Extracting Temporal Patterns and Analyzing Peak Events
K. R. Premlatha, T. V. Geetha · 2010
Temporal text mining (TTM) is the process of discovering sequential and temporal patterns in text information collected over time. This is useful in application domains where each entity of text in a text stream (usually a document or publication) has a meaningful timestamp. In this paper, extraction, formalization and comparison of temporal terms has been performed. The extraction of temporal expressions (explicit, implicit, and vague) has been performed by using FSA. In formalization, the natural language expression is converted in to calendar based time-line. Vague expressions have been handled based on the reference time and the duration of indexical features. Peak event has been analyzed by using the start-end time of the event and reference article for the particular event. Start-end time of the particular event has been used as timestamp while analyzing peak event. Finally temporal similarity between events has been calculated and used to convey temporal relation between documents. This similarity has been calculated using Allan's approach. The similarity has been shown as document event matrix conveying temporal relations. This matrix also highlights the peak events across the documents. Hot topic of the documents has also been exposed while comparing temporally similar events. Index Terms—text mining, temporal similarity, clustering, peak event