Learning algorithms for intelligent retrieval of multi-modal documents

• Unité de recherche: INNOSUISSE
• Numéro de projet: 7024.2;10 ESPP-ES
• Titre du projet anglais: Learning algorithms for intelligent retrieval of multi-modal documents

Description succincte (Anglais)

Learning algorithms for intelligent retrieval of multi-modal documents

Description succincte (Français)

Learning algorithms for intelligent retrieval of multi-modal documents

Résumé des résultats (Abstract) (Français)

Classic search engines persistently fail to capture 'what the user wants' and return many irrelevant results. However, many applications of search technology that serve urgent public and business needs depend on improved methods for 'getting better results faster'. The project attempts to optimize algorithms and methods for ad-hoc ranking and classification of multimodal documents from a very large dataset. Application of latest machine learning techniques will allow better data analysis, resulting in faster and more accurate system response. Furthermore, document analysis applications are mostly multi-modal by nature, requiring clustering of at least text and images or sketches for analysis. The project therefore specifically attempts to identify adequate algorithms and methods for ranking and classification of multi-modal documents.