Xiaodan Zhu
name stamp

I am an Assistant Professor of the Department of Electrical and Computer Engineering at Queen's University, leading the Text Analytics and Machine Learning Lab (TAML). I received my Ph.D. from the Department of Computer Science of the University of Toronto in 2010 and my Masters of Engineering from the Department of Computer Science and Technology of Tsinghua University in 2000. Before, I have also worked with various industry labs, either as a full-time researcher (Intel's China Research Center), visiting scholar (Microsoft Research Asia), or as a research intern (Google Inc., IBM T.J. Watson Research Center).
Assistant Professor, ECE, Queen's University
Adjunct Professor, EECS, University of Ottawa
Natural Language Processing, Social Media, Medical Informatics, Big Data;

Deep Learning, Machine Learning.

Some highlights
Our ACL-2017 tutorial slides on Deep Learning for Semantic Composition are available here: [Link]. Video will be made available later once ACL releases it. Also, Edward and I may make some small updates to the slides.
Tree LSTM: we proposed tree LSTM, which was published at the International Conference on Machine Learning; see the paper at [ICML], or at [ArXiv] for an earlier version.
Sentiment Analysis for Tweets: we built top-ranked systems in a number of SEMEVAL competitions on sentiment analysis for tweets, see here for details, and see publications for papers.
Our paper on semantic compositionality received the Adam Kilgarriff *SEM Best Paper Award for Lexical Semantics at the Joint Conference on Lexical and Computational Semantics (*SEM), Denver, Colorado. [paper]

I will give a keynote talk at the WISDOM Workshop at KDD-2017 on neural network models for contrasting meaning representation and sentiment composition.
I will give a tutorial (together with Edward Grefenstette, Google DeepMind) on "deep learning for semantic composition" at ACL-2017. [slides][description]

Congratulations to Qian! We developed a model (alpha) that is ranked among the top in RepEval-17 Shared Task [link]. The neural networks attempt to encode the meaning of sentences into fixed-sized vectors; the effectiveness of the models is evaluated with Natural Langauge Inference. [paper]
From July 1, 2016 to June 30, 2019, I will serve a three-year term (part-time) for NSERC (Natural Sciences and Engineering Research Council of Canada) as an Evaluation Group (EG) Member to evaluate Discovery Grants applications in the area of Computer Science (link).

Employing our Tree-LSTM, Qian Chen's recent work on Natural Language Inference (paper link) has achieved the the state-of-the-art results on the Stanford Natural Language Inference benchmark (link). (Qian is a vistinting student I co-supervised. He will visit me and Diana Inkpen at University of Ottawa.)

Our paper exploring neural nets for semantic compositionality, "DAG-Structured Recurrent Neural Networks for Semantic Compositionality" has been published at NAACL-2016 (link).

On Feb. 22, 2016, I gave a talk at Ottawa Machine Learning Meetup on "Deep Learning for Text Mining: Case Studies on Social Media and Medical Text" (talk announcement).

Between July and November 2015, I gave a couple of talks at the University of Toronto, McGill University, UMass medical school, University of Ottawa, and Baidu Inc. about our recent efforts on neural networks for semantics.
Our paper on semantic compositionality received the Adam Kilgarriff *SEM Best Paper Award for Lexical Semantics at the Joint Conference on Lexical and Computational Semantics (*SEM), Denver, Colorado. [paper]

Our work on long short-term memory over tree structures has been accepted to International Conference on Machine Learning. Here is the ICML version with updated results [ICML]. The older arXiv version can be found at [ArXiv];
Our paper presented at NIPS-2014 Workshop on Representation and Learning Methods for Complex Outputs [paper]
Gave a tutorial at EMNLP-2014 in Doha on "Sentiment Analysis of Social Media Texts."
In Semeval-2014 Task 9: Sentiment Analysis in Twitter, we ranked first in five of the ten subtask-domain combinations among about 40 teams. In Semeval-2014 Task 4: Aspect Based Sentiment Analysis, our models ranked first in three of the six subtasks among about 30 teams. [paper-1][2]
Our ACL-2014 paper on sentiment analysis (negation modeling). [paper]
Our paper on machine translation of sentiment, presented at EACL-2014. [paper]