First page Back Continue Last page Overview Graphics
Avoid Throwing Away Information
Because everything counts, CRM114 can use a very gentle conditional probability formula, so statistical outlier features have low impact.
CRM114’s per-feature conditional probabilities are limited to roughly the range:
[ .47 ... .53 ] for hapaxes
[ .44 ... .56 ] for 10 occurrences
[ .43 ... .57 ] for 1000 occurrences