Enhancing Turkish Sentiment Analysis Using Pre-Trained Language Models

Sentiment Analysis is a particular natural language processing task that is widely used in obtaining the customer's opinion for a product and measuring customer satisfaction. As with other natural language processing tasks, there are language-specific difficulties in sentiment analysis in Turkish. In this paper, we evaluate and compare various techniques used in Turkish sentiment analysis. We have used pre-trained language models to improve the classification performance of the Turkish sentiment analysis. We compared the results we obtained with previous studies using a data set that was previously used in studies that conducted emotional analysis studies with different techniques. The results of our study showed that we achieved the best classification performance in Turkish sentiment analysis by using pre-educated language models compared to other studies conducted on the data set we used.