Image: © Sundry Photography/Stock.adobe.com
The two organisations have created a machine translation model that focuses on specific categories of information such as product details.
LinkedIn has teamed up with the Science Foundation Ireland (SFI) Adapt centre to improve the level of multilingual content on the platform.
The two organisations have created a machine translation (MT) system that can focus on specific categories of information, such as information about products. This system can currently translate English data into French and German, which will be used to train larger and more accurate classification models.
LinkedIn said the focus on multilingual content comes as its global userbase has grown to more than 900m in more than 200 countries.
LinkedIn staff AI engineer Dr Tatiana Habruseva said the goal is to create an inclusive experience on the platform for everyone, “no matter what part of the world they call home”.
“We want our members to be able to access the professional knowledge and content they are looking for in their native language wherever possible, and this collaboration is one step on our journey to achieving this,” Habruseva said.
However, there are challenges in creating a machine learning system for a specific domain such as LinkedIn, as these models require a large amount of existing translation data.
Working with LinkedIn, researchers at the SFI research centre for AI-driven digital content created a system that can help augment existing LinkedIn data. This ML system can also develop multilingual product classification and recommendation systems for the platform.
The researchers created a parallel test set of product descriptions and then compared multiple ML systems, including “domain-adapted systems trained on specialised data”. Adapt’s deputy director Prof Andy Way said this domain-adapted model appeared to outperform baseline models.
Declan McKibben, Adapt executive director, said the organisation was the “perfect partner” to help LinkedIn with its translation objectives.
“Through our collaborative approach we have succeeded in advancing the MT capabilities within LinkedIn, pioneered new technologies to meet the evolving needs of their users, and strengthened the international research network between Ireland and the US,” McKibben said.
Last September, Adapt researchers conducted a study with the Dublin City University Anti-Bullying Centre to help AI crack down on cyberbullying in different languages.
10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.
