Extracting Temporal Entity from Urdu Language Text

Abstract

The detection of temporal entities within natural language texts is an interesting information extraction
problem. Temporal entities help to estimate authorship dates, enhance information retrieval capabilities, detect and
track topics in news articles, and augment electronic news reader experience. Research has been performed on the
detection, normalization and annotation guidelines for English temporal entities. However, research for Urdu
language lags far behind and there is a need for lot of work to be done in this regard especially when huge quantity
of Urdu data is being generated on online social networks on daily basis. In this paper, we propose a rule-based
approach for temporal entity extraction for Urdu language. Comparing our approach with existing Urdu temporal
entity extraction approaches, our approach dominates on behalf of accuracy and on tackling with all types of Urdu
temporal entity types. We use a publicly available Urdu data corpus for our experiments which consists of 206 date
tags. We extend this dataset by adding 200 Urdu Fully Qualified Date (UFQD) tags. We also introduce a new date
type for Urdu language called Urdu Partially Fully Qualified. Our proposed system achieved average (0.97, 0.98
and 0.98) (Precision, Recall and F1-Measure) respectively for UFQD and Urdu Partially Fully Qualified Date.
Some challenges and issues of other date types in Urdu Textual Language i.e. Deictic and Anaphoric are also
discussed in detail.