RESEARCH

  • Sentiment Analysis by Capsules                               Sep. 2017 – Nov. 2017
Proposed RNN-Capsule, a capsule model based on Recurrent Neural Network for sentiment analysis. RNN-Capsule is capable of outputting words with sentiment tendencies reflecting capsules’ attributes without using any linguistic knowledge. The words well reflect the domain specificity of the dataset. To the best of our knowledge, this is the first capsule model for sentiment analysis. Accepted by WWW’18.
  • Explainable Sentiment Analysis                                 Feb. 2017 – Present
Explainable Sentiment Analysis (ESA) is an emerging field that tries to make sentiment analysis more understandable to users. The objective of ESA is to ensure that a model explains its rationale behind certain analysis.
  • Joint Model                                                                              Apr. 2016 – Feb. 2017
  • Proposed a joint model that performs aspect detection and sentiment classification simultaneously. The model has a shared representation layer for the two tasks, and the classification losses for the two tasks are optimized jointly.
  • Aspect-level Sentiment Analysis                              Jan. 2016 – May 2016
Proposed attention-based LSTM for aspect-level sentiment classification. The models are able to attend different parts of a sentence when different aspects are concerned. Accepted by EMNLP’16.
  • Phrase Extraction                                                               Mar. 2014 – June 2014
Proposed boundary and stickness algorithm for building better features, which can perform extracting phrase in Sina Weibo better.  Graduation Project. National Invention Patent (CN201410265383.5).
  • Price of Commercial House                                          June 2013 – Dec. 2013
Used factor analysis and partial least squares to explore how factors affect the price of commercial house. Accepted by Future and Development’ 14.