My research interest is primarily text mining, information retrieval system, mathematical optimization. I am currently working on a faceted entity ontology learning, gene-based information retrieval system, and relational pattern mining. My previous researches were mainly about topic modeling, community detection, and learning to rank.

Topic-based Academic Information Retrieval System

Supervised by Prof. Xinbing Wang • Sep, 2015 — Jun, 2016

I am currently developing a topic-based search engine to improve academic search experience.

The objectives of this system are mainly:

  • Return paper search results based on both word-level and topic-level similarity with user’s query.
  • Rank papers according to their influence scores as well as their relevance to the query.
  • Visualize the topic distribution of each paper and topic evolution among the whole corpus.

Modeling Academic Influence in Scientific Literatures

Supervised by Prof. Xinbing Wang • Apr, 2015 — Sep, 2015

Scientific articles are not born equal. Some generate an entire discipline while others make relatively fewer contributions. We studied the problem of identifying those important articles and understanding how they influence others.

The main contributions of this work include:

  • Devised a generative model to utilize both the textual content and citation information in scientific literatures.
  • Proposed a fast inference algorithm to learn this generative model based on collapsed Gibbs Sampling.
  • Introduced a novel quantitative metric named J-Index to model academic influence in scientific literatures.
  • Designed experiments on a collection of over 420,000 research papers to validate the effectiveness of J-Index.

This work was submitted to AAAI 2016.

Exponential Interest Aggregation for Hop-by-hop Congestion Control in Named Data Networking

Supervised by Prof. Weijia Jia • Apr, 2015 — Jul, 2015

Named Data Networking (NDN) is a Future Internet architecture aiming to satisfy the new requirement of content communication. We studied the problem of congestion control when designing NDN transport protocol.

The main contributions of this work include:

  • Established the Interest aggregation state transition framework in Named Data Networking.
  • Proposed Exponential Interest Aggregation (EIA), an adaptive forwarding strategy addressing NDN congestion control problem in a hop-by-hop manner.
  • Analyzed the effectiveness of EIA algorithm mathematically under the Interest aggregation state transition framework.
  • Conducted simulation to evaluate the performance of EIA and showed that EIA improves average delay by 13%, average number of retransmission by 25%, and cache hit ratio by 61%.

This work was submitted to IEEE INFOCOM 2016.