[Adapt] Seminar

[Kaiqi Zhao]

Hi guys,


In this seminar, I will introduce our work:
Clustering Web Image by Entity Disambiguation


Abstract:
Image search engine returns results in a mixed bag of different entities and concepts without meaningful ordering or grouping because it doesn’t understand the semantics or recognize the objects in the images. This poses inconvenience for users who are looking for a particular entity without knowing the exact right term to search.  We propose a novel framework that understands the context by disambiguating the terms in the context into the corresponding concepts from an external knowledge base in a process called conceptualization. The framework also incorporates any type of visual signals to complement the text signals in a tri-stage clustering algorithm which produces better clustering results than state-of-the-art approaches. In addition, the framework automatically annotates each cluster of images by its key concepts which allows users to quickly identify the images they want.


I hope you will enjoy this talk.


Best regards,
Kaiqi
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://cs.sjtu.edu.cn/pipermail/adapt/attachments/20121212/55a52349/attachment.html>