Machine Translation Workshop Series: What to Expect
With the successful Machine Translation Workshop Series kick-off behind us, we’d like to take a moment to thank the diverse panel for sharing their MT troubles and triumphs. The panel’s use cases stimulated a productive brainstorming session with the event’s participants. The collective mission was to uncover the main pain points and blockers of MT, which we’re preparing to address together in future workshops.
Digging into the data
During the brainstorming session, we accumulated over 60 topics of interest to our participants. We analyzed these ideas, identified recurring themes, and grouped the suggestions accordingly. The overarching themes were:
- Source content and expectations for MT
- Efficient MT post-editing
- Evaluating quality of different MT systems
- New evaluation methods: BLEURT, BERTScore and COMET
- Is training a customized MT engine worth it?
- Integrating MT into l10n workflows
- Memsource MT features
First up: Budgeting and financials
The next workshop will address Financial Implications and Opportunities with Machine Translation. Participants will delve into different methods for calculating the cost of machine translation workflows and the impact of MT on ROI analysis.
Is the edit-distance payment fair for linguists? How do you reassess post-editing costs as MT improves?
Workshops to look forward to
As we finalize our plans for future workshops, let’s look back at the topics the participants asked us to address.
Source content and expectations for machine translation
Has the time of raw machine translation arrived? What about creative content–will MT ever be a suitable option?
Efficient machine translation post-editing
With machine translation post-editing (MTPE) being the dominant translation strategy for Memsource users in 2020, it’s no surprise that it was a hot topic during the brainstorming session. Is MT plus human review the new normal? Where do you draw the line between light and full post-editing?
Evaluating quality of different machine translation systems
With continuous MT engine improvements, how do you effectively evaluate the quality of different MT providers? Which language pair and domains are best suited for machine translation?
New evaluation methods: BLEUERT, BERTScore and COMET
How are the newer approaches to the scoring different to the standard metrics? If these correlate better with human judgement, can we go ahead and simply replace the standard metrics with them or do we need to change how we understand them?
Is training a customized machine translation engine worth it?
Custom machine translation engines need maintenance. How often do you need to retrain your engines? What are the best practices for improving the quality?
Integrating machine translation into localization workflows
How do you integrate your MT engine with your translation management system? How can you avoid possible obstacles when configuring MT into your workflow for the first time?
Memsource machine translation features
What are the benefits of using Memsource Translate compared to working directly with an MT provider? How is Memsource keeping up with the fast-paced MT industry?
We look forward to tackling these challenges together! Join us for hands-on workshops to equip you and your business to leverage machine translation to its fullest potential.