Information extraction is a key challenge in developing intelligent systems. One tool that can be useful for this task is semantic parsing. Semantic parsers extract the logical representation of a sentence. These parsers are typically trained using an annotated corpus that maps sentences to their logical representation. However, alternative methods of supervision are possible. This project aims to train a parser using supervision from video scenes that are about the sentence.

I like projects that combine computer science with other disciplines. More recently, I’ve had the opportunity to work on projects in natural language processing. Through my SuperUROP I hope to gain more experience in this field and contribute to an interesting project.