Research On Sampling Methods For Web Accessibitity Evauation

In the information era,Internet has promoted the development of human society and brought great improvement to people's lives.As an important part of Internet,the website becomes the main platform for information acquisition,interaction,utilization.However,most websites are designed for ordinary people without taking the needs of special groups into account,making existing websites containing a large number of inaccessible Web pages for disabled people,elderly people and children.Therefore,using Web accessibility evaluaton method to find the inaccessible Web pages accurately and obtain the accessibilility level of the website quickly have become important research topics in the field of information accessibility.However,Web accessibility evaluation cannot be fully automated currently and requires expert inspection,which is prohibitively expensive to directly evaluate the website.To solve this problem,sampling methods have been widely introducet into the process of Web accessibility evaluation by selecting a representative sample of pages.Under this circumnstancs,the cost is greatly reduced.Because of the particularity of Web accessibility evaluation,most of the existing sampling methods cannot be directly used.Meanwhile,there are some unsolved or difficult problems int the sampling methods widely used in Web accessibility evaluation to be solved:1)Sampling methods are not optimized for accessibility metrics.2)the quality of the sampling methods is affected by many factors,such as the size of website,the number of samples,the accessibility metrics.3)there exists inappropriate assignment on human sources and samples when evaluating multiple websites.This article focuses on the unsolved problems in the existing sampling methods on Web accessibility evaluation,and proposes some novel sampling methods,including an optimal sampling method for Web Accessibility Quantitative Metric,the active learning for web accessibility evaluation,and the sampling method based on multi-website collaborative evaluation.The main work of this article are as the following:(1)As existing studies show the accuracy of sampling method depends heavily on the metric,this paper proposes an optimal sampling method OPS-WAQM that is optimized for Web Accessibility Quantitative Metric(WAQM).To support quick accessibility evaluation and real-time website accessibility monitoring,this paper proposes an online extension for the sampling method OPS-WAQM.(2)The quality of the sampling methods is affected by many factors,such as the sampling size,the sampling pages,the accessibility metrics.Under these circumstances,many unwanted factors can creep in and bias the evaluation results.To address this issue,this paper proposes a semi-supervised method active-prediction.The method firstly uses active learning method to select representative web page,then uses these sample pages to learn prediction models,and finally obtains the accessibility results of all web pages through the prediction model.(3)For multi-website accessibility evaluation tasks,the existing sampling number assignment method will cause inappropriate assignment on human sources.To address this issue,this paper proposes a sampling method based on multi-website collaborative evaluation.Considering the website's own characteristics,this method can assign appropriate samples for each website.