Temporal IR

Temporal information retrieval studies how to find evidence that matches both a query’s meaning and its time requirements, such as what was true at a past date or what is current now. Approaches include extracting query time constraints, ranking with temporal signals or recency decay, and representing evidence with intervals, version histories, or temporal graphs; related work applies these ideas to legal and news search, conversational memory, video, and time series. Benchmarks increasingly test changing corpora, historical versions, long-context retrieval, and whether systems return evidence valid at the requested time. A central challenge is balancing freshness with historical accuracy: old evidence may be obsolete for a current query but essential for a question about the past.