A Deep Learning Approach for Government Automated Bengali Note Writing Service

Abstract

The world is undergoing various changes despite the revolution of modern technologies. With the help of automation tools, human tasks are been made very simple and automated. People are adopting Artificial intelligence in healthcare, retail and Ecommerce, food and tech, travel, banking and financial services, entertainment, and various sectors in the developed countries. But in the underdeveloped or developing countries still, now there are some issues to solve. In this world, there are approximately 228 million people who communicate in the Bengali language. And in Bangladesh Bengali is also used as an official language. Govt. or non-govt. offices use Bengali to perform their official tasks. They have to write down their official tasks, letters, or maybe important notes on their own. As there is no automation tool available regarding Bengali note writing, the office staffs have to spend their valuable time by writing down the large notes or documents. This particular thing not only hampers the time but also is a bad usage of valuable human resources. To bridge this gap we have proposed a model that will be able to generate large notes by typing hints of a couple of words. We have used Bag of words and three-layer Artificial neural network to resolve this issue. The Bengali data was processed in a special way so that our model can perform well with about 1431 text documents. After running a couple of experiments our model was able to generate large notes from short notes with 91% accuracy