Hao Zhu

I am a third-year doctoral student at Language Technologies Institute, Carnegie Mellon University. I am fortunately advised by Graham Neubig and Yonatan Bisk. Before joining CMU, I enjoyed four fantasic undergrad years at Tsinghua University, advised by Zhiyuan Liu. I also worked closely with Jason Eisner and Matt Gormley. You can find my cv here (2019/3).

My ultimate goal is to understand human intelligence. Believing in Feynman's famous quote, "What I cannot create, I do not understand.", I am working on teaching Machine Learning models to gain human intelligence. More specifically, I am currently interested in teaching machines to speak human language, as well as to do human-level logical reasoning. I also have broad interests in other cognitive science fields.

Email: {last.name}{first.name} [at] cmu.edu

Twitter: @_Hao_Zhu

GitHub: ProKil

[Full Publication List & Preprints]   [Google Scholar]



Volunteer/Review Assistant: IJCAI 2017/2018

Research Hightlights

Unified Grammar Induction

We demonstrate that context free grammar (CFG) based methods for grammar induction benefit from modeling lexical dependencies. This contrasts to the most popular current methods for grammar induction, which focus on discovering either constituents or dependencies. Previous approaches to marry these two disparate syntactic formalisms (e.g. lexicalized PCFGs) have been plagued by sparsity, making them unsuitable for unsupervised grammar induction. However, in this work, we present novel neural models of lexicalized PCFGs which allow us to overcome sparsity problems and effectively induce both constituents and dependencies within a single model. Experiments demonstrate that this unified framework results in stronger results on both representations than achieved when modeling either formalism alone.

The Return of Lexical Dependencies: Neural Lexicalized PCFGs Transactions of the Association for Computational Linguistics (2020).