Back to Our Roots for Retrieving Very Short Passages

摘要

This article tackles the task of retrieving very shortdocuments via even shorter queries. The problem on handmay relate to the retrieval of tweets, image and tablecaptions, short text messages (SMS) and sponsored retrievalamong others. In such cases, document and/or queryexpansion using thesauri and other external resources (e.g.,Wikipedia) usually available on the World Wide Web(WWW) are proven to be effective approaches. However,the focus of this paper is on documents that are written inlesser known languages for which the WWW is of limiteduse. Our experiments are based on two main corporaextracted from historical manuscripts written in Latin andMiddle High German. We found that retrieving very shortdocuments whose lengths are quite similar via short queriesgiven that no external enrichment resources are available,the classical tf-idf model performs as satisfactorily as themore complex models do, if not better sometimes.