Ellen Spertus. ParaSite: Mining the Structural Information on the World-Wide Web. PhD thesis, MIT, February 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... HTML pages contain an internal structure, and hypertext links describe an external structure composed by the structure of Web sites and the Web macroscopic structure (external to Web sites) Several works have showed that it is possible to extract a hierarchical structure describing Web sites [6] [7], 8] while other deal with macroscopic structure [9] 10] An index has to represent the semantic content of documents, including the structure. Thus, an IR model has to integrate links and their impact directly into the document model. II. Structured IR on the Web We distinguish 3 ....

Ellen Spertus, ParaSite: Mining the Structural Information on the World-Wide Web, Ph.D. thesis, Massachusetts Institute of Technology, Cambridge, United States, February 1998.

....has a distinct social security number, but different employees may have the same first name or last name. The following description of the Squeal schema is a slight oversimplification. The real tables are more normalized and have additional fields. For full information on the Squeal schema, see [Spertus 1998 ] The page table, which describes a Web page, has the following fields: contents: the text on the page bytes : the size of the page when: the date and time when the page was last retrieved The following are examples of legal queries (with SQL keywords in capital letters and comments ....

....tables in the local database. 3. Pass the original query (SELECT source url FROM link WHERE destination url = http: www9.org ) to the local database server, which will return a list of URLs, which the Squeal interpreter returns to the user. The interpreter is described in more detail elsewhere [Spertus 1998 , Spertus and Stein 1998 ] 3. Applications We call Squeal applications ParaSites because they exploit information on the Web in a manner unintended by the information s authors. Our goal is not to argue that these are the best possible applications but to show the variety of useful structural ....

[Article contains additional citation context not shown here]

Ellen Spertus. ParaSite: Mining the Structural Information on the World-Wide Web. PhD Thesis, Department of EECS, MIT, Cambridge, MA, February 1998.

....pages. The term sitation has been coined by Gerry McKiernan to describe the study of links among Web pages [4] Jon Kleinberg and his colleagues have developed algorithms to find high quality Web pages by examining link hierarchy [2] 3. BACKGROUND Our recommender system is written in Squeal [8], a system we developed for making queries on the Web in Structured Query Language (SQL) The Squeal relations used for the recommender system appear in Figure 2. We represent relation names in SMALL CAPS, column names in bold face, and parameters in italics. The VALSTRING relation is used to ....

....vdest, URL usource, URL udest, LINK l where vsource.textvalue = www.ai.mit.edu and usource.value id = vsource.value id and l.source url id = usource.url id and udest.url id = l. dest url id and vdest.value id = udest.value id How Squeal determines this information is described elsewhere [8]. The hstruct and lstruct relations indicate where in the page s header and list hierarchy the link appears; for example, under the first H1 header and in a doubly nested list. Figure 3 shows the portions of the LINK, URLS, and VALSTRING tables for the information from the page shown in Figure 1. ....

[Article contains additional citation context not shown here]

Ellen Spertus. ParaSite: Mining the Structural Information on the World-Wide Web. PhD Thesis, Department of EECS, MIT, Cambridge, MA, February 1998.

No context found.

Spertus, Ellen. ParaSite: Mining the Structural Information on the World-Wide Web. PhD Thesis, Department of EECS, MIT, Cambridge, MA, February 1998.

....to be applicable when a page was entirely irrelevant. For this reason, when averaging ratings, novelty was treated as zero when relevance was zero. To capture the simultaneous importance of all three measures, we also computed their products. Full details about the experiment appear elsewhere [7]. On average, the Excite pages were judged more relevant (1.84 vs. 1.36) and interesting (1.63 vs. 1.47) than the ParaSite pages, while the ParaSite pages were judged more novel 3 (1.32 vs. 1.12) and had a higher product (4.58 vs. 4.29) The results of the evaluation of each set of ....

Ellen Spertus. ParaSite: Mining the Structural Information on the World-Wide Web. PhD Thesis, Department of EECS, MIT, Cambridge, MA, February 1998.

....namedArgument = columnName = fetchExpression fetchExpression = expression expression expression expression expression Figure 2: The grammar for fetch statements. Nonterminals tableList, logicExpression, columnList, orderList, and expression are defined elsewhere [15] and have the same meanings as the corresponding SQL nonterminals. 3.3 Virtual and Physical Tables We use the term virtual table to describe a table whose contents are computed only as needed. Every virtual table must have at least one defining column. The term physical table describes ordinary ....

....s.base = p.number and p.number 100 The select p statements are interpreted by the underlying RDBMS. 3.5 Algorithms At the core of the JIT interpreter is a set of algorithms for converting select v statements into fetch and select p statements. The algorithms are described in full elsewhere [15]. The process involves finding an ordering t 1 , t n for referenced tables such that for every table t , either: 1. t i is a physical table, or 2. a defining column of t i is bound to a simple data object (such as an integer) or to a column of a table t j , where i j. This allows a ....

[Article contains additional citation context not shown here]

Ellen Spertus. ParaSite: Mining the Structural Information on the World-Wide Web. PhD Thesis, Department of EECS, MIT, Cambridge, MA, February 1998.

No context found.

Ellen Spertus. ParaSite: Mining the Structural Information on the World-Wide Web. PhD thesis, MIT, February 1998.

No context found.

Spertus Thesis 98 Ellen Spertus. ParaSite: Mining the Structural Information on the World-Wide Web. PhD Thesis, Department of EECS, MIT, February 1998.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC