This is an interesting paper, but it leaves a lot to imagination. The authors really need to explain their terms when they use them. For instance, in Figure 1, what are the example sets S and E? I presumed they stood for "start" and "end", but then the learner appears to learn prefixes and suffixes at the same time. I agree with sameer, the weak learner should be better described. Perhaps this paper requires a more complete knowledge of wrapper induction.

Additionally I would like to see results comparing BWI to standard WI. This is sort of shown in Figure 4, if we take early boosting rounds to indicate real WI performance. However, I doubt this is the case, because the performance is abysmal -- it doesn't indicate that WI is a useful method (for zero boosting rounds, that is).

- greg