About


I'm a third year PhD student at the University of Cambridge, supervised by Professor Steve Young at the Dialogue Systems Group. I was previously a Machine Learning Consultant at VocalIQ, where I collaborated with Diarmuid O Seaghdha and Blaise Thomson


My research is currently focused on belief tracking in dialogue, where I am interested in moving towards building open-domain dialogue state tracking models. I am also interested in deep learning, natural language processing, Bayesian nonparametrics, unsupervised and semi-supervised learning.  

I completed my Master's thesis at the Cambridge Machine Learning Group, where I worked on Kernel Structure Discovery for Gaussian Process Classification with Professor Zoubin Ghahramani. Prior to that, I graduated from the Cambridge Computer Science Tripos, where my final year project with Dr Sean Holden focused on semi-supervised learning.



Publications 
  1. Nikola Mrkšić, Diarmuid Ó Séaghdha, Blaise Thomson, Milica Gašić, Pei-Hao Su, David Vandyke, Tsung-Hsien Wen and Steve Young. 2015. Multi-domain Dialog State Tracking using Recurrent Neural NetworksIn Proceedings of ACL 2015. Beijing, China. [poster] [bib]
  2. Nikola Mrkšić, Diarmuid Ó Séaghdha, Blaise Thomson, Milica Gašić, Lina Rojas-Barahona, Pei-Hao Su, David Vandyke, Tsung-Hsien Wen and Steve Young. 2016. Counter-fitting Word Vectors to Linguistic Constraints. In Proceedings of NAACL 2016. San Diego, US. [talk] [GitHub] [bib] 
  3. Nikola Mrkšić, Diarmuid Ó Séaghdha, Tsung-Hsien Wen, Blaise Thomson and Steve Young. 2017. The Neural Belief Tracker: Data-Driven Dialogue State Tracking. In Proceedings of ACL, Vancouver, Canada. [woz 2.0 dataset] [corrected dstc2 transcriptions]
  4. Ivan Vulić, Nikola Mrkšić, Roi Reichart, Diarmuid Ó Séaghdha, Steve Young, Anna Korhonen. 2017. Morph-Fitting: Fine-Tuning Word Vector Spaces using Simple Language-Specific Rules. In Proceedings of ACL, Vancouver, Canada. 
For a full list of publications, please see my Google Scholar profile.


Dissertations
  1. Nikola Mrkšić. 2014. Kernel Structure Discovery for Gaussian Process Classification. Master's Thesis, Computer Lab, University of Cambridge.
  2. Nikola Mrkšić. 2013. Semi-supervised Learning Methods for Data Augmentation. Bachelor's Thesis, Computer Lab, University of Cambridge.

Talks
  1. The Probably Approximately Correct Framework. Cambridge Machine Learning Group, November 2014.
  2. Kernel Structure Discovery for Gaussian Process Classification. Cambridge Computer Lab, June 2014.
  3. Semi-supervised Learning for Data Augmentation. Max Planck Institute for Intelligent Systems, Tübingen, February 2014. 
  4. Counter-fitting Word Vectors to Linguistic Constraints. NAACL, San Diego, June 2016.
  5. Neural Belief Tracker. Toshiba Research, Cambridge, December 2016; Apple (Siri) Cambridge, February 2017; Technion, Israel Institute of Technology, March 2017; General Motors Advanced Research Center, Israel, March 2017. 
  6. Vector Space Specialisation, EACL Tutorial (with Ivan and Taher). Valencia, April 2017. 

Teaching
  1. [October 2014 - June 2016] Supervisor for Trinity College, Cambridge. Supervised 15 first and second year undergraduates for the Computer Science Tripos. Courses supervised:  Algorithms, Artificial Intelligence 1, Mathematical and Numerical Methods, Logic and Proof, Concurrent & Distributed Systems.
  2. [October 2010 - now] Petnica Science Centre, Serbia. Associate of the Physics Seminar. Gave several lecture series to high school students. Topics lectured on include Quantum Computing, Artificial Intelligence and Advanced Mathematical Methods.     

Honours and Awards
  1. Trinity College Internal Graduate Scholarship (full funding for the duration of my PhD studies). October 2014.
  2. Research Scholar (Trinity College). October 2014.
  3. Senior Scholar (Trinity College). October 2013.
  4. Trinity College Overseas Bursary (full funding for my undergraduate and Master's studies). October 2010 - June 2014.
  5. Silver Medal, Balkan Olympiad in Informatics. Shumen, Bulgaria, 2009.

Contact

Department of Engineering
University of Cambridge
Trumpington Street
Cambridge CB2 1PZ

Office BN3-011
Email nm480 at cam.ac.uk