Sarcasm, a form of figurative language (FL), often conveys meaning contrary to its literal interpretation in all social media discussion forums and chats, adding complexity to natural language understanding. This paper focuses on detecting sarcasm in Assamese, a low-resource language by utilizing advanced deep learning methods. In particular, we employ bidirectional encoder representation transformers (BERT), a pre-trained transformer-based classifier, to accurately identify the sarcastic undertones within sentences. Additionally, we conducted a comparative analysis with two other deep learning models, i.e., Long Short-Term Memory (LSTM) and Bidirectional Long Short-Term Memory (Bi-LSTM) to evaluate the performance of our proposed BERT transformer-based framework. Upon evaluating the classifiers’ performances in our dataset, BERT demonstrated a better accuracy and efficiently distinguished between sarcastic and non-sarcastic remarks, showcasing its potential as a powerful tool in sarcasm detection.