Machine Translation at Allegro – Use your model wise but data wiser
A datacentric story about developing an in-house Machine Translation engine specifically for e-commerce purposes, aiming to provide better value compared to off-the-shelf solutions. Our focus is on accurately translating industry-specific terms and jargon, while also creating a scalable and cost-efficient solution. We employ state-of-the-art machine learning methods and we heavily involve human evaluators to continually enhance translation quality. We’ll talk about how: we are making our marketplace accessible to non-Polish speakers globally and contributing to the machine translation community at large.
Bio
Meet Mikołaj Koszowski, an expert in the world of natural language processing. He’s a Warsaw University of Technology alumnus who graduated with honors back in 2019. During his short academic journey, he studied Applied Physics and made waves with his thesis for using the ML approach, he even presented it at The Polish Symposium on Physics in Economy and Social Sciences.
With over five years of experience in the field, Mikołaj has carved out a niche for himself in the realm of machine translation. Notably, he played a role in the success of the WMT 2020 competition, where his team’s system emerged as the winner in the English-Polish translation direction during his tenure at Samsung R&D Poland.
Mikołaj’s journey continued at Allegro, where he brought his expertise to the forefront by working on developing models for automating customer service. Presently, he’s embarked on an exciting venture, spearheading the development of an internal machine translation solution that stands as a linchpin for the company’s international expansion.