Selected topics in Machine Learning and Natural Language Processing

Human computer dialogue modelling can be viewed as a machine learning problem where we try to build a statistical model of dialogue and train it using the data consisting of human-human or human-machine dialogues. The area of statistical dialogue modelling lies in the intersection of machine learning and natural language processing. This area however concerns problems that are also subject of many other areas. These problems include interaction, adaptation, zero-shot learning, representation learning, deep learning algorithms, etc. This seminar covers a number of topics related to dialogue modelling that have impact beyond dialogue modelling.

• 24.04.20:         Towards Ontology-Independent Dialogue State Tracking, Dr M. Heck (RE)
• 08.05.20:         Variational Auto-Encoders, Dr N. Lubis (TUT)
• 15.05.20:         Selected papers from ICLR conference, Prof Dr M. Gasic (RE)
• 22.05.20:         Transformers, M. Moresi (TUT)
• 29.05.20:         Generative Adversarial Networks, H.-C. Lin (TUT)
• 05.06.20:         Reward Estimation in Reinforcement Learning, C. Greishauser (RE)
• 12.06.20:         Uncertainty in Dialogue Belief Tracking, C. van Niekerk (RE)
• 19.06.20:         Emotion In Human-Computer Interaction, Dr N. Lubis (RV)
• 26.06.20:         Semantic Similarity, M. Moresi (TUT)
• 03.07.20:         Correlations between sets of words, C. van Niekerk (RE)

TUT = tutorial; RE = research talk; RV = review talk