A New Approach To Improve Web Search Results For Multilingual Documents

With an ever-increasing number of resources on the Web,different language resources such as newspapers,magazines,and articles have become available.This demands for an improved technique in the retrieval of information to extirpate the communication barrier among languages.Search engines are the cornerstones for informational contents and the link between the two parties of publishing and retrieval of content.Moreover,the search for information is no longer narrowed to users of the native language of the resource.Many non-English users,such as Arabic speakers are unable(fail)to express the terms in their own language,as most of the terminologies written in the native language of the resource are acquired from the English language.The terms in mixed/multilingual documents,mostly Arabic documents(or non-English documents),especially in the scientific domain are often accompanied by their translations that lead to the co-occurrence of terms in different languages.When users use mixed queries,the results of the multilingual documents will dominate the top retrieved document rather than the relevant documents,since the scores of the mixed documents get extra weights that are not really part of their weighting value.The proposed approach in this thesis applies a new method to adjust the score of multilingual documents according to the value of co-occurrence of terms,in addition to developing a tool to implement this approach to reduce the problem of documents retrieval when utilizing a mixed query.Consequently,this approach led to the enhanced performance of retrieved documents making them relevant,effective and accurate,proving that our approach is more reliable than other traditional approaches.