Building a text extractor to extract particular type of text from unstructured text data

There is a spectrum of methods. On one end is the tried n' true method of performing pattern matching with rules. Write a series of regular expressions (i.e., regex). On the other end is the newer methods of weak supervision, a system that learns over time to get better. Snorkel one of implementation of weak supervision. Here is a Snorkel demo for identifying mentions of a chemical causing a disease from academic literature.