Humor Detection in Persian: A Transformers-Based Approach Fateme Najafi-Lapavandani Faculty of Mathematics & Computer Science Amirkabir University of Technology Tehran, Iran fatemenajafi135@gmail.com Mohammad Hassan Shirali Shahreza  Faculty of Mathematics & Computer Science Amirkabir University of Technology Tehran, Iran hshirali@aut.ac.ir Received: 5 May 2022 – Revised: 25 August 2022 - Accepted: 27 November 2022 Abstract—Humor is a linguistic device that can make people laugh, and in the case of expressing opinions, it can transform a phrase's polarity. Humorous sentences presenting ideas and criticism, occasionally using informal forms, have made their way to social media platforms like Twitter in almost every domain. Persian speakers likewise express their opinions through humorous tweets on Twitter. As one of the early efforts for detecting humor in Persian, the current research proposes a model by fine-tuning a transformer-based language model on a Persian humor detection dataset. The proposed model has an accuracy of 84.7% on the test set. Moreover, This research introduced a dataset of 14,946 automatically-labeled tweets for humor detection in Persian. Keywords: Humor Detection; Sentiment Analysis; Natural Language Processing; Deep learning; Persian language Article type: Research Article © The Author(s). Publisher: ICT Research Institute I. INTRODUCTION Using humor in speech or text is a way of expressing opinions that may have a different sense than what it seems because its creative characteristics prevent the correct understanding of the text. Social networks and online stores have increased with the growth of using the Internet. And it becomes common to publish opinions on these networks, write reviewes about products on websites, or post ideas about different domains. Owner of these websites are  Corresponding Author interested to analyzing the opinions and feelings of users in these cases. Because these opinions are sometimes mixed with humor, irony, and sarcasm, the effectiveness of sentiment analysis is affected. In the Cambridge Dictionary, humor is defined as “the ability to be amused by something seen, heard, or thought about, sometimes causing you to smile or laugh, or the quality in something that causes such amusement.” [1] Irony is defined as “the use of words that are the opposite of what you mean, as a way of being funny” and “a type of usually humorous expression in which you say the opposite of what you [ Downloaded from ijict.itrc.ac.ir on 2023-07-07 ] 1 / 7