Temporal Representation and Reasoning for the Semantic Web
Sebastian Hübner, Ubbo Visser · The Florida AI Research Society · 2008
We introduce period names, a new approach for quantitative and qualitative temporal representation and reasoning of resources in the Semantic Web. Our work has been implemented into the Bremen University Semantic Translator for Enhanced Retrieval (BUSTER). This enables us to perform a query of the form concept@location in time. Motivation and requirements An important part of search in the Web is time-dependent. The W3C recommends a standard that is machine oriented, limited to single points in time, and offers neither meaningful labels nor qualitative relations. Humans however, make use of these features very often. There is thus a need for the development of a new temporal representation scheme and a reasoning engine. A small study of online news articles from BBC News and The New York Times revealed that the temporal terms can be classified into the following groups: (a) Qualitative vs. quantitative. The qualitative altervative covers terms like “now” and “for a long time”, whereas the quantitative altervative covers terms like “on 21 November” and “in the last year”. (b) Absolute vs. relative. The absolute altervative includes terms like “in 2001” and “beginning of November”, whereas the relative altervative covers terms like “earlier” or “in two months”. Taking this into account, requirements for temporal annotation (and reasoning) in the Semantic Web can be formulated: a) Intuitive labelling: label time intervals with intuitive names. b) Boundaries: boundaries of time intervals should be flexible. The required types are exact, fuzzy, persistent, and unknown. c) Structures: an interval can be based on another interval, can be self-defined or imported. Functions (e.g. earliest beginning of) are needed to extract the significant time points from the intervals. d) Explicit qualitative relations: relations between intervals are needed when using persistent or unknown boundaries.