Masked Word Prediction with Statistical and Neural Language Models

Language modeling is one of the main tools used in most of the natural language processing problems. Neural language models became popular with the approach of transfer learning (using a model trained for one task to another task). However, studies with neural language models on Turkish have not yet reached the desired levels. In this study, statistical and neural language models were compared using the prediction of a word that was closed randomly in a sentence. Neural language models were found to have higher success.