In this episode of Reality Series, we focus on the myths and realities of machine translation quality estimation, hosted by Mei Zheng, data scientist, and Alex Yanishevsky, director of AI and machine translation.

Over the past decade, the evolution of machine translation has significantly changed the industry. With that, there are a lot of factors to consider when we look at the quality estimation of machine translation – and this discussion is dives right in, addressing ways that quality is estimated and calculated, and questions like:

- Does a high score mean high quality?
- What role do semantics play in edit rate?
- How does a fuzzy match affect translators?
- How (and when) should we edit and translate profanities?

Jump into the Episode
[01:39] Busting myth #1: Is 90% quality good enough to publish?
[09:58] Busting myth #2: "MT stinks!"
[17:51] Busting myth #3: There is one MTQE.
[24:13] Q&A

Sample translations discussed during Myth #2

en: “Dr. Smith was found guilty of keeping a protected animal in the Atherton Magistrates Court after being charged with removing a scrub python from a resident’s property. She then went through the legal process to appeal the court ruling.”

zh: 史密斯博士在阿瑟顿地方法院被裁定饲养受保护动物罪，因为她被指控从一个居民的财产中移走一条灌丛蟒蛇。随后，她通过法律程序对法院的裁决提出上诉。

es: La Dra. Smith fue declarada culpable de tenencia de un animal protegido en el Tribunal de Magistrados de Atherton tras ser acusada de retirar una pitón de matorral de la propiedad de un residente. A continuación recurrió la sentencia judicial.

ru: Доктор Смит была признана виновной в содержании охраняемого животного в магистратском суде Атертона после того, как ей было предъявлено обвинение в том, что она убрала камышового питона с участка жителя. Затем она прошла через судебный процесс, чтобы обжаловать решение суда.

Resources and links:

• Smartling Website 

In the first episode of our Reality Series, Smartling leaders – Olga Beregovaya (Vice President of AI and Machine Translation), Andrew Batwash (Associate Director of Language Services), and Jack Welde (President) – uncover the myths and realities of LanguageAI.

Language AI is the topic of the hour for the translation and localization industry. To uncover how it improves business outcomes and where the industry is headed, we explore:

• What is Language AI? Is it the same as Machine Translation?
• How can we leverage LanguageAI to build efficiencies and cost-savings into our processes?
• What outcomes have early adopters of LanguageAI seen? What best practices can we learn from them?
• Where do human translators fit in?

Jump into the Episode

[01:30] About Olga and her background.

[02:28] About Andrew and his background. 

[03:41] What is Language AI?

[08:18] How to leverage Language AI to build more efficient cost and time savings.

[11:25] What ways are humans part of the translation process and what value do they add?

[14:22] How companies like Smartling compete with cloud MT providers like Google, Microsoft, and Amazon.

[17:27] Why companies should be trying to incorporate machine translation.

[25:17] Skillsets required to stay relevant in the age of AI.

[28:46] What is MT not good at?

[36:23] Measures and benchmarks that prove Andrew is having success. 

[37:07] Is it possible to use context for content that’s localized with MT?

[41:16] Aside from machine translation, how else is AI being used? 

[45:05] Final thoughts from Olga and Andrew.

Resources and links:

• Smartling Website