Abstract: Satirical news is a type of parody presented in the format of news. The word satire itself means the use of humor, irony, exaggeration, or ridicule to expose and criticize peoples stupidity or vices. The satirical news is particularly used in the context of contemporary politics and other topical issues. It is a double edged sword, on one hand it is entertainment and on the other hand it is deceptive and harmful. And it could be potentially misleading, in either of the following cases. First one is Satirical cues are too subtle to be recognized. And the other one is reader lack the contextual or cultural background. Satirical news detection is considered as a problem since not everyone can recognize it as a satire one. And the spreading of the false news may lead to hurt the credibility and trust in the social media websites. Existing works only consider document level features to detect the satire, which could be limited. This paper considers all paragraph-level linguistic features to find the satire by incorporating neural network and attention mechanisms. There exist different methods to find out whether the article is satire or not. In this paper the comparative study on the different approaches for the detection of the satire is performed. Its also investigate the difference between paragraph-level features and document-level features.

Keywords: Satirical News Detection, Natural Language Processing (NLP), Deep Learning, Attention Mechanism