Survey On Next Word Prediction Techniques In Natural Languages

Next word prediction is the core problem of NLP, which has numerous applications, to predict which word will be said next. Researchers used several word-level statistical and deep learning models to generate the text. In the past, before the advancement of deep learning techniques, statistical techniques were used to predict the next word. These methods have the drawback of generating text without ensuring the syntactic accuracy of the sentences. Neural networks have also become more popular than traditional methods due to their accuracy in word prediction. Nowadays, word prediction is often done using LSTM, Bi-LSTM, and GRU. This paper provides a survey of different techniques with their classification for predicting the next word and evaluates them. Also, this work identified research gaps and provide some future directions based on the findings.