Deep Learning for Image Search in NAVER
NAVER is the most famous internet company in Korea and the developing company of global messenger LINE. NAVER Search is the key service of NAVER and deals with over 300 million search queries a day. Deep Learning plays a crucial role for improving image search quality and the user interface of NAVER Search recently. First, we extract several local features from document collection by analyzing text context and image content using Deep Learning. Second, by using these features, Deep Learning contributes to global documents semantic analysis and new search interface development. Third, general image tagger with pre-trained Deep Learning model helps search engine to index additional description for the image.
Geunhee has over 10 years of successful experience in large scale data processing for data mining/modelling/searching/crawling. During that time, he has developed several applications and systems for the data utilization such as information retrieval systems, data analytic systems, recommendation systems and so on. He is currently in charge of developing a new image search system in NAVER which utilizes numerical vectorization technique for visual analysis. Prior to that, he was employed in Samsung Electronics and started to work at a start-up developing a large scale search engine after finishing graduate school. He obtained a master degree in industrial engineering and computer engineering in 2007.