Features of Octopus™ |
กก
- Multifaceted Knowledge
Base. Multifaceted knowledge base is used to models the knowledge on
the relevance among media objects of various types (text documents,
images, video clips, audio clips). Specifically, the knowledge base is
built on top of a Layered Graph Model (LGM) constituted by a
set of superimposed layers, with each layer representing the inter-object
relevance defined from a certain perspectives, ranging from low-level
features, structural relations, to user perception.
กก
- Link Analysis based
Retrieval Algorithm. Link analysis technique is developed to
analyze the "relevance links" (among multi-modal objects) recorded in the
LGM, in order to support various retrieval-related tasks, including
search, browsing & navigation, clustering, and classification. Such
retrieval algorithms can take advantage of the broad range of knowledge in
the LGM, and deal with multiple types of media.
-
Learning-from-interactions Strategy. This strategy encourages user
interactions in various forms (e.g., querying, browsing, feedback) and
learns user characteristics (e.g., interests, preferences) from user
behaviors in the interactions. Such user knowledge is incorporated into
the LGM and reused in the subsequent queries to retrieve user-adpated
results.
|
|
| News |
| Our
paper about Octopus has been accepted by IEEE ICIP 2002 recently.
|
|
Links |
| กก |
กก |
