Improving search via personalized query expansion using social media

摘要: Social tagging systems have gained increasing popularity as a method of annotating and categorizing wide range different web resources. Web search that utilizes social data suffers from an extreme example the vocabulary mismatch problem encountered in traditional information retrieval (IR). This is due to personalized, unrestricted users choose describe tag each resource. Previous research has proposed utilization query expansion deal with this rather complicated space. However, non-personalized approaches based on relevance feedback personalized co-occurrence statistics only showed limited improvements. paper proposes novel framework individual user profiles mined annotations resources marked. The underlying theory regularize smoothness word associations over connected graph using regularizer function terms extracted top-ranked documents. intuition behind model prior assumption term consistency: most appropriate for are likely be associated with, influenced by documents ranked highly initial query. also simultaneously incorporates through Tag-Topic latent graph. experimental results suggest can produce better than both classical approach other methods. Hence, significantly benefits leveraging users' media data.