A new IR Method for Mixed Document Scoring Adjustment Using Multilingual Document

Abstract

Non-English-speaking users, such as Arabic speakers, are not able to express terminology in their native languages. It leads to mixing languages together. Therefore, such non English speaking users may express their queries in a mixed form between two languages, mostly English and the native language, in order to precisely present their concepts to search engines. This also makes the web essentially cross lingual and/or multilingual. These effects specifically in the non-English scientific documents, which contain some, English terms to express terminology. It shown that in non-English documents, those in Arabic for example, terms often accompanied by their translations resulting in the co-occurrence of the same term in two different languages. When searching using mixed query the result is that this documents dominate the top retrieved documents. However, the approach is to add another level to adjust the score using the co-occurrence to minimize it. After implementing the proposed method, it show improvement in the ranking. In this thesis, we have designed a new method to set the matching level in retrieving information from multilingual documents. The method tested on a set of documents (watan-2004 of a textual nature). The results in this research show that the proposed new method has improved the accuracy of retrieval after comparing it to a search engine that does not have this feature.