- Quick Links for
To complement the individual research projects featured in the Research Projects directory, we have created a separate web site to provide an overview of our publications. On this web site, you can find the topics that we have worked on and the papers that we have published. You can see our most cited (according to Scopus and Web of Science) and most recent work on each topic, and how our efforts on the topics evolve with time. Please proceed to the web site by clicking on the following picture.
The web site is created automatically based on the collection of publications by our faculty members and students that is kept at the HKUST Institutional Repository. The key technology behind it is a hierarchical topic detection method named hierarchical latent tree analysis (HLTA).
HLTA is developed by a research team led by Prof. Nevin L. Zhang and the software package is kept at GitHub: https://github.com/kmpoon/hlta. HLTA first learns a latent variable probabilistic graphical model based on word co-occurrences, then builds a topic tree from the model, and finally uses the tree to index the documents. Here is the part of the structure of the model learned from a collection of articles published at New York Times.
In the era of big data, people are often confronted with large collections of unstructured text documents. Gaining an overview of the contents is important to the realization the potential values of such data. However, this is difficult task. HLTA is developed to meet the challenge. It has been previously applied on research papers published in recent year at top AI and Machine Learning venues. The results are shown at aipano.cse.ust.hk to help researchers track research trends and identify papers to read, and help conference organizers and journal editors identify reviewers for submissions.