Project: On the Use of Relevance Feedback and Visual Representations to Help in Classifying Multimedia Information
Description
In this fast digital content world, it is not uncommon for people to have thousands of images, texts, or songs in their possession that are usually not discarded but often not consumed. Some may create different strategies to organize content, but classifying and finding interesting multimedia items (e.g., songs) without labeling information of what is interesting and what is not has become challenging (mainly because "interesting" varies from person to person). Among the existing solutions, relevance feedback approaches allow users to classify (or rank) extensive collections of items by only tagging a few items.
This project aims to create a visual analytics solution that uses relevance feedback or active learning strategies coupled with visual representations to help users explore multimedia information to retrieve the content of interest, reducing the number of items (e.g., songs, photos, etc.) someone needs to browse. This project's main challenge is creating dynamic visual representations, e.g., dimensionality reduction, that adapt based on user expectations (distances change as users give feedback).
Details
- Student
-
EKElian Klaassen
- Supervisor
-
Fernando Paulovich