There are two different approaches to managing the preference set, ordered management and weighted management. Ordered management is based on discarding those antecedents that do not fulfill a preference if there is any candidate that fulfills it. Weighted management is based on assigning a weight to each preference and then selecting the candidate with the maximum value.
For these experiments, the system was trained so as to obtain the best set of preferences. Subsequently, we applied both of the approaches to preference management to obtain the best result with the training corpus. Once we obtained the best set of preferences and its best management, we evaluated the system with the test corpus in order to obtain independent results.
So, for experiments 0, 1, and 2, ordered management was applied to obtain the best set of preferences. Then, in experiment 3, we applied weighted management to improve the results in the training corpus.