ABOUT THIS EPISODE

#24: What Makes Machine Translation Work? - Olga Beregovaya

Boost your CRM with Salesforce – tdv.co/sf

In this episode of The Deep View Conversations, we explore the world of machine translation and artificial intelligence with Olga Beregovaya — Smartling’s VP of Machine Translation and Artificial Intelligence. Olga takes us through the evolution of machine translation, from rule-based systems to statistical models to today’s neural networks and large language models.

She sheds light on how translation has shifted from being a purely linguistic endeavor to one that now sits at the intersection of data science, AI, and human creativity. We dive into hallucinations, under-resourced languages, the rise of synthetic data, and what it truly means to maintain quality in a multilingual world. If you've ever wondered what makes AI translation tick — and where it's going — this is the episode for you.

Get The Deep View — your daily source for in-depth, fact-based reporting on artificial intelligence — in your inbox every morning. Subscribe here! (https://www.thedeepview.co/subscribe)

Connect with us on X, TikTok and Instagram.

Artificial intelligence is a complicated topic, bound by a number of complex threads — technical science, ethics, safety, regulation, neuroscience, psychology, philosophy, investment and — above all — humanity. On The Deep View: Conversations we break it all down, cutting through the hype to make clear what's important and why you should care.

Technology

Science

English

United States

Use API to fetch this episode in JSON

IN THIS EPISODE

Olga Beregovaya

Edit these tags

SHOW NOTES 🔗

00:00:00– Intro

00:01:20– Meet Olga Beregovaya: Linguist to AI executive

00:03:00– Rule-based machine translation: Where it all began

00:05:15– From rules to statistics: The SMT era

00:07:30– Neural networks and the Transformer revolution

00:10:20– Where linguists fit in today’s AI loop

00:12:45– Ad Break

00:14:00– The limits of automation in high-risk content

00:16:30– Has the translator job been reinvented?

00:19:00– The future of post-editing and model training

00:21:30– Can we trust AI without hallucinations?

00:24:00– Building translation systems at Smartling

00:27:00– Terminology, glossary enforcement, and sentence fluency

00:30:00– Translation memory and AI-enhanced fuzzy matching

00:31:30– Smartling’s model strategy

00:33:30– What quality training data looks like

00:35:00– Supporting under-resourced languages

00:38:00– Quality estimation, assurance, and model evaluation

00:40:00– Dealing with linguistic complexity

00:42:30– Rules meet AI: A hybrid approach

00:45:00– Staying at the cutting edge

00:47:00– Final thoughts: Translators, technologists, and tomorrow

00:49:00– Why translation is still a human story at heart

TRANSCRIPT 🔗

Are you the producer of this podcast? Add a podcast transcript

Need Audio-to-Text? Transcribe with Listen411 in Just 60 Seconds

SEARCH PAST EPISODES

00:55:45

SAVE

CLIP

Disclaimer: The podcast and artwork embedded on this page are from The Deep View, which is the property of its owner and not affiliated with or endorsed by Listen Notes, Inc.