
“Lifelong learning describes the vision of developing intelligent systems that are able to learn new tasks or skills over their lifetime. This thesis aims to contribute to this vision by exploring how the principles of lifelong learning can be applied to dialogue systems.”
Title: Evaluation of Lifelong Machine Learning for Dialogue Systems
Supervisors:
Prof. Dr. Martin Volk (UZH)
Prof. Dr. Mark Cieliebak (ZHAW)
Keywords: machine learning, deep learning, lifelong machine learning, dialogue systems, question answering, natural language interfaces to databases, natural language processing
Abstract: This Thesis is concerned with the application of lifelong learning (LL) onto dialogue systems. Lifelong learning describes the vision of developing intelligent systems that are able to learn new tasks or skills over their lifetime. The more practical goal is to avoid the need of having a machine-learning expert retrain the system manually. Rather, the system notices its deficiencies and learns the new task autonomously. Currently, we are still far away from this grand vision. This thesis aims to contribute to this vision by exploring how the principles of lifelong learning can be applied to dialogue systems. An important part of lifelong learning are the metrics used by the system to assess its current quality in order to trigger the necessary steps required to improve itself.
Planned Duration: 1.1.2018 – 1.6.2021
Publications:
https://www.researchgate.net/profile/Jan_Deriu https://scholar.google.ch/citations?user=PvHXh9wAAAAJ&hl=de