Florian Kreyssig

Florian Kreyssig 

Florian Kreyssig DE
Ph.D Candidate (CV, April'22)
Machine Intelligence Laboratory
Cambridge University Engineering Department
University of Cambridge

Research Interests: Active Learning, Semi-Supervised Learning, ASR, Diarisation


  1. Biased Self-Supervised Learning for ASR
    GSF.L. Kreyssig, Y. Shi, J. Guo, L. Sari, A. Mohamed, GSP.C. Woodland
    Accepted to Interspeech 2023

  2. Transferable Dialogue Systems and User Simulators
    GSB. Tseng, Yinpei Dai, GSF.L. Kreyssig, GSBill Byrne
    ACL, Bangkok , 2021

  3. A Distributed Optimisation Framework Combining Natural Gradient with Hessian-Free for Discriminative Sequence Training
    A. Haider, C. Zhang, GSF.L. Kreyssig, GSP.C. Woodland
    Neural Networks, 2021

  4. Discriminative Neural Clustering for Speaker Diarisation
    Q. Li*, GSF.L. Kreyssig*, C. Zhang, GSP.C. Woodland
    SLT, Shenzhen, 2021, [Best Student Paper Award]
    Link to Code and Data
    * Indicates first authors. Order determined by coin flip.

  5. Cosine-Distance Virtual Adversarial Training for Semi-Supervised Speaker-Discriminative Acoustic Embeddings
    GSF.L. Kreyssig, GSP.C. Woodland
    Interspeech, Shanghai, 2020

  6. PyHTK: Python Library and ASR Pipelines for HTK
    C. Zhang, GSF.L. Kreyssig, Q. Li, GSP.C. Woodland
    ICASSP, Brighton, 2019

  7. Deep Learning for User Simulation in a Dialogue System
    GSF.L. Kreyssig
    M.Eng. Thesis, University of Cambridge, 2018, [Best Presentation Award]
    Supervisor: GSProf. Dr. M. Gašić

  8. Neural User Simulation for Corpus-based Policy Optimisation of Spoken Dialogue Systems
    GSF.L. Kreyssig, GSI. Casanueva, GSP. Budzianowski and GSM. Gašić
    SIGDIAL, Melbourne, 2018

  9. Variational Cross-domain Natural Language Generation for Spoken Dialogue Systems
    GSB. Tseng, GSF.L. Kreyssig, GSP. Budzianowski, GSI. Casanueva, GSY. Wu, GSS. Ultes and GSM. Gašić
    SIGDIAL, Melbourne, 2018

  10. Feudal Dialogue Management with Jointly Learned Feature Extractors
    GSI. Casanueva, GSP. Budzianowski, GSF.L. Kreyssig, GSS. Ultes, GSB. Tseng, GSY. Wu and GSM. Gašić
    SIGDIAL, Melbourne, 2018

  11. Improved TDNNs using Deep Kernels and Frequency Dependent Grid-RNNs
    GSF.L. Kreyssig, C. Zhang, GSP.C. Woodland
    ICASSP, Calgary, 2018

  12. Modular Construction of Complex Deep Learning Architectures in HTK
    GSF.L. Kreyssig, C. Zhang, GSP.C. Woodland
    UK Speech, Cambridge, 2017

  13. Experimental Studies on Teacher-student Training of Deep Neural Network Acoustic Models
    Q. Li, C. Zhang, GSF.L. Kreyssig and GSP.C. Woodland
    UK Speech, Cambridge, 2017


Undergraduate Supervisions (100+ hours)

Professional Activity

Reviewer (30+ papers)