A novel boot strapping algorithm for text extraction in a self-organising neural network model

Abstract

With rapid growth in internet and its associated communication protocols, need for printed documents to be carried over from one place to another has been reduced to minimise the cost and time. Research contributions in the past have paved the way for implementation of smart and intelligent algorithms to further minimise manual intervention in processing of documents. One such area is the automation of text extraction from documents with increased accuracy and least number of false detections. A wide range of algorithms and methodologies have been reported in the past towards efficient extraction of text from documents which may be online or offline. This research paper proposes an efficient extraction algorithm of text from given set of documents which may or may not be graphic through utilisation of a hybrid SOM-ANN algorithm. The experimentation has been done over a wide variety of inputs and convergence of error in extraction is found to be minimum when compared to other conventional extractors.