INDTime: Temporal Tagger––First Step Toward Temporal Information Retrieval

Abstract

In today’s era, a large amount of digitized information is generated every day. Retrieving information about a specific time period is one of the demands of information retrieval system. Such time-related information can be helpful to add time dimensions in existing information processing and information retrieval systems. In this paper, we represent INDTime, a rule-based system that is aimed to recognize and normalize temporal expressions present within the text document. It is designed to automatically annotate temporal expressions as per TIMEX3 annotation standard. The system has been evaluated on three datasets, one of them is WIKIWAR dataset and the other two are manually annotated datasets. We achieved an average precision of 94.15 %, recall of 92.55 %, and f-measure of 91.37 % on three datasets for recognizing temporal expressions and average precision of 90.95 %, recall of 92.55 %, and f-measure of 91.37 % for normalization.