What’s an ICE match?

An in-context exact match, or ICE match, is a blocked segment that will appear on our interface in reviews. ICE matches are triggered when the content of a segment is recognised as something that’s been translated with us before, and where the segments preceding and following it are exactly the same as they were in the previously completed translation.

ICE matches will appear as blocked segments in grey, though note that they are different from medium confidence segments, which also appear in grey. These are differentiated from normal editing segments which appear in white. 

We’ve created a video that should display all the different types of segments you’re likely to see when working with reviews from now on, so please take a few minutes to look at it below:

ICE matches are a feature that are standard across the translation industry, and we’re introducing them on the Unbabel interface to make life easier for you and increase the quality and consistency of our translations.

These matches and their deployment on the interface are decided by the customer and their individual preferences, and not by Unbabel or our systems. As with any other task or review, it’s your job to work with the materials you have, always taking into account the customer’s preferences via the client instructions.

One final thing to note is that at the beginning and end of texts, the system only has one segment either preceding or following the matched segment to work with, so it will trigger an ICE match from just one segment, where applicable.

We understand you might have questions about this, so please take a look at these FAQs below:

What are the different kinds of segments I can now see in reviews?

What do I do when I see an ICE match?

How often can I expect to see ICE matches?

Why are we introducing ICE matches to the Unbabel interface?

What happens if there’s an error in an ICE match?

What are the different kinds of segments I can now see in reviews?

We really recommend taking the time to watch our video on these new types of segments, but we’ve also laid out all the different examples of the segments you might see on the interface for reviews from now on.

1. A normal segment - just work with this as you normally would on any other review task.
2. A medium-confidence segment - you can read more about these here, but in brief we recommend you just scan them for errors as quickly as you can and move onto a more pressing segment.

3. ICE match segment - this is where you cannot edit this. Check out the section below for more info.
4. ICE match with character limits - this is where an ICE match coincides with a character-limited segment. Here, you’re allowed to edit it where necessary, but remember that this is an ICE match so is highly likely to be well-translated already.
5. ICE match with formatting tags - here you should use the same guidance as you normally would with formatting tags whilst being aware this is an ICE match.
6. ICE match with a glossary - here, the segment remains uneditable, even with the glossary term included. You should not change anything about it, though you can report if you need to using the flag button.

What do I do when I see an ICE match?

Hopefully, you’ll just think ‘great!’, give it a quick scan, then move on to the next segment. ICE matches are there precisely because they don’t need attention, so just check them out for context and then move on to the segments that do require your expert attention.

How often can I expect to see ICE matches?

The frequency of ICE matches depends entirely on the customer’s preferences and the nature of the content, and so it will depend on the availability of reviews from customers that use this feature. This will, in turn, vary by language pair. However, if you’re working frequently with many different customers, you can expect to see them quite a lot.

This doesn’t mean that every segment will be an ICE match - remember that in order for a match to exist, it must have the same context before and after, so by definition there will be plenty of segments that are normal. And, finally, segments that have formatting tags, other tags, or character limitations can be used for ICE matches, but you’ll be able to edit them (you can see examples of this above).

Why are we introducing ICE matches to the Unbabel interface?

ICE matches are standard across the industry - you might be more familiar with them under the terms context match, 101% match, or guaranteed match. A quick search on Google will find you lots of different results.

As they’re standard across the industry, this means that both clients and translators are looking for this feature on the Unbabel platform, and it’s up to us to provide it. For those of you who haven’t come across this feature before, we think it will be a practical tool to help you use your time efficiently and focus on what matters in producing an excellent translation.

What happens if there’s an error in an ICE match?

Essentially, this isn’t something you should worry about. While you are able to report it by pressing the flag button, you should do this, then continue to do the work as best as you can.

These segments were blocked by the customer and it was their choice to approve it, so anything you produce will fit their specifications. On top of this, if your work is evaluated at some point in the future, you will not be marked down for any errors in the segments.

If you choose to unclaim a job due to an error in an ICE match, which we would not advise, you can also report it via the unclaim function.