Can artificial intelligence mirror the human’s emotions? A comparative sentiment analysis of human and machine interpreting in press conferences

Abstract

The advancement of machine interpreting (MI) has the potential to revolutionise the field of interpreting. However, concerns persist regarding the capacity of MI to accurately convey emotions in original discourse, especially in high-stakes settings such as press conferences. To address this, the present study investigates the sentiment mediation in human interpreting (HI) and MI when rendering source speeches during press conferences. Employing the Linguistic Inquiry and Word Count (LIWC), specifically LIWC2015 and LIWC-22, a comprehensive sentiment analysis was conducted on a self-built corpus comprising Chinese source speeches and their corresponding English interpreting produced by human interpreters and machines. The findings reveal that both HI and MI demonstrate a comparable capacity to convey sentiments, with MI exhibiting human-like patterns in handling emotional content in original discourse. However, compared to human interpreters, MI plays a more active role in attenuating negative emotions while accentuating positive emotions. These results underscore the enhanced ability of AI-powered MI to emulate human interactions and generate renditions that align with human values, highlighting its potential to effectively facilitate cross-linguistic communication. The study contributes to the growing body of research on the impact of interpreting technology on communication dynamics and the evolving nature of interpreting in the AI era.