TengQi Ye - Transfer Talk - 14th August 2015

Video Category: 
Transfer Talk
TengQi Ye

Title: A Content-based Search Engine for Lifelogging using a Collaborative Common Knowledge Base

Supervisor: Dr. Cathal Gurrin, Dr. Kevin Casey

Abstract: Lifelogging is the process of gathering, processing and exploiting life experience data, which involves various main branches of computer science, including: multimedia, personal sensing and human–computer interaction. As the result of the technological progression of sensor and storage devices, gathering and storing lifelogging data is not longer a technical challenge. However, there has not yet been significant progress in processing and organising lifelogging data, especially how to achieve effective retrieval. It is my belief that lifelogging data is inherently multimodal and the lifelogging data is more suitable to learn concrete multimodal concepts, rather than abstract lifelogging concepts. I argue that the retrieval can be accomplished within two intrinsic stages: converting all views into semantic concept vectors (in text format) and then using a collaborative common knowledge base to associate those semantic concepts to support retrieval. This will be evaluated on the NTCIR-12 lifelogging test collection. There are two dominant contributions from my work. Firstly, I propose the first content-based search engine for lifelogging data. Secondly, my approach introduces a collaborative common knowledge base for multimedia hybrid fusion which will enhance the performance of the content-based search engine over a state-of-the-art baseline.