As more and more information-seeking activities are moving to visual interfaces, the way of interacting with QA systems is changing. For a QA system, this poses two key challenges:
- understanding when and if the question is related to the context
- ensuring that the answers are contextually-relevant when required.
This talk will show how traditional QA approaches fail in such scenarios, and present solutions that serve information-seeking in the presence of rich contextual information. This includes the efficient modeling of conversations, the design of a realistic dataset with contextual information, and experiments showing that, by incorporating context from different sources, strong baselines can outperform SOTA models from traditional QA.
Speaker(s): Gianni Barlacchi
Gianni Barlacchi is a machine learning scientist at Amazon working on creating new solutions for Alexa and doing research on question answering. He received a Ph.D. in Computer Science at the University of Trento and a MSc in Computer Engineering at the University of Siena. His research interests span from conversational question answering to continual learning for NLP and how to enable systems to learn new knowledge from a stream of data across time.