Statistical profiles of highly-rated web sites

Abstract

We are creating an interactive tool to help non-professional web site builders create high quality designs. We have previously reported that quantitative measures of web page structure can predict whether a site will be highly or poorly rated by experts, with accuracies ranging from 67-80%. In this paper we extend that work in several ways. First, we compute a much larger set of measures (157 versus 11), over a much larger collection of pages (5300 vs. 1900), achieving much higher overall accuracy (94% on average) when contrasting good, average, and poor pages. Second, we introduce new classes of measures that can make assessments at the site level and according to page type (home page, content page, etc.). Finally, we create statistical profiles of good sites, and apply them to an existing design, showing how that design can be changed to better match high-quality designs.

Building up usability-engineering capability by improving access to automated usability evaluation

“…• Processing scheme lists the type of information and sketches the algorithm that is used to process the source information. (Card et al, 1983;John and Kieras, 1996) Webtango (Ivory-Ndiaye, 2003;Ivory et al, 2001b;Ivory and Hearst, 2002;Ivory and Megraw, 2005) Cognitive Walkthrough (Blackmon et al, 2002;Kitajima, 2005;Kitajima et al, 2000;Landauer and Dumais, 1997;Pirolli and Card, 1999) Text on links/ text on headings of the pages ProcessLens (Heftberger and Stary, 2004;Dittmar et al, 2004;Stary, 2000Stary, , 2006Stoiber and Stary, 2003) UML specification…”

Implementation and evaluation of a quality-based search engine

“…The system WebTango extracts more than 150 atomic and simple features from a Web page and tries to reveal statistical correlations to a set of sites rated as excellent [24]. The extracted features are based on the design, the structure and the HTML code of a page.…”

“…A third fell into the categories good, average and poor. Classifiers could learn this assignment with an accuracy of 94% in the test set [24]. A different application domain is addressed by the Intelligent Essay Assessor which based on Latent Semantic Indexing (LSI) [20].…”

“…Many of the features ultimately selected for AQUAINT were also used in other experiments [2,24,39], others have been developed for the system presented. Most of the atomic features are determined by the graphic and visual design of the page, however, some also are based on the content of the page.…”

“…These include the number of graphics on a page, the number of links to graphics, the relation between the number of graphics and the file size. Some of these features are motivated by the work in [24], however, our approach has no pre-defined notion of the meaning of a specific feature. The effect of the feature on the quality model is purely defined through machine learning based on the examples.…”