Impact of Sentence Length on the Readability of Web for Screen Reader Users

Abstract

Readability of text is generally believed to be connected to sentence length. Most studies on readability are based on visual reading. Less is known about text readability for users relying on screen readers, such as users who are blind. This study therefore set out to investigate the effect of sentence length on the readability of web texts accessed using screen readers. A controlled withinsubjects experiment was performed with twenty-one participants. Participants used a screen reader to read five texts with different sentence lengths. The participants' comprehension and perceived workload were measured. The findings reveal that there is a significant effect of sentence length and most participants exhibit the highest comprehension and lowest workload with sentences comprising 16-20 words. Implications of these results are that web content providers should strive for sentence length of 16-20 words to maximize readability.

Assessment of Readability Level of Reading Materials in Indonesia EFL Textbooks

“…Similar to this, the result findings by Suheri, Azhar, and Afrianto (2018) have also shown that the text with the lowest total number of sentences and the lowest total of words illustrating the text is difficult, with 10 total sentences and 141 total number of words. Kadayat and Eika (2020) and Putri (2021) have added that the readability of a text is connected to sentence length. This is also supported by the finding of Kamarudin and Sugianto (2020) and Fata and Aprilya (2021), who have mentioned that readability is affected by sentence and word aspects of the materials.…”

Beyond Alternative Text and tables: Comparative Analysis of Visualization Tools and Accessibility Methods

“…webAIM publishes screen reader usage statistics for blind people [Blia] and low vision people [Web]. There have been numerous research reports on how people use screen readers to navigate the Web and what kinds of barriers they face (e.g., the impact of sentence length [KE20] and redundant information [GTS18], common frustrations [LAKM07], browsing strategies [BBDR10], and coping tactics [VH13, VH14, MVPA18]). A few research studies have reported the accessibility status of government websites [HR13], social media [VAWL16], and workplaces [WWB18].…”

Multi-language Information Extraction with Text Pattern Recognition

“…Therefore, the baseline has to be a "reasonable smart" random model that would pair each noun with each verb and consider all combinations of these as triples. Assuming there are one verb and two nouns per sentence the random model would yield an accuracy as shown in Equation4, but since the sentences should strive to be between 16-25 words [29] the accuracy is drastically reduced. Let's say that the baseline picks 3 words randomly from sentences, the accuracy would converge towards one fraction of 16 in best case and one fraction of 25 in worst case, see Equation6.…”