Using a boosted tree classifier for text segmentation in hand-annotated documents

► We propose a tree-structured classifier for text identification. ► Increasing classification accuracy for imbalanced data problem. ► Avoid over-fitting for imbalanced set.