CS 224U - Natural Language Understanding
CS 224U is about extracting meaning from language in a form that a computer program can act on. For our project, Yushi Wang and I investigated using the output of a syntactic dependency parser (the Stanford parser) to provide useful features for SEMPRE, a flexible semantic parsing framework.
We achieved substantial gains over a strong baseline: 38.3% F1 on the WebQuestions dataset, compared to 31% for a baseline consisting of a diverse subset of the the features used in the paper that introduced the dataset (Berant et al., 2013).