An Updating Scheme Based on Long-Term Relevance Feedback Learning in VAST System

In our earlier works on VAST (visuAl & semantic image search) system, the semantic network effectively associated keywords and visual feature clusters. However, we only concerned about the construction of the semantic network before, and did not consider the updating of the semantic network. In this paper, an updating scheme base on long-term relevance feedback learning is proposed to update the semantic network in VAST system. The updating scheme keeps up the characteristic of automatic retrieval for the semantic network, and further makes full use of the userpsilas feedback information. Therefore, the semantic network with the updating scheme gives a good tradeoff between utilizing the user\´s feedback and avoiding the "lazy user" problem. The experimental results show the effectiveness of the proposed updating scheme.