Unbabel, a provider of translation platforms for multilingual customer service at scale, is partnering with Carnegie Mellon University, INESC-ID, and Portugal's Instituto de Telecomunicações, to improve multilingual conversational chat. The Multilingual AI Agent Assistants (MAIA) research project will seek to augment customer service agents with artificial intelligence to deliver chat in 30 languages and improve customer satisfaction through human empathy.
"While chat is becoming one of the preferred means for customer service, it currently faces important limitations that revolve around supporting customers in their native language to drive better customer satisfaction, while maintaining the human-to-human experience," said Andre Martins, vice president of AI research at Unbabel, in a statement. "The MAIA research project will enable customer service and support professionals to capture context and establish empathy with customers by infusing a machine and human approach, making the process more seamless and accurate."
During the next three years, the research team proposes to build a toolbox of machine learning technologies, including context-aware machine translation, automatic answer generation, and conversational quality estimation, for online multilingual customer service. It will also help agents with new dialogue-oriented productivity tools and expand Unbabel's quality estimation technology to assess conversational quality.
"As the global economy becomes even more connected, it is more important than ever to be able to communicate effectively across cultural borders," said Graham Neubig, an associate professor at Carnegie Mellon University, in a statement. "Tackling this task through machine translation is challenging due to the need to consider relevant conversational context in order to generate precise, polite, and culturally appropriate translations. At the same time, the interactive nature of conversations poses a number of new opportunities for translation technology. The MAIA project will advance natural language processing and machine learning technology to tackle these challenges and opportunities, increasing the speed, quality, and user experience of multilingual machine-mediated conversations."
In addition to MAIA, Unbabel's research team is also researching the following:
- Scribe: Partnering with INESC-ID, this project aims to develop a technical solution for automatic transcription, translation, and subtitling of audiovisual content.
- Unbabel4EU: This project aims to build an advanced language engine capable of human-quality translation between any pairing of the 24 official languages of the European Union.