Text Separation from Mixed Documents Using a Tree-Structured Classifier

In this paper, we propose a tree-structured multiclass classifier to identify annotations and overlapping text from machine printed documents. Each node of the tree-structured classifier is a binary weak learner. Unlike normal decision tree(DT) which only considers a subset of training data at each node and is susceptible to over-fitting, we boost the tree using all training data at each node with different weights. The evaluation of the proposed method is presented on a set of machineprinted documents which have been annotated by multiple writers in an office/collaborative environment.

• Type: Journal
• Year: 2010
• Where: ICPR
• Authors: Xujun Peng, Srirangaraj Setlur, Venu Govindaraju, Ramachandrula Sitaram
• Keywords: Binary Weak Learner, Computer Vision, ICPR 2010, Training Data, Tree-structured Multiclass Classifier