Set up

To experiment with the clustering evaluation tools, we will use the various contain.rsf files we created in the Clustering lab.

We will also utilize an authoritative decomposition for the TAB2PS system. You can get it here.

Koschke - Eisenbarth (KE)

Transform the contain.rsf files you would like to compare to the KE format with

unitrans input.rsf output.kos

For example, to evaluate the output of ACDC for the TAB2PS system, give

unitrans tab2ps.contain.acdc.rsf tab2ps.contain.acdc.kos
unitrans tab2ps.contain.auth.rsf tab2ps.contain.auth.kos

To run KE, give

ke -cand tab2ps.contain.acdc.kos -ref tab2ps.contain.auth.kos

Observe the recall rate towards the end of the output, as well as the rest of the provided information. To see even more details, run the same command with the -detailed flag on, as in:

ke -cand tab2ps.contain.acdc.kos -ref tab2ps.contain.auth.kos -detailed

MoJo

To run MoJo for the same two decompositions, give

mojo tab2ps.contain.acdc.rsf tab2ps.contain.auth.rsf

To get the corresponding MoJoFM value, give

mojo tab2ps.contain.acdc.rsf tab2ps.contain.auth.rsf -fm