Efficient Time-Travel Search over Web Archives

text-indexing techniques do not provide efficient support for time-travel queries. Further, the high level of redundancy between document versions is not exploited. In [3] we proposed a temporal text-indexing technique having three key components: Time-Travel Inverted File Index We build on the inverted file index and extend the structure of its postings by a validity time-interval: 〈 d, score, [tb, te) 〉 Approximate Temporal Coalescing (ATC) Subsequent document versions tend to have highly similar scores for the same term. ATC reduces the index size by coalescing such sequences, while retaining a tunable approximation guarantee. Sublist Materialization (SM) When processing a time-point query qt many postings are read unnecessarily, since t ∉ [tb, te]. SM tackles this problem by systematically materializing smaller sublists. There are two variants of choosing sublists. Performance Guarantee: The minimal amount of space is required, while retaining a tunable performance guarantee. Space Bound: The best expected performance is achieved, while not requiring more than a tunable space bound