A RNN based Approach for next word prediction in Assamese Phonetic Transcription
Partha Pratim Barman, Abhijit Boruah · Procedia Computer Science · 2018
In this paper, we present a Long Short Term Memory network (LSTM) model which is a special kind of Recurrent Neural Net-work(RNN) for instant messaging, where the goal is to predict next word(s) given a set of current words to the user. This method is more complex in other languages apart from English. For instance, in Assamese language, there are some equivalent synonyms of the word ‘you’, that is used to address a second person in English. Here, we have developed a solution to this issue by storing the transcripted Assamese language according to International Phonetic Association(IPA) chart and have fed the data into our model. Our model goes through the data set of the transcripted Assamese words and predicts the next word using LSTM with an accuracy of 88.20% for Assamese text and 72.10% for phonetically transcripted Assamese language. This model can be used in predicting next word of Assamese language, especially at the time of phonetic typing.