The MT Post-editing Experiment: Questions And Answers
We received a lot of questions during our webinar that kicked off the MT Post-editing Experiment. We were not able to answer all of them, so we are providing the answers below:
Q: Where is the link for setting up the MT post-editing starter kit?
Q: Does the post-editing effort calculation consider sentence length while making the calculations per word? A single word corrected in a two-word sentence vs a 10-word sentence?
A: Yes, it does.
Q: What happens if a client wants a quote before we proceed with the translation?
A: It is true that the post-editing analysis can only be launched after the translation is made. It is clear that you will in most cases need to produce a quote before you start translating. You should first run a standard (we call it a default analysis in MemSource) analysis that will analyze the source to produce a quote. Based on your experience you could adjust the quote by an additional estimated brought about by machine translation post-editing. To be on the safe side, you could leave the quote as it is and if you make any additional saving during the translation process, this will increase your margin.
Q: Is there any possibility how to deploy Trados in your workflow?
A: Not really. Memsource is a replacement for Trados, not an addition.
Q: As to the experiment: Are there any rules for PE? Understandability - grammar - style? Highest possible target text quality?
A: We would not be able to objectively measure these. The objectives of the experiment are listed here.
Q: Could you please confirm if it’s possible to use just the analysis feature in Memsource without the rest of the system?
A: This is not possible as the analysis draws data from a number of Memsource components, e.g. Memsource XLIFF file, Memsource Editor, Memsource Cloud…
Q: Any chance we can use another MT solution than Bing or Google? Asia Online?
A: Yes, any MT engine with a decent API can be set up to work in Memsource Cloud. Contact Memsource support.
Q: Do we have to buy the Google API or any other API to use machine translation?
A: Yes, you have to get your own API key, Microsoft has a free trial period.
Q: What are the privacy settings for the PE environment?
A: Any data that you upload to Memsource Cloud remain private, confidential, and secure.
Q: What exactly are these “MT match ranges”?
A: These are e.g. 95% to 99% or 85% to 94% as can be seen in a post-editing analysis.
Q: Will it be possible to get a recording of this Memsource demo?
A: Yes, it has been posted on our YouTube channel.
Q: This is really cool; however, I am concerned about sending customer segments off to Microsoft Translator or Google - they both have license agreements that give them rights to the content submitted “in perpetuity” - your customer might not be happy about that…
A: If you have concerns, I suggest you use another machine translation service. Any MT technology with a decent API can be easily connected to Memsource Cloud.
Q: Could it be possible to create an option that would allow us to choose whether to have tags displayed in MT output or not?
A: Yes, this would be possible.
Q: Can Memsource integrate Systran Desktop 7 premium?
A: We can integrate Systran server products, not desktop.
Q: What would be the process to connect to a “secure” MT engine (rather than MS translator or Google Translate)?
A: To connect another MT engine, even a custom one, is possible. Please contact Memsource support for details.
A: These or better - please contact firstname.lastname@example.org for details.