Description
In this talk, I will discuss how to address some of the most likely causes of failure for new Natural Language Processing (NLP) projects. My main recommendation is to take an iterative approach: don't assume you know what your pipeline should look like, let alone your annotation schemes or model architectures.