Introduction     Features Demonstration Publications Contact Us
What's Octopus?
   
    Octopus™
is an aggressive mechanism for user-adaptive search of multi-modal information based on multifaceted knowledge base, an ongoing research project at Dept. of Computer Engineering and Information Technology of City University of Hong Kong.

    Octopus™ aims to promote an innovative scenario of multimedia retrieval -- users retrieve semantically meaningful and user-adapted multi-modal objects (including text documents, images, video clips, audio clips) through easy-to-compose queries -- which brings about brand-new user experiences in searching the Web, digital libraries, or other multimedia repositories.

    To achieve the above objective, we have designed some unique features for  Octopus™, including a multifaceted knowledge base, link analysis based retrieval algorithms. and  a learning-from-interactions strategy. An prototypical system is under implementation, the online demonstration of which will be available soon. The published papers about (or related to) Octopus™ can be downloaded from this site.

  

Why Octopus?
    The proliferation of multimedia data in a variety of modalities challenges the traditional retrieval technologies in the following aspects:
  • Users prefer multi-modal information (medias of various types)  to a single type of media. 
  • Users prefer semantically meaningful results.
  • Each individual user has particular need that differs from each other.       

     The three requirements go beyond the scope of conservative traditional retrieval technologies, which focus on single types of media using low-level data features. In contrast, Octopus™ has an rather aggressive spirit in that it deal with multi-modal objects by exploring a broad range of knowledge as well as user interactions. Therefore, Octopus™ can meet all the requirements of multimedia retrieval in a variety of environments.

 News
Our paper about Octopus has been accepted by IEEE ICIP 2002 recently.
 Links
กก


This site is updated on Apr.13th, 2002 by Jun Yang.