Answering General Time-Sensitive Queries

摘要: Time is an important dimension of relevance for a large number searches, such as over blogs and news archives. So far, research on searching collections has largely focused locating topically similar documents query. Unfortunately, topic similarity alone not always sufficient document ranking. In this paper, we observe that, class queries that call time-sensitive queries, the publication time in archive should be considered conjunction with to derive final Earlier work improving retrieval “recency” target recent documents. We propose more general framework handling automatically identify intervals are likely interest Then, build scoring techniques seamlessly integrate temporal aspect into overall ranking mechanism. present extensive experimental evaluation using variety article data sets, including TREC well real web analyzed Amazon Mechanical Turk. examine several detecting query incorporating information process. show our robust significantly improve result quality compared state-of-the-art techniques.