The Macquarie NLP group carries out research in Natural Language Processing and related areas like Machine Learning, both foundational work and applications to other domains. Some of the research themes are as follows:
NLP and privacy. NLP technologies can place individuals’ privacy at risk, either from being included in training data for NLP systems or from inferences that can be made by NLP systems. We’re looking at understanding privacy leakage in those contexts and at how to support privacy for individuals.
NLP and security. We examine and identify various attacks that can be made against NLP systems, such as adversarial attacks designed to disrupt their operations and backdoor attacks that enable undesirable behaviour. Our aim is to develop robust defense mechanisms to protect NLP systems against such vulnerabilities.
NLP, norms and neuro-symbolic computation. These components work together to create systems that understand natural language while following societal rules. Norms define expected behaviours, which NLP models can interpret with human-in- the-loop involvement. Neuro-symbolic computation enhances this by combining neural networks and symbolic reasoning to manage structured knowledge, like legal or ethical norms. This integration improves decision-making, accuracy, and transparency in NLP applications.
NLP and artificial agents in team contexts. We work with colleagues from Psychology who are interested in how humans behave in team contexts, both all-human teams and human-AI teams.
NLP of medical texts. We are exploring the application of NLP techniques on medical texts. We pay special emphasis on finding methods to help the practice of evidence-based medicine. For example, we have developed a corpus of medical questions and their answers and pointers to clinical evidence, and we have participated in the BioASQ challenges. More recently we have started looking at multimodal information (text and images).
Information seeking for task-oriented conversational agents. We are looking into developing methods that would help a conversational agent to ask questions to the user who wants to solve a complex task such as design an optimisation problem.
NLP for Social Good. We are developing NLP techniques to enhance the understanding and generation of human language in socially relevant contexts, focusing on developing innovative representation learning methods that enable machines to comprehend complex human communication. By exploring the intersection of NLP and social computing, we aim to contribute to the development of responsible NLP technologies that support effective and trustworthy interactions in various applications, fostering positive societal impact.
Multimodality. We study the interdisciplinary domain of integrating textual processing with speech and visual modalities. This holistic approach enhances our understanding of human communication and provides more effective solutions across diverse applications, contributing to our goal of fostering positive societal impact.
NLP for Low Resource Languages. We address the unique challenges of NLP in low-resource languages, emphasising fairness and inclusivity in technology development. Our research focuses on innovative techniques for language modeling and representation learning, ensuring equitable access to NLP tools for underrepresented languages. By prioritising these efforts, we aim to create robust and accessible NLP solutions that support diverse linguistic communities.
NLP for Social Media. We develop cutting-edge NLP techniques specifically designed for the intricate dynamics of social media. Our research addresses critical challenges such as sentiment analysis, misinformation detection, toxicity moderation, trend identification, and user engagement analysis. By harnessing advanced algorithms and machine learning models, we aim to enhance the accuracy and effectiveness of AI-driven insights, ultimately contributing to safer and more engaging social media environments.