Abstract With the vast amount of information available online, searching results for a given query requires the user to go through many titles and snippets. This searching time can be reduced by clustering search results into clusters so that the user can select the relevant cluster at a glance by looking at the cluster labels. For web page clustering, terms (features)... With the vast amount of information available online, searching results for a given query requires the user to go through many titles and snippets. This searching time can be reduced by clustering search results into clusters so that the user can select the relevant cluster at a glance by looking at the cluster labels. For web page clustering, terms (features) can be extracted from different parts of the web page. Giansalvatore, Salvatore and Alessandro [1] have extracted terms from entire web page for clustering. Number of terms returned in this case is more and it produces lengthy vectors. To reduce the size of the vector, Stanis law Osinski et al.,[2] and Ahmed Sameh and Amar Kadray [3] have considered terms from the snippets. In this work, we extracted terms from Uniform Resource Locator (URL), Title tag and Meta tag and we used optimized Kmeans algorithm for clustering. Optimization of K-means algorithm is done by selecting the number of clusters using knee finding algorithm instead of selecting it randomly. We compared our method with existing methods in terms of Intra-cluster distance and Inter-cluster distance. Read More Read Less

Abstract Performance evaluation of classical information retrieval systems usually aims to assess the ability of these systems to find documents considered relevant to a certain search query based on a specific evaluation criteria. This approach, however, is not suitable to adequately evaluate some information retrieval applications such as web search engines. The web special characteristics make information retrieval tasks and the... Performance evaluation of classical information retrieval systems usually aims to assess the ability of these systems to find documents considered relevant to a certain search query based on a specific evaluation criteria. This approach, however, is not suitable to adequately evaluate some information retrieval applications such as web search engines. The web special characteristics make information retrieval tasks and the evaluation of search engines on the web face multiple challenges. Different web-specific, user-specific and language-specific requirements should be considered when designing and performing evaluation tests on operational web search engines. This paper discusses the special requirements for comprehensive comparative evaluation of different web search engines and highlights some languagespecific considerations for evaluation in Arabic language. Read More Read Less