What is "statistical semantics"

n. (context computational linguistics English) The study of estimation of the meanings of words by looking at patterns of words in huge collections of texts, using statistical methods.

Statistical semantics is the study of "how the statistical patterns of human word usage can be used to figure out what people mean, at least to a level sufficient for information access" . How can we figure out what words mean, simply by looking at patterns of words in huge collections of text? What are the limits to this approach to understanding words?