Data wydarzenia:

ConvLab-3: A Flexible Dialogue System Toolkit Based on a Unified Data Format

Data: wtorek, 16.01.2024, godz. 11:00-12:00

Prelegent: Chris Geishauser (Heinrich Heine University Düsseldorf)

Streszczenie: Task-oriented dialogue (TOD) systems function as digital assistants, guiding users through various tasks such as booking flights or finding restaurants. The development of TOD systems requires multiple components and learning paradigms such as supervised and reinforcement learning. Existing toolkits for building TOD systems often fall short of delivering comprehensive arrays of data, models, and experimental environments with a user-friendly experience. Our multifaceted dialogue system toolkit ConvLab-3 is crafted to bridge this gap. The unified data format of ConvLab-3 simplifies the integration of diverse datasets and models, significantly reducing complexity and cost for studying generalization and transfer. Our enhanced reinforcement learning toolkit features a streamlined training process, in-depth evaluation tools and a selection of user simulators, supporting the rapid development and evaluation of robust dialogue policies. We regard ConvLab-3 as a powerful tool for seasoned researchers and an accessible platform for newcomers. In this talk I will introduce task-oriented dialogue systems and their components as well as the most important key features of ConvLab-3.

Miejsce: B1-7/8 oraz online