What is SmartMatch?
SmartMatch is a feature that is unique to Smartling which can help you get the most out of your translation management platform.
When a new source string matches perfectly with a unit in your Translation Memory (i.e. the exact same source text has already been translated before), SmartMatch applies the existing translation from your Translation Memory in a fully automated way.
With SmartMatch, the Translation step of your workflow is skipped. Depending on your preferences, the existing translation from your TM can be sent to a revision step, so it can be amended if needed, or it can be automatically published.
This means that SmartMatch can produce even greater time and cost savings than a fuzzy match (which still needs to go through the Translation step of the workflow).
A string needs to fulfill the following criteria to qualify for SmartMatch:
- The same exact source text has been translated before and the corresponding translation is stored in your Translation Memory.
- The previous translation needs to be published.* Translations which are still in progress are not used for SmartMatch.
*With the exception of the SmartMatch setting “Restore translation”. - SmartMatch can only be used for an entire string. If a string has multiple segments, a qualifying match needs to be available for the entire string or for each segment.
- The new source string, as well as the translated text from your TM, need to use the same number of placeholders as the original source string, in the same position within the string. If a TM entry presents inconsistencies in placeholders between the source and the translation, it won't be reused for SmartMatch.
- If there is a character limit set up for the new source string, SmartMatch only works if the translation from your Translation Memory entry does not exceed that limit.
- If the source string contains HTML tags, all of the same HTML tags need to be present in the TM entry.*
*Unless the SmartMatch settings “Text ignore markup” or “Text remove markup” are enabled. See SmartMatch Rules for more information. - The new source string needs to have the same whitespaces as the TM entry.*
*Unless the SmartMatch setting “Text ignoring whitespace” is enabled. See SmartMatch Rules for more information. - The new string needs to follow the same capitalization as the TM entry.*
*Unless the SmartMatch setting “Text ignore case” is enabled. See SmartMatch Rules for more information. - If desired, SmartMatch can be limited to strings with the same string variant as the translation unit from your Translation Memory (by enabling the setting “Text with variant”). See SmartMatch Rules for more information.
How SmartMatch Works
When SmartMatch is enabled, once new content is authorized for translation, content moves into the Processing step where SmartMatch initiates, and automatically checks your Translation Memory for an exact match. If an exact match is found, it is automatically applied and the string skips the translation step and moves to the step configured in the SmartMatch Rules - no human effort is required.
All SmartMatched translations write to the Translation Memory, regardless of the result configured.
Tip: Even after a string has been authorized and gone through the Processing step, SmartMatch continues to check for a potential match. SmartMatch only works if the string has no saved translation; it cannot overwrite existing translations.
Account Owners and Project Managers can customize SmartMatch behavior with SmartMatch Rules, and it can be customized for each Translation Memory source within a Leverage Configuration. Once a Translation Memory source is created, you can then enable and configure SmartMatch.
How to Enable or Disable SmartMatch
In your Leverage Configuration, the SmartMatch link text ("Enabled" or "Disabled") indicates whether at least one SmartMatch rule is currently enabled or disabled for each Translation Memory source.
Enabling SmartMatch
To enable or customize SmartMatch, follow these steps:
- Access your Leverage Configuration (either from Account Settings > Linguistic Assets > Leverage, or from your Project Settings > Linguistic Package > Leverage: Read from).
- In the list of Translation Memory sources, under SmartMatch, click Disabled/Enabled. This opens the SmartMatch Settings dialog, where you can configure SmartMatch Rules.
- Select the desired result for each SmartMatch rule from the dropdown.
- If you choose First Revision Step, you have the additional option of selecting or deselecting Enable Pre-Publish.
- For recommendations on which result to select for each rule, please visit this article.
- Click Save Changes to put the customizations into effect, or Cancel to exit.
Disabling SmartMatch
To disable SmartMatch, follow these steps:
- Access your Leverage Configuration (either from Account Settings > Linguistic Assets > Leverage, or from your Project Settings > Linguistic Package > Leverage: Read from).
-
Click on the link text indicating that SmartMatch is currently "Enabled" for the Translation Memory source.
- This opens the SmartMatch Settings dialog, where you can configure SmartMatch Rules.
- For each SmartMatch rule, select the result "Disabled" from the dropdown.
- Click Save Changes. SmartMatch is now disabled for this Translation Memory source.
SmartMatch Rules
SmartMatch has different levels of precision. Customizable rules allow you to consider the context in which the previous translation was used, as well as some formatting elements like spaces or upper and lower case letters.
Info: For detailed information about all available SmartMatch rules, please visit this article.
SmartMatch Results
Each SmartMatch rule can be enabled with one of the following results.
When SmartMatch Rules are set to an Edit or Review step, ensure your workflow contains this type of step. Otherwise, SmartMatch will automatically publish any matches. For example, if a SmartMatch rule uses the result "First Revision Step (Review)", but your workflow only contains an Edit step and no Review step, any SmartMatches for this rule will be automatically published.
- Published: SmartMatched strings bypass all workflow steps and will immediately be published.
- First Review Step: SmartMatched string is automatically moved to the first Review-type step, bypassing all previous workflow steps. If there is no such workflow step, strings will be immediately published.
-
First Edit Step: SmartMatched string is automatically moved to the first Edit-type step, bypassing all previous steps. If there is no such workflow step, matching strings will be immediately published.
- Workflow steps with the step type "Post-Edit" are not included in this setting. To send SmartMatches to a "Post-Edit" step, use the First Revision Step (Edit, Review or Holding) setting.
- First Revision Step (first post-translation step): SmartMatched string is automatically moved to the first step of any type: Edit, Post-Edit, Review, AI Review, Hold, Internal Review, Quality Evaluation, LQA, or Desktop Publishing. If the workflow only contains a Translation step, matching strings will immediately be published.
- First Hold Step: SmartMatched strings will jump to the first Hold step in the workflow. If there is no Hold step in the workflow, the SmartMatched strings will move to Published.
- Last Revision Step (final step before Published): SmartMatched strings will be sent to the step directly preceding the Published step. This may be any type of step. If the workflow only contains a Translation step, matching strings will immediately be published.
-
First Internal Review Step: SmartMatched strings are automatically moved to the first Internal Review-type step.
-
Disabled: If you prefer not to use SmartMatch for a given Translation Memory, ensure that all SmartMatch Rules are set to Disabled and then click “Save Changes”.
- When SmartMatch is Disabled, your translators can still see suggestions from this Translation Memory, but SmartMatch will not automatically move the content to a different step.
When choosing to send content to any Edit, Revision, or Hold Step (2-7 above), you can also choose to enable pre-publish. This feature is useful if SmartMatched content remains in a Review, Edit or Hold step longer than desired. Enabling this setting will allow SmartMatched content to "prepublish" so translations are available to use and are visible in their destination (e.g. on a website via the GDN) before reaching the Published step of the workflow.
Considerations
Checks Continuously
SmartMatch runs continuously on untranslated strings that are in progress. This happens not just the moment strings are authorized, but also after they have been authorized. More precisely, after a translation is published or prepublished, it checks to see if untranslated strings that have been authorized for translation could SmartMatch from the newly published translation. However, this is not guaranteed to be instantaneous. Therefore, SmartMatched strings may appear in the Processing or Translation step briefly.
SmartMatch is a String-Level Action
In order for SmartMatch to kick in, you need to have a translation available in your Translation Memory for the entire string. If a translation is not found for the entire string, SmartMatch attempts to find a match for each segment separately. If matches are found for all segments, then SmartMatch is deemed successful.
If there is an existing translation available for just one segment, SmartMatch won’t work. This is to avoid SmartMatching only one sentence within a paragraph, out of context.
Only Previously Published Translations Will SmartMatch
SmartMatch only works if your previous translation has been fully completed and is sitting in the Published step. This ensures that we don’t automatically reuse translations that that have not previously gone through the full workflow and reached the Published step.
How SmartMatch Handles Plural Strings
Smartling supports two CLDR forms for English: ‘One’ and ‘Other’ (corresponding to the singular and plural forms in English grammar). SmartMatch only uses the ‘Other’ form to find matches for plural strings. This means that if a TM Match is found for the ‘Other’ form, SmartMatch will apply all form translations from the TM Match, even if the ‘One’ form does not match the ‘One’ form of the new source string. This can sometimes lead to the wrong number of placeholders or text being applied by SmartMatch for the ‘One’ form.
SmartMatch does not match different source forms for many reasons, mainly because forms mean different things in different languages, and some forms may or may not exist in various languages. For this reason, SmartMatch relies on matching using the only form required for every language: the ‘Other’ form.
Translation Memory Quality
As SmartMatches bypass your translators completely, they can only ever be as good as the translations stored in your Translation Memory. This means that the quality of your previous translations is a critical factor to consider.
Think of SmartMatch as a magnifying glass on translation quality. If your Translation Memory contains great quality translations, SmartMatch will help you re-use them at a larger scale and reduce the risk of incorrect re-translations. On the other hand, if your Translation Memory contains poor or even incorrect translations, SmartMatch will keep propagating any mistakes that your translators may have made in the past.
For this reason, SmartMatches should only be sent directly to the Published step if you can fully trust the quality of your Translation Memory. If you are unsure about the quality of the translations stored in your translation memory, it is recommended that SmartMatch Rules are set to an Editing or Review step, so a human can validate that quality.
Restoring Previously Saved Translations
If a string is unauthorized while still in progress, any saved translation for that string will be removed. A string can be unauthorized in several ways, including:
- The Job with the string is canceled.
- The language is removed from the Job.
- The source file is deleted from Smartling.
- A new version of the source file is uploaded to Smartling, and the string is removed in the updated version.
- Translation is stopped via the Stop Translation option in the Strings View actions menu.
If the string’s translation was saved to the Translation Memory, you can restore it by enabling the SmartMatch rule Restore translation: Exact match for previously removed unpublished translations. Once this rule is enabled, you can re-trigger SmartMatch using the Recheck for SmartMatches option in the Strings View. This is the only way a string can SmartMatch against itself.
If enabling the Restore translation: Exact match for previously removed unpublished translations rule, consider selecting the First Revision Step (first post-translation step) result to minimize the probability of an incorrect translation being copied to multiple strings.
Recheck for SmartMatches
SmartMatch compares authorized but untranslated strings against existing translations in your leverage configuration to automatically apply translations. If your project is configured to use SmartMatch, strings are checked for a SmartMatch when they authorized.
However, there are times when you might want to recheck your authorized strings for a SmartMatch. Examples are:
- You've imported translations or added a TM after the strings were authorized.
- You've enabled or updated SmartMatch settings since the strings were authorized.
Rechecking SmartMatch checks what matches exist currently in the TM for untranslated strings.
To recheck strings for SmartMatches:
- Go to the Strings View
- Select the string(s) you want to recheck for SmartMatches
- Click Actions > Recheck for SmartMatches
Strings must first be authorized before they can be SmartMatched.
SmartMatch Reporting
Account Owners and Project Managers can get a data visualization of the number of SmartMatches across the entire account with the SmartMatch Leverage Report.
Video Tutorial
Timestamps:
What is SmartMatch? 00:14
How does it work? 02:55
Important considerations 03:43
➤ SmartMatch is a string level action 03:57
➤ SmartMatch only uses (pre)published translations 04:32
➤ The quality of your TM matters 05:17
➤ Your workflow needs to match your SmartMatch configuration 06:24
How to configure SmartMatch 06:43
a. How to disable SmartMatch 07:23
b. How to enable SmartMatch 07:44
➤ Main settings: Text with variant / Text without variant 07:51
➤ 4 additional settings to ignore formatting elements: 11:24
• Text ignoring whitespace 12:06
• Text ignore markup 12:54
• Text ignore case 15:55
• Match any 16:55
➤ Restore Translation 17:45
Enable Prepublish: How does it work? 20:07
When does SmartMatch kick in? 20:58
➤ When new content gets authorized 21:05
➤ Continuous SmartMatch 21:24
Troubleshooting: Reasons why a string might not SmartMatch
There are several possible reasons why a string with a matching Translation Memory entry might not SmartMatch. If you expected a certain string to SmartMatch but it didn't, please check the following:
- Has the string been authorized and gone through the Processing step?
SmartMatch only kicks in once a string has been authorized, in the Processing step of the translation workflow (or in the Translation step, as a result of Continuous SmartMatch). - Is SmartMatch enabled in the Linguistic Package?
SmartMatch only works if it is enabled for the Translation Memory which contains the matching translation.
In the CAT Tool, you can use the Language Resources panel to check which Translation Memory the match originates from. - Are all criteria for the enabled SmartMatch setting(s) met?
If one or more criteria are not met, SmartMatch does not kick in.
For example, SmartMatch might be enabled for "Text with variant" only but the string might use a different variant from the matching Translation Memory entry.
The variant of the Translation Memory entry can be checked in the Activity log. - When was SmartMatch enabled?
If SmartMatch was enabled only after the string was authorized and went through the Processing step, then it will not kick in automatically. Instead, it needs to be triggered manually from the Strings View, by selecting the action "Recheck for SmartMatches". Once you recheck for SmartMatch, the latest SmartMatch settings will be taken into account. - Does the string already have a saved translation?
SmartMatch only applies a translation to strings which do not have a saved translation yet. It does not overwrite any existing, saved translations. - Is the matching Translation Memory entry published?
As a precaution, SmartMatch only works if a TM entry is "published". This means that the original translation has either reached the Published step, has been pre-published or imported from a TMX file. Unpublished translations are not used for SmartMatch, since they are still in progress and subject to change. The status of the matching TM entry can be checked in the CAT Tool, under "Language Resources". - Is there a matching translation available for the entire string?
In the CAT Tool, a string might be split into multiple segments, if it contains more than one sentence. If a match is available for a certain segment only (but not for the entire string), SmartMatch will not kick in. SmartMatch always works on a string level, not on a segment level. - Are any placeholders missing in the translation?
If a string contains placeholders for variable elements, SmartMatch only works if the translation uses the same placeholders as the source text. If any placeholders are missing in the translation, which is stored in the TM, that TM entry cannot be used for SmartMatch. - Does the translation exceed the character limit?
If there is a character limit set up for a string, SmartMatch only works if the Translation Memory entry does not exceed that limit. This is an important precaution to avoid any UI or formatting issues.