Closed jonadsimon closed 2 years ago
Currently am normalizing overlaps with respect to maximum possible overlaps, giving a "ratio"
Might be better to normalize with respect to expected overlaps, where expectation is computed as a function of baseline letter co-occurrence frequencies.
Dumb idea, not helpful. See https://github.com/jonadsimon/wonder-words-generator/issues/40 for a better idea
Currently am normalizing overlaps with respect to maximum possible overlaps, giving a "ratio"
Might be better to normalize with respect to expected overlaps, where expectation is computed as a function of baseline letter co-occurrence frequencies.