A Mathematical Theory of Communication
Shannon defined information quantitatively and introduced entropy as a measure of a source's uncertainty.
Why it matters
Prediction acquired a unit it could be measured in, the basis of all later language modelling.
The paper solves an engineering problem, sending messages through a noisy channel. What matters here is a side result: modelling a source as a sequence of symbols with probabilities, and approximating text with n-grams. It is the first quantitative account of how predictable the next symbol is.