What is it about?
While bi-directional long short-term (BLSTM) neural network have been demonstrated to perform very well for English or Arabic, the huge number of different output classes (characters) encountered in many Asian fonts, poses a severe challenge. In this work we investigate different encoding schemes of Bangla compound characters and compare the recognition accuracies. We propose to model complex characters not as unique symbols, which are represented by individual nodes in the output layer. Instead, we classify only basic strokes and use special nodes which react to semantic changes in the writing.
Featured Image
Why is it important?
This is a new research direction towards compound character encoding.
Perspectives
We have showed that our approach outperforms the common approaches to BLSTM neural network-based handwriting recognition considerably for Bangla. In future, we will consider other scripts.
Dr. Nilanjana Bhattacharya
Read the Original
This page is a summary of: Improved BLSTM Neural Networks for Recognition of On-Line Bangla Complex Words, January 2014, Springer Science + Business Media,
DOI: 10.1007/978-3-662-44415-3_41.
You can read the full text:
Contributors
The following have contributed to this page
