Recently large-scale pre-trained language models, such as GPTs or BART have achieved successful performances in generating grammatically fluent text and capturing knowledge present in training corpus. However, it is pointed out that generating multi-sentence text (i.e., stories, narrations) with internal logical consistency is still far from being solved , with existing solutions merely scratching the surface in simple settings .
Honours and Minor Thesis projects
Medical guidelines provide human experts with steps on conducting diagnosis. While existing online knowledge graphs such as Unified Medical Language System (UMLS)  provide a wide coverage of various biomedical entities, their coverage is limited in certain specific domains such as Ophthalmology. In this project we will build an automated system that convert the offline medical guidelines of Glaucoma  and Diabetic Retinopathy  to machine readable knowledge graphs.
Knowledge graphs (KGs) play an important role in Natural Language Processing (NLP) and store information in a structured and machine-accessible way . They are used in different domains and fields. Generating texts from KGs is an important NLP task which transforms graph into natural language. For example, given a subgraph from a KG, we aim to get a corresponding description. Texts are easier to understand for human than graph-structured data.
In Ophthalmology, patients are routinely scanned with multiple retinal imaging systems that provide complementary information to the clinicians. However, unlike other specialties, the images are not analysed by a radiologist and the treating ophthalmologist or optometrist is expected to analyse this data on their own. This is extremely time consuming, and difficult to achieve in clinical settings. Thus, AI models for disease detection have been extremely popular.
Machine learning models have significantly improved the ability of autonomous systems to solve challenging tasks, such as image recognition, speech recognition and natural language processing. The rapid deployment of such models in safety critical systems resulted in an increased interest in the development of machine learning models that are robust and interpretable.
Planning is the reasoning side of acting in Artificial Intelligence. Planning automates the selection and the organization of actions to reach desired states of the world as best as possible. For many real-world planning problems however, it is difficult to obtain a transition model that governs state evolution with complex dynamics.
Social media has become a dominant means for users to share their opinions, emotions and daily experience of life. A large body of work has shown that informal exchanges such as online forums can be leveraged to supplement traditional approaches to a broad range of public health questions such as monitoring depression, domestic abuse, cancer, and epidemics.
Linguistic phenomena have emerged and evolved over the span of thousands of years leading to many variations. Through this evolution, many linguistic structures and compositions have emerged or disappeared. In this project we will deploy an information-theoretic perspective to investigate the connections between linguistic phenomena (survival), and communication efficiency and emergence.
Machine Learning (ML) models are deployed in many safety-critical systems (such as self-driving cars, cancer detection software, etc.) to improve human decision-making. Therefore, safety is central to the success of many human-in-the-loop systems that deploy such ML models.
Since the 1990s, researchers have known that commonly-used public-key cryptosystems (such as RSA and Diffie-Hellman systems) could be potentially broken using efficient algorithms running on a special type of computer based on the principles of quantum mechanics, known as a quantum computer. Due to significant recent advances in quantum computing technology, this threat may become a practical reality in the coming years. To mitigate against this threat, new `quantum-safe’ (a.k.a.