Using focus of attention in classification

The goal of a classifier is to accurately predict the value of a class given its context. Often the number of classes "competing" for each prediction is large. Therefore, it is necessary to "focus attention" on a smaller subset of these. We investigate the contribution of a "focus of attention" mechanism using enablers to the performance of a word predictor. We then describe a large scale experimental study in which the approach presented is shown to yield significant improvements in word prediction tasks.