Recall Systems: Effcient Learning and Use of Category Indices

We introduce the framework of recall systems for efcient learning and retrieval of categories when the number of categories is large. A recall- system here is a simple feature-based interme- diate ltering step which reduces the potential categories for an instance to a small manage- able set. The correct categories from this set can then be determined using traditional classiers. We present a formalization of the index learning problem and establish NP-hardness and approxi- mation hardness. We proceed to give an efcient heuristic for learning indices, and evaluate it on several large data sets. In our experiments, the index is learned within minutes, and reduces the number of categories by several orders of magni- tude, without affecting the quality of classica- tion overall.