I translated a list of SEO keywords from German to English (UK) using Baseline LC and Studio's pre-translate batch function. Then I had the same list of keywords translated by Baseline LC and DeepL, using the MT Comparison tool. To my surprise, the LC translations differ vastly (the results obtained with MT Comparison tool being significantly worse than via pre-translate).
How is that possible? Maybe somebody from SDL cast a light on this? I shall gladly email my results to you, but I can't post them on this forum.
Hello Daniel Hug ,
I have spoken to one of the LC developers about this-
We can't be sure how DeepL works exactly, but suspect that DeepL connects to different API's for the baselines.
With regard to Adaptive MT from LC and the LC Baselines, these should be the same and will be looking at investigating this further but I can't give any concrete answers just yet.
Hello Steven Whale
I actually did not notice a discrepancy with DeepL, just with LC.
And sorry for not communicating more clearly, but I used the LC Baseline Engine both times, and got very different results just depending on whether I used "Pre-Translate" (the batch function) or the App "MT Comparison". I was just wondering. Using the app, many German words did not get translated at all.
Hi again Steven Whale
reading this: https://community.sdl.com/product-groups/translationproductivity/f/openexchange_applications/27082/feedback-for-mt-comparison-plugin I think this whole thread belongs to the Apps section since the behavior is caused by the MT Comparison Plugin, not by the LC.