Can Neural Machine Translation Reduce Gender Bias?-0001 November 30
A recent blog published by Google looked at their efforts to reduce gender bias in Google Translate. Google Translate learns from millions of translated examples that exist on the web. The new development in the system addresses gender bias by providing feminine and masculine translations for some gender-neutral words. Previously only one translation would be returned for a query, even if the translation could have a feminine or masculine form. But is this development enough?
ADAPT researcher at DCU, Eva Vanmassenhove, recently spoke to Slator, an organisation focused on the language industry, to discuss the limitations in the Google approach and discuss her research in Neural Machine Translation (NMT). Neural networks try to mimic how the human brain works and NMT significantly outperforms old statistical machine translation models. Google claims this new NMT system can “reliably produce feminine and masculine translations 99% of the time.”
Vanmassenhove, who is focusing on how to integrate more linguistic knowledge into a machine translation system as part of her research in the ADAPT Centre, noted: “It is not translating from languages such as French, Italian, Portuguese or Spanish into English that is problematic, but the other way around. Different languages have different ways of expressing gender and it is important to realise that there won’t be one solution that fits all”.
She continued: “Even context-aware NMT systems that can take some context into account while translating would have a hard time getting this right, as (cross-genre) gender prediction, especially for languages that do not mark gender explicitly (such as English), remains an unsolved task.” For example, ‘I am a nurse’ will be given feminine translations while ‘I am a surgeon’ will result in masculine ones.
Although Google’s system is much improved there are still shortcomings due to limited language coverage but the blog post indicates that they are just at the first stage of reducing gender bias in machine translation.
Posted by Catherine O'Connor, Head of External Relations, School of Computer Science and Statistics, Trinity College Dublin