Machine translation and automatic subtitling of gender-focused messages on Instagram. A case study of a Raquel Willis’ speech

As social media platforms increasingly disseminate LGBT and feminist discourses across cultural and geographic boundaries, translation plays a crucial role in ensuring the accessibility of these gender-focused messages. Nevertheless, the widespread use of machine translation may hinder their accurate transmission, largely due to their linguistically biased and insufficiently representative training corpora. This research examines how Instagram’s automatic subtitling features based on machine translation technologies translate LGBT and feminist terminology from English into Spanish, according to the Multidimensional Quality Metrics CORE typology. It also intends to discover if gender bias is present in this tool, which is analysed through a manual evaluation of inclusive language translation practices, drawing on official guidelines. For the analysis, this research takes as a case study a Gender Liberation Movement cofounder’s speech and evaluates its 67 automatically generated subtitles by Instagram’s machine translation tool into Spanish. Finally, the results show that Instagram’s machine translation tool for automatic subtitling inconsistently translates LGBT and feminist neologisms with correctness, the ‘accuracy’ errors being the most prevalent ones. This tool also appears to be gender biased, as it frequently lacks gender-inclusive translation practices.