly and
enables the visually impaired users to access news articles effectively.
2 SYSTEM ARCHITECHTURE
2.1 System Overview
Traditionally, visually impaired people resort to the screen readers to get the online news. The large
volume of daily news articles with a loose organization of relevant contents makes the visually
impaired people difficult to access the interested news. In particular, the visually impaired people
have to listen to a news article line by line to understand the main topic due to the lack of content
summary. It is almost impossible for these special people to track and read a serial of news articles in