Research of the Dialog Systems and Machine Learning Working Group
2025
- Sources of understanding and confusion in conversational systems and how to learn from them, Milica Gasic, Fondazione Bruno Kessler, Trento, 2025
2024
- Is a large language model a (new) model for dialogue?, Milica Gasic, Trivago, Düsseldorf, 2024
- Why LLMs need to worry?, Milica Gasic, Bamberg, 2024
- Continually learning Conversational AI, Tohoku, Japan, 2024
- Continually learning Conversational AI, Milica Gasic, DFKI, Saarbrücken, 2024
2023
- Beyond Hype: Technical Insights into ChatGPT and the Future of Conversational AI, Renato Vukovic, HeiCAD Brown Bag Lunch HHU Düsseldorf, 2023, Aligned_LLMs_HeiCAD_presentation
- Explaining the Sensation: An Accessible Introduction to ChatGPT, Michael Heck, HeiCAD Special Event, 2023, ChatGPT2023
- Dialogue Evaluation via Offline Reinforcement Learning and Emotion Prediction, Inivited talk at the 5th Workshop on NLP for Conversational AI, Toronto, Canada, 2023, Evaluation-072023_convai5
- Dialogue Evaluation via Offline Reinforcement Learning and Emotion Prediction, Keynote at the 32. Tagung der Computerlinguistik-Studierenden (TaCoS), Düsseldorf, Germany, 2023, Evaluation-062023_tacos
- Exploring the Shape of Word Spaces with Topological Data Analysis, Benjamin Ruppik, - Pitt NLP Seminar, University of Pittsburgh Computer Science department, Pittsburgh, PA, USA, - MIT CSAIL Spoken Language Systems Group, Cambridge, MA, USA, - Columbia University NLP Seminar, New York, NY, USA, 2023
2022
- Topological Data Analysis in Word Embedding Spaces, Benjamin Ruppik, Invited talk at the Geometry graduate colloquium, ETH Zurich, Switzerland, 2022
2021
- Shifts and other vices of human-computer dialogue, Milica Gasic, NeurIPS 2021
- Neural Dialogue Models: Growing up confident and independent, Milica Gasic, KONVENS 2021
- User-centric dialogue models, Milica Gasic, NSF Future Directions Workshop, Emory NLP Research Lab, 2021
- Towards Dynamic Dialogue Models, Milica Gasic, Annual Conference of the SLTs, University of Sheffield, 2021
- A topological view of polysemy, Milica Gasic, Lectures on Computational Linguistics 2021, University of Roma, 2021
- Towards Dynamic Dialogue Models, Milica Gasic, TU Darmstadt, 2021
2020
- Dialogue as a sequential decision-making process, Milica Gasic, ML in PL Conference, 2020
- Dialogue as a sequential decision-making process, Milica Gasic, HeiCAD AI Conference, 2020
- 10 things you should know about dialogue, Milica Gasic, 7th Frederick Jelinek Memorial Summer Workshop on Speech and Language Technology, JSALT 2020
- On the track of multi-domain dialogue models, NeurIPS Human in the loop workshop, virtual, 2020
2019
- General Dialogue Topics, Koichiro Yoshino, HHU, 2019
- User Simulation for Dialogue Systems, Hsien-Chin Lin, HHU, 2019
- Policy, Nurul Lubis, HHU, 2019
- Breaking open Belief Tracking, Carel van Niekerk, HHU, 2019
- The World of Word Vectors, Michael Heck, HHU, 2019
- Fast and curious dialogue policy optimisation, Milica Gasic, WCIS, Paris 2019
- Deep Learning for natural dialogue modelling, Milica Gasic, University of Potsdam, 2019
- Dialogue Systems – Can computers have emotions? Futurium, Berlin, Germany, 2019
- Curiosity in dialogue policy optimisation, Google Conversational Search and Recommendation Workshop, London, UK, 2019
- ConvLab3 public repo, https://github.com/ConvLab/ConvLab-3
- TripPy-R: Codebase for "Robust Dialogue State Tracking with Weak Supervision and Sparse Data", https://gitlab.cs.uni-duesseldorf.de/general/dsml/trippy-r-public, published in TACL 2022
- PyDial 3.0
- FeudalGain: Codebase for "What Does The User Want? Information Gain for Hierarchical Dialogue Policy Optimisation", published in IEEE ASRU 2021.
- SetSUMBT: Codebase for “Uncertainty Measures in Neural Belief Tracking and the Effects on Dialogue Policy Performance”, published in EMNLP 2021.
- TUS: Codebase for "Domain-independent User Simulation with Transformers for Task-oriented Dialogue Systems", published in SIGDIAL 2021.
- Calibrating Dialogue Belief State Distributions: Codebase for "Knowing What You Know: Calibrating Dialogue Belief State Distributions via Ensembles", published in Findings of EMNLP 2020.
- TripPy: Codebase for "TripPy: A Triple Copy Strategy for Value Independent Neural Dialog State Tracking" is a new approach to dialogue state tracking (DST) which makes use of various copy mechanisms to fill slots with values, published in SIGDIAL 2020.
- LAVA: Codebase for "LAVA: Latent Action Spaces via Variational Auto-encoding for Dialogue Policy Optimization", published as a long paper in COLING 2020.
- Prof Marcus Zibrowius, Heinrich Heine University Düsseldorf
2024
- Slow and steady wins the race: Reinforcement learning in Conversational AI, Pint of Science, Düsseldorf, 2024
2023
- Podcast Interview, Renato Vukovic, Lessing Gymnasium, 2023, https://open.spotify.com/episode/0PxIFKCN3uksB6qTmMfI0T
- Michael Heck, in conversation with Christopher Trinks, "Wir müssen die Gefahren von KI sehen", Rheinische Post, vol. 78, no. 133 D/D-S, June 2023,
https://rp-online.de/nrw/staedte/duesseldorf/forschung-in-duesseldorf-chatgpt-und-co-welche-gefahren-ki-birgt_aid-91376447