"... In PAGE 7: ...Table1 shows the list of 50 tested web sites and the corresponding results. The number shows the percentage of perfect cases in each web site.... ..."

"... In PAGE 3: ... Table 1: Fortune 500 Firms with Public Web Sites - by SIC Division SIC Division # Firms SIC Division # Firms A: Agriculture, Forestry, Fishing 0 F: Wholesale Trade 30 B: Mining 2 G: Retail Trade 55 C: Construction 7 H: Finance, Insurance, Real Estate 81 D: Manufacturing 192 I: Services 43 E: Transportation, Communication, Electric, Gas, and Sanitary Services 80 6. Site Content Table2 lists the number and percentage of sites with each type of content. Three items appear on more than 90% of the sites: quot;Product/service description, quot; quot;View financials, quot; and quot;Send e-mail to firm.... In PAGE 7: ... Findings Summary These Findings answer the Research Questions. Question 1, quot;What are the categories of site content on Fortune 500 Web sites, quot; is answered by identifying four categories of content on Fortune 500 Web sites - Customer Sales, Customer Service, Communication, and Informational ( Table2 ). Informational content is the most prevalent content area, answering Question 2, quot;Which Web site content categories most frequently appear on Fortune 500 Web sites quot; (Table 2).... In PAGE 7: ... Question 1, quot;What are the categories of site content on Fortune 500 Web sites, quot; is answered by identifying four categories of content on Fortune 500 Web sites - Customer Sales, Customer Service, Communication, and Informational (Table 2). Informational content is the most prevalent content area, answering Question 2, quot;Which Web site content categories most frequently appear on Fortune 500 Web sites quot; ( Table2 ). Question 3, quot;Do Fortune 500 industries differ in Web site content, quot; is answered affirmatively.... In PAGE 7: ... Currently, the public Web sites of the Fortune 500 emphasize information presentation. The sites contain a high percentage of Informational content ( Table2 ). A few items in the other three categories, such as the ability to send e-mail, are available on the sites.... In PAGE 8: ... Aggregating data across all 490 firms, or analyzing data at the SIC Divisional level, masks some important findings because of the differences that exist between industry groups. For example, when considering all 490 Web sites, the percentage of transactional sites is within one-half point of the percentage of sites that publish information to suppliers or resellers ( Table2 ). This comparison might imply that firms look equally to customers and suppliers in the supply chain when they use the Web to communicate with trading partners.... ..."

"... In PAGE 7: ...mages in a test set. We used C4.5 [13] to build the decision tree. Specifically, we performed forty tests, in which images at a Web site among the forty sites shown in Table1 were chosen as the test set and images at the rest of thirty nine sites were chosen as the training set. We show the results for the following five cases.... ..."

"... In PAGE 7: ...days and the results (see Table4 ) were almost the same. We considered only documents in certain formats that were accessible via HTTP protocol.... In PAGE 7: ... For instance, we ignored audio and video files, which is natural, but also doc, ppt, rtf, and txt documents, which may be arguable. (Omitting these formats in one of the experiments caused only one change in the middle part of the chart in Table4 .) To prevent the Web agent from getting stuck in Web traps, we decided the maximum height of the document tree to be eight, which is a good estimate by experience.... In PAGE 7: ...Table 4: Authoritative Web sites In total, the agent gathered over 250 thousand documents of requested types and it created a corpus of about 7 GB. In Table4 , we can see that the Web sites (or departments if we identify institutions with their Web sites) are split into three clusters. At the top, there are three sites distinctly ahead of the others.... In PAGE 7: ...o citations. Citations are links from documents on other servers. Also, we removed duplicate citations or self-citations. As quality ranking is somewhat tricky we do not disclose the relationship between Table 3 and Table4 at present. 3.... In PAGE 8: ...significantly distinct results. Another trouble is dynamically generated Web pages (see the server with a much larger number of documents in Table4 ). Two and more URLs (and thus two and more possible citations) may represent one document and citations by them should be counted only once then.... ..."

"... In PAGE 9: ... In those cases our heuristics perform at least as well, if not better. However, there are some cases where the four heuristics in [6] perform poorly; for example, on the sites listed in Table12 , their heuristics only achieved a success rate of only 59%. In these cases our heuristics still perform very well, with a 93% success... ..."

"... In PAGE 16: ... Those interested in experimental performance evaluation should continuously monitor the state-of- the-art. Table7 provides sources for the benchmarks described in this article. The references at the end can provide new information on tools and benchmarks.... ..."