Ask whether the pair is surprising

A journey through pointwise mutual information with Botie.

Natural Language Processing · A journey with Botie

First visit? Learn the workshop controls

Botie's notebook

Are Those Words Associated, or Just Common?

Part 1 of 3

Two words may appear together frequently simply because both appear everywhere. Raw co-occurrence counts cannot tell us whether the pairing is especially informative. We need a baseline for how often they would meet without an association.

Sources for this history

Church and Hanks (1990), Word Association Norms, Mutual Information, and Lexicography

← Previous lesson: Latent Semantic Analysis Read the full lesson and derivations