About Me

Welcome! I am Tiancheng Zhao (赵天成). I received my Ph.D. degree in May 2019, from Language Technologies Institute (LTI) , Carnegie Mellon University, advised by Maxine Eskenazi. Also, I am a member of Dialog Research Center (Dialrc) . Before becoming a PhD student at Carnegie Mellon University, I graduated from the Master of Language Technologies (MLT) program at LTI in 2016, advised by Maxine Eskenazi and and Alan W Black. Prior to that, I obtained my bachelor degree in Electrical Engineering from University of California, Los Angeles and worked on speech signal processing, advised by Abeer Alwan.

My interest in general lies in deep learning and its application in end-to-end dialog systems and natural language processing. My current research focuses on 1) modeling long-term context in human-human and human-machine conversations and 2) end-to-end trainable dialog models that can conduct both goal driven and non-goal driven conversations via deep (reinforcement) learning.

I successfully defended my PhD! [PDF]

You can also find me at GitHub, LinkedIn and Google Scholar.



Carnegie Mellon University

- Ph.D. in Language and Information Technology, Computer Science, 2016 - 2019

- GPA 4.15

Carnegie Mellon University

- Master of Language Technologies, Computer Science, 2014 - 2016

- GPA 3.97

University of California, Los Angeles

- Bachelor of Science in Electrical Engineering, 2010 - 2014

- Overall GPA 3.86, Major GPA 4.0

- Graduated with Summa Cum Laude (Top 5% of the School of Engineering)

- Outstanding Bachelor of Science Award, 1 out of 152 students in Electrical Engineering Department

Selected Publications

PhD Dissertation

Tiancheng Zhao Learning to Converse with Latent Actions 2019


1. Jianheng Tang, Tiancheng Zhao, Chenyan Xiong, Xiaodan Liang, Eric P. Xing, Zhiting Hu Target-Guided Open-Domain Conversation, long paper at ACL 2019. [Code and Data]

2. Shikib Mehri, Evgeniia Razumovsakaia, Tiancheng Zhao and Maxine Eskenazi Pretraining Methods for Dialog Context Representation Learning, long paper at ACL 2019.

3. Tiancheng Zhao, Kaige Xie and Maxine Eskenazi Rethinking Action Spaces for Reinforcement Learning in End-to-end Dialog Agents with Latent Variable Models, long paper at NAACL 2019. [Oral] [Code and Data]

4. Weiyan Shi, Tiancheng Zhao, Zhou Yu Unsupervised Dialog Structure Learning, long paper at NAACL 2019.


1. Tiancheng Zhao and Maxine Eskenazi Zero-Shot Dialog Generation with Cross-Domain Latent Actions, in proceedings SIGDIAL2018 as a long paper. [Oral] [Best Paper Award] [Code and Data] [Slides]

2. Kyusong Lee, Tiancheng Zhao, Alan W Black and Maxine Eskenazi DialCrowd: A toolkit for easy dialog system assessment, in proceedings SIGDIAL2018 as a demo paper.

3. Jiaping Zhang, Tiancheng Zhao, Zhou Yu Multimodal Hierarchical Reinforcement Learning Policy for Task-Oriented Visual Dialog, in proceedings SIGDIAL2018 as a long paper. [Oral]

4. Tiancheng Zhao, Kyusong Lee and Maxine Eskenazi, Unsupervised Discrete Sentence Representation Learning for Interpretable Neural Dialog Generation, in proceedings ACL2018 as a long paper. [Oral] [Code and Data] [Slides]

5. Zhiting Hu, Zichao Yang, Tiancheng Zhao, Haoran Shi, Junxian He, Di Wang, Xuezhe Ma, Zhengzhong Liu, Xiaodan Liang, Lianhui Qin, Devendra Singh Chaplot, Bowen Tan, Xingjiang Yu, Eric Xing, Texar: A Modularized, Versatile, and Extensible Toolbox for Text Generation, in Proceedings of Workshop for NLP Open Source Software (NLP-OSS) 2018.

6. Yijun Xiao, Tiancheng Zhao, William Yang Wang, Dirichlet Variational Autoencoder for Text Modeling, arXiv preprint arXiv:1811.00135 2018.


1. Tiancheng Zhao, Allen Lu, Kyusong Lee and Maxine Eskenazi, Generative Encoder-Decoder Models for Task-Oriented Spoken Dialog Systems with Chatting Capability , in Proceedings of SIGDIAL 2017 as a long paper. [Oral] [ Live demo ] [Presentation Video]

2. Kyusong Lee, Tiancheng Zhao, Yulun Du, Edward Cai, Allen Lu, Eli Pincus, David Traum, Stefan Ultes, Lina M. Rojas Barahona, Milica Gasic, Steve Young and Maxine Eskenazi,  DialPort, Gone Live: An Update After A Year of Development , in Proceedings of SIGDIAL 2017 as a demo paper.

3. Tiancheng Zhao, Ran Zhao and Maxine Eskenazi, Learning Discourse-level Diversity for Neural Dialog Models using Conditional Variational Autoencoders, in Proceedings of ACL 2017 as a long paper. [Oral] [Code and Data]

4.Kyusong Lee, Tiancheng Zhao, Stefan Ultes, Lina Rojas-Barahona, Eli Pincus, David Traum and Maxine Eskenazi, An Assessment Framework for DialPort, in Proceedings of IWSDS 2017 as a short paper.


1. Tiancheng Zhao and Maxine Eskenazi, Towards End-to-End Learning for Dialog State Tracking and Management using Deep Reinforcement Learning, in Proceedings of SIGDIAL 2016 Conference. [Oral] [Best Paper Nominee Award] [Presentation Video]

2. Tiancheng Zhao, Kyusong Lee and Maxine Eskenazi, The DialPort Portal: Grouping Diverse Types of Spoken Dialog Systems, in Proceedings of Second Workshop on Chatbots and Conversational Agent Technologies (WOCHAT). [Oral]

3. Andrew Wilkinson, Tiancheng Zhao and Alan W Black, Deriving Phonetic Transcriptions and Discovering Word Segmentations for Speech-to-speech Translation in Low-resource Settings, in Proceedings of the Interspeech 2016.

4. Tiancheng Zhao, ReinForest: Multi-Domain Dialogue Management Using Hierarchical Policies and Knowledge Ontology, Technical Report.

5. Tiancheng Zhao and Mohammad Gowayyed, Algorithms for Batch Hierarchical Reinforcement Learning , arXiv preprint arXiv:1603.08869 (2016).


1. Tiancheng Zhao, Alan W Black, and Maxine Eskenazi, An Incremental Turn-Taking Model with Active System Barge-in for Spoken Dialog System, in Proceedings of SIGDIAL 2015 Conference. [Oral] [Presentation Video]

2. Tiancheng Zhao and Maxine Eskenazi, Human-System turn taking analysis for the let’s go bus information system , The Meeting of the Acoustical Society of America, Pittsburgh, May 2015.


DialPort: a Distributed Multi-domain SDS Portal (2015-Present) [Click Here]

Innovation in spoken dialog technology requires a shared platform where new ideas can flourish, new dialog problems can come to the surface and communities can form to leverage different parts of the solution for better spoken dialog systems.

DialPort will allow for new application ideas and newly available technologies to be shared which will in turn attract real users, creating real data. The creation of this platform is the goal of DialPort. Therefore, we need a community website and portal that can help the community to create these platforms for future applications and provide a stream of real users. DialPort will work with the spoken dialog community to create and share real user applications that serve as data streams and research platforms.


Teaching Assistant, 10-707 Deep Learning, Fall 2017.

Teaching Assistant, 10-703 Deep Reinforcement Learning , Spring 2018

Teaching Assistant & Recitation Leader, 10-601 Introduction to Machine Learning , Fall 2019

Professional Activities

Reviewer & Program Committee: EMNLP 19, NAACL 19, AAAI 19, ACL 18, EMNLP 18, IWSDS 17, DSTC-6 17, YRRSDS 16,17