| 研究概要 |
In this year we have provided methodology for estimating reference time of documents, which we call "focus time". Focus time of documents represents the time periods to which the document content refers to. This concept is orthogonal to the notion of document timestamp or its creation time. Our proposed method uses statistical data derived from large document collections, thanks to which, we can estimate the association of given word with time. Using this data it is possible then to estimate the association of an entire document with the time, even in case of absence of any temporal expressions such as dates. The results of this work have appeared in CIKM2013 conference.
We have also measured the extent to which users use temporal features in their search activities, for example, we analyzed how much users search for past, recent or future-related information on the Web. This survey has been conducted on 110 Web users and its results were published at TempWeb2013 workshop.
To further foster the research in temporal information retrieval we established “Temporal Information Access” task challenge at NTCiR. This initiative provides test bench for evaluating performance of retrieval systems over the same datasets and tasks. The task challenge and the datasets that we provide are described in paper published at TempWeb2014 workshop.
|
