JSNU Logo
CECPC_Core / Collocation Strength Word cooccurrence and association measures

CECPC_Core / Collocation Strength

Count words around a node and calculate MI, t-score, z-score, chi-square, and log-likelihood.

Analysis Settings

Metric Notes

MI may rank rare pairs highly. Keep minimum cooccurrence at 5 or above, and prefer log-likelihood for sparse data.

Fixed definitions: MI=log₂(O/E), t=(O-E)/√O, z=(O-E)/√E; χ² and G² use the same 2×2 contingency table.