Beyond the Binary: Making Machine Translation Inclusive

Track: Multilingual AI | AI1 |
Thursday, June 6, 2024, 1:30pm – 2:00pm
Held in: Pembroke room
Presenters:
Cristina Anselmi - Electronic Arts 
Amanda Hawthorne - Electronic Arts
Host: Johan Lourens

Over the past year, the presenters took a focused look at how they could make their machine translation engines more inclusive and above all less binary. They had a few ideas but the one that seemed most viable was creating their own data sets to facilitate non-binary language training internally. It was a massive task, which needed the collaboration of many teams and detailed planning. Did they make huge steps towards ending binary machine translations? Yes. Have they still got more work to do? Also yes. In this session, the presenters will talk you through how they went about training their machines to stop being binary and the lessons they learned.

Takeaways:

• A fundamental understanding of how generative AI can be used to create multilingual product descriptions and other kinds of content, resulting in cost savings and increased efficiency;
• The hows and whys of leveraging language data to train and implement a customized generative AI system for content generation;
• A clear roadmap for working with a technology provider to define and execute a proof of concept for real-world generative AI use cases