Interpretation of sentiments by popular machine models: A study with Saudi student-interpreters

Abstract

This mixed-methods study explores which of the four free AI applications, Google Translate, Microsoft Translator, Microsoft Bing AI, and QuillBot, are most successful in decoding emotional messages in the Arabic-English language pair. A dataset of positive, negative, neutral, and mixed emotional tones was used for interpretation test with the AI programs, and results were rated by three language experts for sentiment accuracy, contextual understanding, and cultural appropriateness. Findings revealed that AI tools are quite effective in recognizing obvious positive and negative emotions but experience some issues when it comes to ambiguous or culturally specific speech content. Microsoft Bing AI showed the least difference with human judgments and hence found to be the best for Arabic-English emotive interpretation. The results indicate that AI can serve as a resource for Saudi student-interpreters and confirm the importance of human-AI cooperation to obtain the correct context-specific and culturally sensitive interpretation.