Дискриминационно обученная модель тегирования

We present a technique for augmenting annotated training data with hierarchical word clusters that are automatically derived from a large unannotated corpus. Cluster membership is encoded in features that are …...