“Cold” answers “hot.” Most words don’t get off that easy.
A thesaurus will hand you one antonym per word and move on. But every opposite answers a particular question. The opposite of mother along parental gender is father. Along generational position, it’s child. Along nurturing behavior, it’s neglect — the mother who isn’t one. Three different questions, three different opposites, all correct.
Even “hot,” the textbook case, opposes four ways: cold in temperature, mild in spiciness, unpopular in trendiness, ugly in attractiveness. The mapping covers 1.5 million words and phrases of the raw vocabulary — every opposite tagged with its axis and rated 1–7 for strength. The deep end of that list is noisy, and it’s there for mining: emergent patterns among the over-represented terms, and better odds that an interesting topic surfaces from the long tail. Familiar words like the ones below carry the clean story.
Darker means stronger opposition (rated 1–7 in the data; showing 3+). Notice how often the “obvious” opposite is only the first axis of several.
Single words are just the start. The data also pairs opposing topics — each side a pool of words, the pair scored for whether the opposition would be self-evident to a puzzle player. There are 162,837 scored pairs; here are a few dozen of the cleanest, one per axis. Click around.
Showing 48 of 162,837 scored pairs — one per axis.
For a puzzle, an opposing pair is free structure. A board built on Wired vs Tired reads instantly — players sort mania from lethargy without being told the rule. And because each side ships as a pool of member words, one pair yields many boards.
The antonyms & contrasts table in the release carries both layers: word-level opposites tagged with axis and strength, and the scored topic pairs with their member pools. See real rows from every table, or just take the whole thing — it’s public domain.