MTPE vs Augmented Translation for Better Translation Efficiency

In today’s fast-paced world of translation, the use of technology has become essential to optimize efficiency. Gabriel Fairman dives into the discussion of Machine Translation Post-Editing (MTPE) and Augmented Translation, two pivotal technologies that can shape the future of translation services, especially from a financial perspective.

What is Machine Translation Post-Editing (MTPE)?

Machine Translation (MT) refers to the automatic translation of text using software, without any human interference. Various types of engines can be employed, from statistical and rule-based to neural networks, such as Google, Microsoft, or AWS.

However, as Gabriel Fairman explains, MTPE involves human intervention to refine and correct the machine’s output:

‍“Post-editing refers to the process that any human being engages in to make sure that they’re reading over the machine translation propositions and adjusting them as necessary to ensure they are viable translations.”

The complexity of MTPE varies significantly, and in some cases, it can be as difficult as translating from scratch. Gabriel warns that post-editing doesn’t always guarantee efficiency:

‍“It can often be very misleading, very complex, and a lot more painful than actually translating from scratch.”

The Challenges of MTPE

Although MTPE holds the promise of speeding up the translation process, it doesn’t always deliver as expected. Some of the key issues Gabriel highlights include:

• Unpredictable quality: Machine translations can be inconsistent, sometimes producing flawless sentences, while other times riddled with errors.
• Hidden mistakes: Errors may be buried in the text, such as cultural misappropriations, literal translations, or even embarrassing mistakes.
• Varied subject matter: Performance varies across fields. For example, MTPE might work well for finance but poorly for marketing.

Fairman sums up the limitations of MTPE:

‍“The basic benefit of machine translation is that it delivers this promise of greater efficiency, but the challenge is that MTPE is ungovernable.”

Enter Augmented Translation

Augmented Translation introduces a new way of working with machine-generated content. This technology focuses on context-sensitive translations, ensuring that the translator’s edits are continuously integrated into the system, improving the translation quality over time.

According to Gabriel, the major advantage of augmented translation lies in its dialogical nature:

‍“The beauty of the context-sensitive translation is that while it seems on the surface very similar to MTPE, it’s operating in a very different paradigm. It’s learning from your choices, it’s getting better, and more in tune with your style, diction, tone, and syntax.”

Key Differences Between MTPE and Augmented Translation

• Continuous learning: Augmented translation systems adjust based on the translator's inputs, improving suggestions for future translations.
• Semantic verification: These systems can pick up on nuances that traditional QA tools often miss, including gender bias, cultural misappropriations, and inappropriate tones.
• Improved efficiency: By reducing the number of repeated errors and incorporating real-time learning, augmented translation reduces cognitive effort and increases satisfaction for translators.

Financial Considerations

From a financial perspective, augmented translation can offer more stable and predictable outcomes compared to MTPE. While MTPE can result in high variability in quality and effort, augmented translation streamlines the process by continuously refining the translation based on user input.

Fairman points out that while the edit distance in augmented translation can be a helpful metric to measure effort, the true challenge lies in connecting that to cognitive effort and overall quality. He notes:

‍“It’s still going to be challenging to connect edit distance to actual cognitive effort, and it’s even harder to connect cognitive effort to translation quality.”

Conclusion

In conclusion, while MTPE had a strong promise of efficiency, augmented translation is proving to be a more sophisticated and reliable alternative. It not only reduces the time spent on repetitive edits but also offers a more organic and fulfilling experience for translators.

Gabriel concludes with a strong recommendation to try augmented translation:

‍“It’s different in that it’s significantly better, more sophisticated, dialogical, ever-evolving, and it just feels organic and alive.”