Optimizing Turkish Opinion Mining: A Comparative Study of AI Algorithms

Opinion mining, aka sentiment analysis, is a branch of Natural Language Processing (NLP) that focuses on analyzing and understanding opinions, sentiments, attitudes, and emotions expressed in text data. The goal of opinion mining is to determine the sentiment polarity of a given piece of text, such as a review, comment, or social media post. However, opinion mining faces language-specific challenges that differentiate studies in less commonly researched languages from those conducted in English. This article presents a novel process for Turkish opinion mining by comparing various artificial intelligence algorithms. We conducted extensive experiments using an open-source Turkish opinion-mining dataset to ensure transparency and reproducibility. Our research evaluated traditional machine learning, deep learning-based algorithms, and pre-trained transformer models, focusing on optimizing their parameters. We also compared word embeddings with the traditional bag-of-words method. By fine-tuning hyperparameters, our optimized models significantly improved accuracy and F1 scores. The proposed process outperformed existing methods in the literature, providing valuable insights for future research in opinion mining.