MSU at ImageCLEF: Cross language and interactive image retrieval

In this report, we describe our studies with cross language and interactive image retrieval in ImageCLEF 2004. Typical cross language retrieval requires special linguistic resources, such as bilingual dictionaries. In this study, we focus on the issue of how to achieve good retrieval performance given only an online translation system. We compare two approaches, i.e., a translation-based approach and a model-based approach, and find that the later one performs substantially better than the former one. For interactive image retrieval, we investigated the potential use of user relevance feedback (URF), which was designed to address the mismatch problem between user queries and system descriptions. Our strategy is to let the system select important terms for user feedback before expanding queries. However, our preliminary results appear to indicate that the URF approach developed at the current stage is not working. We report our current investigation and discuss lessons learned from this experience.