Sentiment analysis with natural language processing and deep learning techniques: A study on Turkish texts

E-commerce platforms are widely preferred by many individuals today due to their ability to provide a fast, easy, and contactless shopping experience, aiding consumers in determining their product preferences. During online shopping experiences, the positive and negative reviews about products available on the platform play a crucial role in enabling customers to form realistic opinions and for businesses to enhance the quality of their products/services. However, conducting analysis of these reviews individually can be a complex and time-intensive process, potentially introducing human error risks. Therefore, various deep learning models are being explored for sentiment analysis to enable businesses and customers to conduct rapid and accurate analyses. This study focuses on sentiment analysis of reviews posted on e-commerce sites for specific product categories using deep learning algorithms such as Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), Gated Recurrent Unit (GRU), and Bidirectional Encoder Representations from Transformers (BERT). Single algorithmic and hybrid models are developed to perform sentiment analysis. The results obtained from analyzing e-commerce platform reviews using deep learning algorithms are expected to enhance customer satisfaction on e-commerce platforms and contribute to areas such as marketing strategies.