Best Entry Pages for the Topic Distillation Task
Theodora Tsikrika, Mounia Lalmas · 2013
In a typical web search, users consider entry pages to rele-vant sites as more valuable than isolated pieces of relevant text. The Topic Distillation Task aims at identifying the page at the right level of site hierarchy considered to pro-vide optimal access, by browsing, to relevant pages within the site, i.e. its Best Entry Page. Our aim is to estimate a measure of how good a page is as an entry page to the site it belongs, by aggregating the page’s system-assessed relevance with that of its structurally related, Web pages belonging to the same site. To model this aggregation, we propose a framework which is expressed within Dempster-Shafer Theory of Evidence. Furthermore,we generalise our model by taking into account other system-assessed proper-ties of Web pages. Apart from their relevance, the authority and hub properties of Web pages are considered in the ag-gregation. We evaluate our approach by performing exper-iments using the.GOV test collection. The results of these experiments are promising.