Text extraction using artificial neural networks
Mudassar Raza, Ata Ullah, Khawaja Moyeezullah Ghori, Sajjad Haider · Networked Computing and Advanced Information Management · 2011
Computerized text extraction from a number of static resources speeds up the process in offices, libraries, banks and a variety of other places. Text extraction can be done using a number of different techniques depending upon the need of system and accuracy level. Artificial Neural Networks have a well reputed history in this regard that they have a wonderful accuracy level for automated text extraction. This research paper shows the different modifications that can be made to existing text extraction techniques using backpropagation artificial neural networks. Classification of input patterns to categories on the basis of character width considerably increases the accuracy of results. Hidden layer optimization techniques can also contribute to the accuracy of recognizing patterns. Moreover, learning rates were got by trial and error methodology, which may help in enhancing the overall accuracy of such systems.