Text Completion, Classification and Correction Using Different Types of Neural Networks

Abstract

This article reviews the various methods used in the process of completing and correcting text on a word based as well as a sentence based approach. Firstly, we look at the process of understanding the text from scratch which requires the application of deep learning to text that facilitates the understanding the inputs at a character level and other concepts related to abstract texts, and all this helps us in classifying texts. For the above said purposes we use different types of Artificial Neural Networks which include Convolutional Neural Networks and Recurrent Neural Networks. Both Convolutional and Recurrent Convolutional neural networks can be individually used in the classification of texts and Recurrent Neural Networks along with Convolutional Networks are used in the completion and correction. The result of using temporal and character level Convolutional network for classification are also shown to be competitive to the results produced by traditional techniques. The model for correction and completion enables both the processes to occur simultaneously by ciphering the hidden representations at a character level and decoding the revised sequence. 

Keywords: Machine Learning, Artificial Neural Network, Convolutional Neural Network, Recurrent Neural Network, Computer Vision, Supervised Learning