DEU at ImageCLEFmed 2009: Evaluating re-ranking and integrated retrieval model

Berber T., ALPKOÇAK A.

2009 Cross Language Evaluation Forum Workshop, CLEF 2009, co-located with the 13th European Conference on Digital Libraries, ECDL 2009, Corfu, Yunanistan, 30 Eylül - 02 Ekim 2009, cilt.1175, (Tam Metin Bildiri) identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Cilt numarası: 1175
  • Basıldığı Şehir: Corfu
  • Basıldığı Ülke: Yunanistan
  • Anahtar Kelimeler: Content-based image retrieval, Image- clef medical retrieval task, Information retrieval, Machine learning, Re-ranking
  • Dokuz Eylül Üniversitesi Adresli: Evet

Özet

This paper presents DEU team participation in imageCLEF2009med. Main goal of our participation is to evaluate two different approaches: First, a new re-ranking method which aims to model information system behaviour depending on several aspects of both documents and query. Secondly, we compare a new retrieval approach which integrates textual and visual contents into one single model. Initially we extract textual features of all documents using standard vector space model and assume as a baseline. Then this baseline is combined with re-ranking and integrated retrieval model. Re- ranking approach is trained using ImageCLEFmed 2008 ground truth data. However, re-ranking approach did not produced satisfactory results. On the other hand, our integrated retrieval model resulted top rank among all submissions in automatic mixed runs.