I am currently a postdoctoral research fellow in School of IT, Deakin University, Australia. I got my PhD in 2019 from Monash University, Australia. My research area includes natural language processing, machine learning and biomedical data mining. I received my M.Eng from Zhengzhou University in 2015 and B.Eng from Henan University in 2012, respectively. I am now working closely withLongxiang Gao, Reza Haffari, Lan Du, and Wray Buntine.
Trang Vu, Ming Liu, Gholamreza Haffari, Dinh Phung. Learn to Active Learn by Dreaming, In Annual Conference of the Association for Computational Linguistics 2019.
Michelle Ananda-Rajah, Diva Baggio, Trisha Peel, Anton Peleg, Gholamreza Haffari, Ming Liu, Christoph Bergmeir. Closing the gap in surveillance and audit of invasive mold diseases for antifungal stewardship using artificial intelligence, In Journal of Clinical Medicine 2019.
Ming Liu, Gholamreza Haffari, Wray Buntine. Learning to actively learn for neural machine translation, In Conference of Natural Language Learning 2018.
Michelle R Ananda-Rajah, Ming Liu et al. Deep learning for recognition of invasive fungal disease from chest computed tomography in haematology-oncology patients, In International Immunocompromised Host Society Symposium, Athens 2018.
Ming Liu, Gholamreza Haffari, Wray Buntine. Learning How to Actively Learn: A Deep Imitation Learning Approach, In Annual Conference of the Association for Computational Linguistics 2018.Video Link
Ming Liu, Gholamreza Haffari, Wray Buntine, Michelle Ananda-Rajah. Leveraging Linguistic Resources for Improving Neural Text Classification, In Proceedings of the Australian Language Technology Association 2017, pp: 34-42.
Ming Liu, Gholamreza Haffari, Wray Buntine. Learning cascaded latent variable models for biomedical text classification, In Proceedings of the Australian Language Technology Association 2016, pp: 139-143
Ming Liu, Hongying Zan, Jun Liang. Key Sentiment Sentence Extraction In News Using SVM and RNN, Journal of Shandong University, 2014, pp: 46-50.
Automatic long text summarization (May 2019-Present) Overview: My post-doc project, aiming at automatic summarization for scitific papers and books.
Multi-modal topic modeling (Jan 2019-Present) Overview: Topic models are widely used in the social sciences and humanities to analyze text collections. This project aims to build topic models with multiple modalities, expecially from text and images. We use pre-trained word embeddings to inform how a topic focuses on words, and docment level image embeddings to inform how a document focuses on topics. Our methods improves traditional topic models significantly, which allows to discover more informed and focused topics with more representative words, leading to better modelling accuracy and topic quality.
Clinical text mining for fungal diagnosis support (Oct 2015–Present) Overview: This project (http://www.fungalai.com) built on neural networks in a world first, can make real-time surveillance of fungal diseases possible using: Natural language processing of chest CT reports, Deep learning based image analysis of chest CT scans, and an expert system that integrates lab and drug information.I developed a cascaded latent variable model and a neural model for clinical text classification, the neural model achieved state of art performance, and could support doctors better identify fungal disease.
Cardiorespiratory synchronization analysis for atheletes in Ironman competetion (Dec 2018-Jan 2019) Overview: This project was done in Deakin University, where ECG and respiratory time series signals were recorded for a group of atheletes before and after an Ironman competetion. We conducted EMD and Hilbert Huang Transform on both signals, resulting in the synchronisation of the two phases. It shows ECG and repiratory signals are more synchronised after the Ironman competetion. This finding is potential for fatigue dection for players in the future.
Weak supervision and active learning for deep neural models (Sep 2017- Feb 2019) Overview: Manual annotation is very expensive in the real world for many NLP tasks such as clinical record classification, named entity recognition, machine translation, active learning is a type of semi-supervised learning where a human expert give labels to those specific data with which the machine learning models can get improved performance. In this project, instead of using traditional heuristic-based methods, we propose several weakly supervised and active learning strategies for deep neural models, the strategies includes using structured information, learning for priors, and learn active learning policies. These strategies are quite effective in the real word environment.
Construction of a large-scale Modern Chinese Knowledge Base (Sep 2012-Aug 2014) Overview: I was a member of this project which derived from the Natural Science Foundation of China. The main task was to build a large-scale modern Chinese Knowledge Base based on existing resources including Modern Chinese Dictionary(5th Edition), Chinese Grammar Knowledge Base(GKB), Chinese Function Word Usage Knowledge Base, etc.
Public Opinion Analysis toward Online Communities in Zhengzhou Uni (Nov 2013-Nov 2014) Overview: I was the leader of this project, which was one of 2013 Postgraduate Innovation Projects in Zhengzhou Uni. My work was system design and opinion analysis using some machine learning methods(SVM, NB).
Co-funded Monash Graduate Scholarship (2015-2018)
China National Postgraduate Scholarship (Sep 2014)