IRS13 An Experiment in Automatic Thesaurus Construction chapter R. T. Dattola D. M. Murray Harvard University Gerard Salton Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. VIII-13 Sub- Correlation Cutoff Overlap Class Class Collection Function Ratio Ratio Coefficient 2 Cosine .70 .915 .263 24.0 .71 .034 .280 .95 .76 .034 .300 1.01 .82 0 .311 0 Tanimoto .50 1.03 .263 27.2 .55 .034 .280 .948 .60 .025 .280 .711 .65 .034 .297 1.01 .70 0 .331 0 4 Cosine .70 .484 .325 15.8 .71 .095 .365 3.48 .76 .095 .381 3.62 .82 .016 .444 .705 Tanimoto .50 .556 .317 17.6 .55 .095 .365 3.48 .60 .095 .365 3.48 .65 .078 .381 3.02 .70 .016 .444 .705 6 Cosine .70 .593 .341 20.2 .71 .007 .400 .296 .76 .007 .400 .296 .82 0 .407 0 Tanimoto .50 .593 .326 19.3 .55 .007 .400 .296 .60 .007 .400 .296 .65 .007 .400 .296 .70 0 .407 0 Evaluation Statistics 3 ADI Subcollections Table 2