Challenges
For humans, successful reading comprehension depends on the construction of an event structure that
represents what is happening in the text, often referred to as the situation model in cognitive
psychology. This situation model also involves the integration of prior knowledge with information
presented in text for reasoning and inference.
Language understanding requires a combination of relevant evidence, such as from contextual knowledge,
common sense or world knowledge, to infer meaning underneath. It also requires a constant update of
memory as reading progresses. In machine reading comprehension, a computer could continuously build and
update a graph of eventualities as reading progresses. Question-answering could, in principle, be based
on such a dynamically updated event graph.
Project Aims
The UKRI-funded Turing Ai Fellowship project aims to develop a knowledge-aware and event-centric
framework for NLU, in which event graphs are built as reading progresses; event representations are
learned with the incorporation of background knowledge; implicit knowledge is derived by performing
reasoning over event graphs; and the comprehension model is developed with built-in interpretability and
robustness against adversarial attacks.
Impact
Since spoken and written communication plays a central part in our daily work and life, the proposed
framework will have a profound impact on a variety of application areas, including drug discovery,
intelligent virtual assistants, automated customer services, smart home, and question-answering in the
finance and legal domains, benefiting industries such as healthcare, finance, law, insurance and
education.
