Next Word Prediction In Telugu Sentences Using Recurrent Neural Networks

Abstract

Data that is collected from the social network is almost full of noise and extra characters. Popular social networks like Twitter, Facebook, WhatsApp, YouTube generate lots of data and preprocessing is the primary task to transform the messy data into useful or reliable data. The preprocessing step removes, replaces, modifies, and normalizes the collected data. It also fills the data with suitable words in incomplete sentences. In this paper, we address the procedure to predict suitable words in the incomplete sentence in the Telugu language.   

            Natural Language Toolkit, for example, Indic Language (iNLTK) supports the Indian language for processing 13 official languages in India. Here, Telugu is one of the languages that is supported by iNTK. The prediction of the next word can be implemented by using Recurrent Neural Networks, by predicting the next word in a sequence so that the number of keystrokes of the user can be reduced to save time. This approach can be used effectively for various NLP preprocessing methods, sentence auto-completion in Telugu.