My research goal is to study the content dimension of natural language and the social dimension of it (e.g., who says it, in what context, for what goals)

Keywords: faculty , research

One major focus in our lab is to develop better machine learning models for low resourced NLP tasks. Deep learning has achieved extremely good performance in most supervised learning settings. However, when there is only limited labeled data, these models often fail. This strong dependence on labeled data largely prevents neural network models from being applied to new settings or real-world situations.

It’s easier to build something up from scratch than it is to come to an existing working complex system and try to figure it out

When you build something from scratch you get to build up your mental model bit by bit, feeling in control and understanding the whole thing. It might be(come) very complex, but it makes sense and it’s ‘correct’ in your perspective. If you are dragged into a complex working system, you have to build up the mental model based on what you see. Imagine you get to a legacy system and you’re supposed to figure it out. How else would you be able to work on it? That means you have to uncover all the mental models that built up to this system. Next to that, you need to connect them and figure out why certain decisions were made.