Emotion analysis with data of turkish product comments on E-commerce sites

In this study, the aim is to conduct sentiment analysis from user comments on e-commerce sites. With the widespread use of the internet and the popularity of social media platforms today, users have more opportunities to share and express their thoughts about products, services, and various topics. However, analyzing this big data source is not feasible with human effort alone, hence the need for methods like sentiment analysis. Sentiment analysis is a method used to understand and infer human emotions from data sources such as text, voice, or images. This thesis aims to classify user comments as positive, negative, or neutral, aiming for more detailed results with target-based sentiment analysis. The findings indicate that the proposed BiLSTM (Bidirectional Long Short-Term Memory) based model outperforms other models. This model achieved high accuracy rates in classifying negative, positive, and neutral sentiment states. Particularly, the addition of the sentiment information layer improved the representation of emotional features and led to more accurate classification. Comparative analysis results demonstrate the strong potential of deep learning models in text-based sentiment analysis, especially highlighting the effectiveness of bidirectional approaches. This study emphasizes the future potential of deep learning-based sentiment analysis applications. Future work should focus on working with larger datasets and developing the model to handle more complex emotional content.