Content area
This study investigated the comparative performance of two popular sentiment analysis tools, Azure Sentiment Analysis and Google Cloud Sentiment Analysis, applied to Polish and English text data. The analysis focused on two content types: hotel reviews and tweet comments. The findings demonstrate that the type of text and language analyzed significantly influence the accuracy of sentiment detection. Sentiment analysis tools performed worse on informal and unstructured text like tweet comments compared to structured and formal text like hotel reviews. Additionally, language can influence the performance of sentiment analysis tools. In this study, English showed a slight edge over Polish, particularly for tweet comments. When analyzing tweet comments, the Google Cloud Sentiment Analysis function outperformed Azure Sentiment Analysis in both Polish and English, suggesting better handling of informal and ambiguous language. Furthermore, this research addressed limitations of relying solely on pre-assigned sentiment labels from the tools. A custom labeling method was developed, analyzing the tools’raw sentiment scores (e.g., positive vs negative sentiment score). This method achieved a more granular sentiment classification, especially for informal text data. Compared to pre-assigned labels, this approach provided a more nuanced understanding of sentiment and facilitated a more accurate comparison of the tools’performance, particularly highlighting the one tool’s advantage in handling informal language complexities. By highlighting the challenges associated with informal and unstructured text, as well as the importance of considering language variations, this research contributes to the field of sentiment analysis. The findings and the proposed labeling method present valuable opportunities for future research along promising directions, particularly in sentiment-based decision-making for social media data analysis.