POSHAN: Cardinal POS Pattern Guided Attention for News Headline Incongruence

Rahul Mishra, Shuo Zhang

Abstract: Automatic detection of click-baits and incongruent news headlines is crucial to maintain the reliability of the Web and has raised much research attention. However, most existing methods perform poorly when news headline contains contextually important cardinal values such as a quantity or an amount. In this work, we focus on this particular case and propose a neural attention based solution, which uses a novel cardinal Part of Speech (POS) tags pattern based hierarchical attention network, namely POSHAN, to learn effective representations of sentences in the news article. In addition, we investigate a novel cardinal phrase guided attention, which uses word embeddings of the contextually important cardinal value and neighbouring words. In the experiments conducted on two publicly available datasets, we observe that the proposed method gives appropriate significance to cardinal values and outperforms all the baselines. An ablation study of the POSHAN, shows that the cardinal POS-tag pattern based hierarchical attention is very effective for the cases in which headline contains cardinal values.