Research Projects



Reinforcement Learning framework for temporal Question-Answer reasoning(June 2017 - Present)

Under the supervision of Prof. William Wang

We defined a novel temporal Question-Answer problems. Then we built a large temporal dataset of document-question-answer pairs for large scale deep learning training. Finally, based on Reading Comprehension, we invented a well-designed reinforcement learning framework to solve temporal QA problems.





Evolving Scholarly Networks : Experiments, Modeling and Analysis(June 2016 - February 2017)

Under the supervision of Prof. Xinbing Wang and Dr. Luoyi Fu

We provide a new and novel evolving scholarly model that jointly captures both intra and inter correlations of papers, authors and topics during the evolving process. We further validate that our model can accurately reproduce the global and local structures of real scholarly networks.






Maximum Value Matters: Finding Hot Topics in Scholarly Fields(Sep 2016 - June 2017)

Under the supervision of Prof. Xinbing Wang and Dr. Luoyi Fu

We draw the Topic Map which show the relationships between topics and use k-core analysis to insight the skeleton of the whole CS field. A series of inner and inter factors are also proposed to determine the state of a topic. On the other hand, we examine the co-evolving relations between each part of topics, especially the author factor’s influences.



FPRank: Learning to Predict the Future Popularity Rankings of Scientific Publications(Aug 2016 - Sept 2016)

Under the supervision of Prof. Xinbing Wang and Dr. Luoyi Fu

A new ranking framework ZeroRank is proposed to tackle two problems: 1) Future Citation Ranking problem which aims to predict the ranking of scientific publications by their predicted future popularity and 2) Zero Citation Ranking problem which is to rank only newly published literatures without citation information. ZeroRank considers the dynamic time-aware representations of papers and takes the future citation count as the ground truth future popularity. LTR technique is leveraged and time is taken as query in ZeroRank to train ranking models.


Acemap : Academic Map System(Mar 2016 - Present)

Under the supervision of Prof. Xinbing Wang.

Acemap is a academic map system which aims to process the big scholarly data, analyze the citation network and visualize the relationship among papers to help researchers grasp the academic big picture more conveniently and more intuitively.






Acenap:Academic Network Analytic Platform(Fabruary 2017 - Present)

Under the supervision of Prof. Xinbing Wang and Dr. Luoyi Fu

Acenap, abbreaviated as Acedamic Network Analytic Platform, is an online platform developed for the purpose of facilitating research for those who are interested in the areas regarding social and academic networks. Particularly, it includes: 1) Some generating models that are commonly used in social network analysis. 2) Some algorithms that are either proposed by our own research group, or adopted from the existing state-of-the-art work in social-related areas. 3) A collection of datasets from different social networks and academic networks. Some of them are crawled by our own research group, while others are collected elsewhere but processed with our own research usage.