Quantification of time in Digital Libraries: Temporal Zipf's law

Abstract

The temporal dimension of a text document defines the temporal scope of its narrated event. This temporal dimension acquires more importance in corpora created along several years of production, such as digital libraries. Temporal aspects of text have been the subject of many researches with specific tasks, notably information retrieval and event detection, while no studies have been conducted to quantify and analyze the richness of the temporal dimension of different text collections. Analysing thirteen text collections we show how the extent and characteristics of the time presence in text varies among collections that have different scopes, although time intervals are mentioned in almost all the text units analyzed. We found that unique intervals follow the same distribution, given by the Zipf's law, that holds for single words.