Sentiment Analysis of Turkish Reviews on Google Play Store

The development of technology and the increase in shared data volume has made the analysis performed on this data valuable. Users expressing their opinions as comments have also been invaluable to companies. In this study, approximately two million Turkish user reviews that are made on Google Play are collected and evaluated. The comments were analyzed as ratings out of five and positive-negative states. For this data, encoder-based BERT model, which has been popular recently in deep learning methodologies, has been fine-tuned and used. The effects of different sizes of training data on success were observed. Classification achievements are presented with tables and graphics. In order to compare the performance of the Turkish BERT model against prior text classification studies in the literature, a similar smaller amount of training data is used