To test whether the results from the new system perform better than an ordinary CBR system, the following test was done:
An expert on the OREDA database was given a list of the queries done, where the information on which tool had produced which result was removed. The order of the presentation of the results was picked at random. The expert was told to choose which of the two retrieved cases was most similar to the new case. The results from the expert is given in Fig 5.5. We gave one point to a system if it was thought to deliver a better result than the other one. The figure presents only the last five series, because the expert thought that the first four series was difficult to evaluate. In the tables presenting the results, the system that is thought to be best is marked with a star (*).
