Query Response Ranking by Temporal Diversity using User Sessions as Feedbacks

Abstract

Retrieval efficiency of temporal issues can be enhanced by getting into account the time aspect. Latest temporal ranking systems use a couple of main strategies: 1) a mixture unit linearly integrating textual resemblance as well as temporal resemblance, and 2) a probabilistic system producing a query from the textual as well as temporal component of report automatically. In this document, we suggest a unique time-aware ranking system according to learning-to-rank strategies. We use two classes of attributes for understanding a ranking system, entity-based as well as temporal attributes, which are based on annotation information. Entity-based attributes are targeted at acquiring the semantic resemblance around a query as well as a document, while temporal attributes determine the temporal resemblance. By using considerable studies we reveal that our ranking system considerably enhances the retrieval efficiency over current time-aware ranking systems.