The central hypothesis of this work is that retrieval effectiveness of multilingual documents can be improved by simultaneously providing the search engine human-translated multilingual queries identified with their source languages. Inforadar^ML enhances Inforadar by adding support for multilingual queries and document collections.

Our research addresses an inherent problem in information retrieval: formulating precise and effective information retrieval queries has always been a difficult task, even for experienced user.

Hence, simple queries often return extremely large and imprecise result sets when used with large corpora. The task of formulating an effective query requires the user to predict which terms appear in documents relevant to the information needed. This often requires extensive knowledge about the document collection, which it’s difficult to obtain in large document corpora.

In addition, users want to avoid retrieving irrelevant documents due to a query that is underspecified or contains ambiguous terms. As a result, users of an information retrieval sys-tem with a large corpus are often faced with the task of manually sifting through very large and often inappropriate result sets.

Our research is also motivated by the incredible rate at which the web is becoming an enormous corpus of multilingual information. Hence, the same information needs can be expressed with semantically equivalent queries in multiple languages.

Although ubiquitous access to this information is sometimes inherently limited by language barriers we believe that, even when such barriers are weak or non-existent, access to multilingual information remains difficult simply because the tools used were not designed with multilingual information in mind. Crosslingual information retrieval (CLIR) systems attempt to provide the best possible set of multilingual relevant documents in response to a query expressed in one language, often called the source language.

Having in mind all the facts mentioned above, our hypothesis is that retrieval effectiveness of multilingual documents can be improved by simultaneously providing the search engine multilingual queries identified with their languages, and allowing multilingual users to translate their queries.

[ Next: System Requirements ]