Towards content expiry date determination: Predicting validity periods of sentences

Abstract

Knowing how long text content will remain valid can be useful in many cases such as supporting the creation of documents to prolong their usefulness, improving document retrieval or enhancing credibility estimation. In this paper we introduce a novel research task of forecasting content’s validity period. Given an input sentence the task is to approximately determine until when the information stated in the content will remain valid. We propose machine learning approaches equipped with NLP and statistical features that can successfully work on a relatively small number of annotated data.