Pei-Hao Su (蘇培豪)
I obtained my PhD in Statistical Spoken Dialogue Systems at Cambridge University supervised by Professor Steve Young and Dr. Milica Gašić.
My PhD was funded by Taiwan Cambridge Scholarship, and a member of Queens' College. I was also fortunate to work with Dr. Yuandong Tian at Facebook AI Rsearch in the last summer of my PhD.
My primary research interest lies mainly in the intersection of deep learning and reinforcement learning for spoken dialogue systems, aiming to set up an unified framework for language understanding/replying and a system that is capable of scaling to broad domains and directly learn from human user interaction.
Prior to Cambridge, I received my Bachelor and Master degree in EECS from National Taiwan University, where I joined the Digital Speech Processing Laboratory held by Professor Lin-shan Lee and focused on statistical dialogue modelling for personalising pronunciation training in a dialogue game.
Curriculum Vitae [] (last update: February 2021)
Awards
- Company of the Year, Cambridge Computer Lab Ring, 2019
- Best Student Paper Award, ACL, 2016
- W. G. Collins Endowment Fund Award, Cambridge University Engineering Dept., 2016
- Interspeech Travel Grant, ISCA, 2015
- Taiwan Cambridge Scholarship, Cambridge Trust & MOE of Taiwan, 2014-17
- Government Scholarship for Studying Abroad (waived), MOE of Taiwan, 2014
- ICASSP Travel Grant, National Science Council of Taiwan, 2013
- Advanced Speech Technologies Scholarship, NTU EECS Department, 2012
- Best Intern, Trend Micro Inc., 2012
- Dean's List, NTU, 2008
Education & jobs
Advisor: Dr Yuandong Tian
Advisor: Professor Steve Young
Cloud storage division
Advisor: Professor Lin-shan Lee
Publication
First author and main work only, see also my full list of publications in my Google Scholar page
Thesis
Journal Article
-
Sample efficient deep reinforcement learning for dialogue systems with large action spaces
IEEE Transaction on Audio, Speech and Language Processing, To appear
Gellért Weisz, Paweł Budzianowski, Pei-Hao Su, and Milica Gasic
-
“Reward estimation for dialogue policy optimisation”
Computer Speech & Language, Volume 51, September 2018, Pages 24-43 [paper ] [bib ]
Pei-Hao Su, Milica Gašić, and Steve Young
-
“A Recursive Dialogue Game for Personalized Computer-Aided Pronunciation Training”
IEEE Transaction on Audio, Speech and Language Processing, January 2015 [paper ] [bib ]
Pei-hao Su, Chuan-hsun Wu, and Lin-shan Lee
Preprint
Conference Paper
-
“ConveRT: Efficient and Accurate Conversational Representations from Transformers”
EMNLP 2020
[paper ]
Matthew Henderson, Iñigo Casanueva, Nikola Mrkšić, Pei-Hao Su, Tsung-Hsien Wen, and Ivan Vulić
-
“A Repository of Conversational Datasets”
ACL NLP for ConvAI workshop 2019
[paper ]
Matthew Henderson, Paweł Budzianowski, Iñigo Casanueva, Sam Coope, Daniela Gerz, Girish Kumar, Nikola Mrkšić, Georgios Spithourakis, Pei-Hao Su, Ivan Vulić, and Tsung-Hsien Wen
-
“Training Neural Response Selection for Task-Oriented Dialogue Systems”
ACL NLP for ConvAI workshop 2019
[paper ]
Matthew Henderson, Ivan Vulić, Daniela Gerz, Iñigo Casanueva, Paweł Budzianowski, Sam Coope, Georgios Spithourakis, Tsung-Hsien Wen, Nikola Mrkšić, and Pei-Hao Su
-
“A Benchmarking Environment for Reinforcement Learning Based Task Oriented Dialogue Management”
NIPS 2017 Symposium on Deep RL
[paper ]
Iñigo Casanueva*, Paweł Budzianowski*, Pei-Hao Su*, Nikola Mrkšic, Tsung-Hsien Wen, Stefan Ultes, Lina Rojas-Barahona, Steve Young and Milica Gašic
-
“Sample-efficient Actor-Critic Reinforcement Learning with Supervised Data for Dialogue Management”
SIGDIAL 2017 (Oral)
Pei-Hao Su, Paweł Budzianowski, Stefan Ultes, Milica Gašić and Steve Young
-
“On-line Active Reward Learning for Policy Optimisation in Spoken Dialogue Systems”
ACL 2016 (Oral)
[paper ] [bib ] [slides ] [data ] [video ] (Best Student Paper Award)
Pei-Hao Su, Milica Gašić, Nikola Mrkšić, Lina Rojas-Barahona, Stefan Ultes, David Vandyke, Tsung-Hsien Wen and Steve Young
-
“Multi-Domain Dialogue Success Classifiers for Policy Training”
ASRU 2015
David Vandyke, Pei-Hao Su, Milica Gašić, Nikola Mrkšić, Tsung-Hsien Wen and Steve Young
-
“Reward Shaping with Recurrent Neural Networks for Speeding up On-Line Policy Learning in Spoken Dialogue Systems”
SIGDIAL 2015
Pei-Hao Su, David Vandyke, Milica Gašić, Nikola Mrkšić, Tsung-Hsien Wen and Steve Young
-
“Learning from Real Users: Rating Dialogue Success with Neural Networks for Reinforcement Learning in Spoken Dialogue Systems”
Interspeech 2015 (Oral)
Pei-Hao Su, David Vandyke, Milica Gašić, Dongho Kim, Nikola Mrkšić, Tsung-Hsien Wen and Steve Young
-
“A Cloud-based Personalized Recursive Dialogue Game System for Computer-Assisted Language Learning”
SLaTE 2013 (Oral)
Pei-hao Su, Tien-han Yu, Ya-Yunn Su, and Lin-shan Lee
-
“NTU Chinese 2.0: A Personalized Recursive Dialogue Game for Computer-Assisted Language Learning”
SLaTE 2013 (Demo)
Pei-hao Su, Tien-han Yu, Ya-Yunn Su, and Lin-shan Lee
-
“A Recursive Dialogue Game Framework with Optimal Policy Offering Personalized Computer-Assisted Language Learning”
Interspeech 2013 (Oral)
Pei-hao Su, Yow-Bang Wang, Tsung-Hsien Wen, Tien-han Yu, and Lin-shan Lee
-
“A Dialogue Game Framework with Personalized Training using Reinforcement Learning for Computer-Assisted Language Learning”
ICASSP 2013 (Oral)
Pei-hao Su, Yow-Bang Wang, Tien-han Yu, and Lin-shan Lee
Thesis
-
“A Dialogue Game Framework Offering Personalized Pronunciation Training for Computer-Assisted Language Learning”
Master Thesis, National Taiwan University 2013
Pei-hao Su
Invited Talk
Tutorial
-
"Deep Learning for Conversational AI"
NAACL 2018 tutorial [Website]
Industy
-
"Pracitcal Approaches to Conversational AI"
ODSC 2018
-
"Pracitcal Approaches to Conversational AI"
South England NLP Meetup 2018
-
"Reward Estmiation for Dialogue Policy Optimisation"
General Motor Advanced Techinal Center Israel, 20 March 2017
-
“Practical Human-in-the-loop Reinforcement Learning in Spoken Dialogue Systems”
Apple Cambridge, 12 Oct 2016
-
“On-line Active Reward Learning for Policy Optimisation in Spoken Dialogue Systems”
Maluuba Research (Now Microsoft). 2 Aug 2016
-
“On-line Active Reward Learning for Policy Optimisation in Spoken Dialogue Systems”
Toshiba Research Cambridge. 8 June 2016
Academia
-
"Pracitcal Approaches to Conversational AI"
Cambridge LTL Group 2018
-
“Reward Estimation for Dialogue Policy Optimisation”
DeepHack.Turing Summer School Hackahton, Moscow, Russia. 25 July 2017 [slides ]
-
“Beyond Siri: towards fully data-driven conversational systems”
Queens' College Cambridge Grad Talk. 20 Feb 2017
-
“On-line Active Reward Learning for Policy Optimisation in Spoken Dialogue Systems”
Cambridge NLIP Seminar. 25 Nov 2016
-
“Practical and Scalable Reinforcement Learning for Spoken Dialogue Systems”
NTU Speech Lab, Taiwan. 24 June 2016
-
“Practical and Scalable Reinforcement Learning for Spoken Dialogue Systems”
Academia Sinica Taiwan. 21 June 2016
-
“Transfer Learning”
Cambridge University Machine Learning Group Seminar. Feb 2015, with Yingzhen Li, [slide ]
Conference
-
“Sample-efficient Actor-Critic Reinforcement Learning with Supervised Data for Dialogue Management”
SIGDIAL 2017. Saarbruecken, Germany.
-
“On-line Active Reward Learning for Policy Optimisation in Spoken Dialogue Systems”
ACL 2016. Berlin, Germany.
-
“Learning from Real Users: Rating Dialogue Success with Neural Networks for Reinforcement Learning in Spoken Dialogue Systems”
Interspeech 2015. Dresden, Germany.
-
“A Cloud-based Personalized Recursive Dialogue Game System for Computer-Assisted Language Learning”
SLaTE 2013. Grenoble, France.
-
“A Recursive Dialogue Game Framework with Optimal Policy Offering Personalized Computer-Assisted Language Learning”
Interspeech 2013. Lyon, France.
-
“A Dialogue Game Framework with Personalized Training using Reinforcement Learning for Computer-Assisted Language Learning”
ICASSP 2013. Vancouver, Canada.