Sarcasm detection is a challenging prediction task because of the myriad ways in which sarcasm may be expressed. Recent work in sarcasm detection uses prompt tuning to detect contradictions underlying sarcastic text.

School

Computer Science and Engineering

Research Area

Natural language processing | Machine learning

The project will involve developing a new approach to detect sarcasm. Recent work here [1]. The supervisory team consists of lecturer in natural language processing (NLP) and an incoming associate lecturer who will both closely work with the candidate.

The candidate will have strong experience in training deep learning models.

The candidate will learn how a typical NLP project is executed. The candidate will also become familiar with fine-tuning large language models (LLMs). The project may involve creation of new datasets, use of resources or extension of existing libraries to suit the task, among other research tasks. The candidate will be expected to submit well-documented code and a report. The report will summarise the approach, results (quantitative and qualitative) will be expected, so as to be submitted to an NLP research venue. The project has potential to be extended as an honours project.

  1. https://aclanthology.org/2023.eacl-main.25/ ; Accessed on 21st August, 2023