Commonsense reasoning refers to the ability of capitalising on commonly used knowledge by most people, and making decisions accordingly. This process usually involves combining multiple commonsense facts and beliefs to draw a conclusion or judgement. While human trivially performs such reasoning, current Artificial Intelligence models fail, mostly due to challenges of acquiring relevant knowledge and forming logical connections between them. This project aims to develop and evaluate machine learning models for commonsense reasoning, with question answering as the key application.
- Improve a state-of-the-art deep learning model for commonsense reasoning.
- Evaluate the state-of-the-art commonsense reasoning models for question answering.
- Storks, Shane, Qiaozi Gao, and Joyce Y. Chai. "Commonsense reasoning for natural language understanding: A survey of benchmarks, resources, and approaches." arXiv preprint arXiv:1904.01172 (2019): 1-60.
- Moghimifar, Farhad, Lizhen Qu, Yue Zhuo, Mahsa Baktashmotlagh, and Gholamreza Haffari. "COSMO: Conditional SEQ2SEQ-based Mixture Model for Zero-Shot Commonsense Question Answering." COLING (2020).
Students should have excellent grades in machine learning and relevant math courses. Strong programming skills are essential. Preference will be given to students who have previous practical research experiences on deep learning, natural language processing, and commonsense reasoning. Students with an interest in pursuing PhD research or careers in research are especially encouraged to apply.