Translation memory (TM)
Translation memory typically consists of the last known good translation for a given segment (text).
Translation memory is assigned and segmented by the following:
See the list below for a summary of how to maximize the use of translation memory in XTM:
Importing translations from other TMS packages
Translation memories can be imported into XTM from a TMX (the standard Translation Memory file format) or from translations stored in XLIFF files or Excel files.
Tip
For best results, we recommend using the TMX v 4b format.
Once the translations are imported via the TM file, the source (segment) text will need to be uploaded to a new project or an existing project's files tab.
Take note of the following items in regards to segmentation:
Source strings in XTM are segmented by individual sentences. This default setting can be changed.
The advantage is that XTM can incorporate ICE matching and provide smaller segments to translators, which increases their quality and efficiency.
Many other TMS vendors segment strings differently, oftentimes providing larger segments to translators. In such cases, the TM that is imported into XTM from the other TMS vendor will not always provide perfect matches.
Applying context to TM
Note
A single word or string may have multiple meanings. For example, an apple can be a fruit or a computer brand. Applying context will allow XTM to provide better suggestions for any segment and improve the accuracy of the TM Auto-Fill functions.
Tip
XTM has the ability to add custom context rules not indicated here. Please contact your XTM point of contact for more details.
The following methodologies can be used to apply context to segments stored in translation memory:
ICE (context) match
An ICE (In Context Exact) match is a translation that when matched against what's stored in TM, guarantees a high level of appropriateness because it has the same context
Important
In XTM an ICE match occurs when the source content in the current, previous and next segment is an exact match with the current, previous and next segment in TM (I.E the same location in a paragraph). ICE Matching is most relevant when source content is segmented by sentence as opposed to paragraph.
XTM Determines if an ICE or Leveraged Match exists using the following logic:
If the source text is identical to the one stored in TM, then XTM checks if a segment has a segment ID.
If the segment has a segment ID then XTM tries to find if an ICE (context) match exists by matching the source text segmentID with the one stored in TM for that string.
Tip
To include the segment ID in the TM please click here for instructions.
If the segment does not have a segment ID then XTM tries to find if an ICE (context) match exists by matching the source text alone.
When neither of the segmentID / context conditions is met, a leveraged match is returned.
When the translation is saved and the translation becomes approved for TM, XTM saves the translation in the TM and includes the segment ID (if activated) along with the source text, target text, and context (ICE match) by default.
Consider the segment We painted it red last week as used in the following paragraph:
My house used to be yellow.
We painted it red last week.
It looks very different now.
Is not a 100% ICE Match but is a Leveraged Match with the following:
My house used to be very ugly.
We painted it red last week.
It looks nicer now.
The original context was about a color scheme, but the context in the second paragraph is about beauty. Therefore the translation for the segment We painted it red last week may contain different adjectives to describe its beauty versus its.
TM penalty profiles
Penalty Profiles penalize the percentage match of a fuzzy match.
Tip
Example: Segment X has an 80% Fuzzy Match against an entry in TM. However, the TM entry was stored for the prescription drug product whereas the translation is applied to over-the-counter drug products. Both are different products and certain segments have slightly different meanings between products. Applying a TM penalty of 5% lowers the fuzzy match from 80% to 75% accordingly.
XTM can apply penalty profiles based on the following criteria:
Tags.
TM Status (Approved or Not Approved).
XLIFF:doc status (Translated, New, Rejected, Validated, Proofed).
Target Language for all projects or per project.
Note
This penalty (Target Language) supersedes all other penalty profiles when more than one is triggered.
Note
The only languages that can have an associated TM Penalty Profile are: Arabic, Dutch, English, French, German, Italian, Portuguese, Spanish
If Multiple ICE or Leveraged matches exist for a given segment.
TM penalty profiles are created in Configuration->Data->Tags->TM Penalty Profiles.
TM matching reference
There are four basic mechanisms for providing translation matches against translation memory and auto-filling against the contents of the translation memory:
ICE (context) match
An ICE (In Context Exact) match is a translation that when matched against what's stored in TM, guarantees a high level of appropriateness because it has the same context
Important
In XTM an ICE match occurs when the source content in the current, previous and next segment is an exact match with the current, previous and next segment in TM (I.E the same location in a paragraph). ICE Matching is most relevant when source content is segmented by sentence as opposed to paragraph.
XTM Determines if an ICE or Leveraged Match exists using the following logic:
If the source text is identical to the one stored in TM, then XTM checks if a segment has a segment ID.
If the segment has a segment ID then XTM tries to find if an ICE (context) match exists by matching the source text segmentID with the one stored in TM for that string.
Tip
To include the segment ID in the TM please click here for instructions.
If the segment does not have a segment ID then XTM tries to find if an ICE (context) match exists by matching the source text alone.
When neither of the segmentID / context conditions is met, a leveraged match is returned.
When the translation is saved and the translation becomes approved for TM, XTM saves the translation in the TM and includes the segment ID (if activated) along with the source text, target text, and context (ICE match) by default.
Consider the segment We painted it red last week as used in the following paragraph:
My house used to be yellow.
We painted it red last week.
It looks very different now.
Is not a 100% ICE Match but is a Leveraged Match with the following:
My house used to be very ugly.
We painted it red last week.
It looks nicer now.
The original context was about a color scheme, but the context in the second paragraph is about beauty. Therefore the translation for the segment We painted it red last week may contain different adjectives to describe its beauty versus its.
Leveraged match
A leveraged match is when a sentence or phrase in a translation memory (TM) is the same phrase in a different context as the sentence or phrase the translator is currently working on.
Tip
Context is determined via ICE matching.
Fuzzy matches
Refers to the situation when a sentence or phrase in a translation memory (TM) is similar (but not a 100% ICE or Leveraged match) to the sentence or phrase the translator is currently working on. The TM tool calculates the degree of similarity or “fuzziness” as a percentage figure.
XTM Workbench (CAT) tool will show up to 5 fuzzy matches with the highest match rate for a given segment.
Note
This is configured under configuration->Settings->Translation->TM in the Fuzzy Matches section.
Fuzzy Match rates are calculated using one of the algorithms described here.
Fuzzy repeats
Refers to segments with no matches in the TM, but that are quite similar to each other within the project to be translated.
XTM Workbench (CAT) tool will show up to 5 fuzzy matches with the highest match rate for a given segment. This is configured under configuration->;Settings->;Translation->;TM in the Fuzzy Matches section.
TM clean up tools
XTM provides an interface for a TM Expert to perform the following tasks on any individual segment in the TM:
Warning
Only the TM Expert role has access to this feature.
Change the translation text (target).
Change the TM Status (Approved or Not Approved)
Delete a Segment.
This is performed in the TM->Manage window on a per language combination and customer level.
Tip
The advantage of making changes in this location is that XTM provides a total of 14 filters (search criteria) to narrow down the entries being verified.
Auto filling translations
Note
Auto filling with TM entries is done first if machine translation is enabled.
Auto filling occurs each time a project starts or is reanalyzed.
The configuration is done here in the following sections respectively:
ICE Matches
Leveraged Matches
Fuzzy Matches
Fuzzy Repeats
Translations can be configured to auto-fill automatically with a TM entry when any single one or any combination of the following occurs :
ICE match autofill options
ICE Matches where inline variables are the same in the segment and in the TM are auto-filled by default.
ICE matches where inline variables in the source segment and TM entry do not match can also be auto-filled. In this case, the inline variable in the source segment is copied to the translation.
Leveraged match autofill options
XTM can also be configured to auto-fill leveraged matches where inline variables in the source segment and TM entry do not match.
In this case, the inline variable in the source segment is copied to the translation.
Fuzzy match autofill options
95-99%
85-94%
75-84%
Fuzzy repeat autofill options
95-99%
85-94%
75-84%
TM approval mechanisms
Note
XTM automatically stores segments in TM as unapproved.
By default, XTM does not approve any translations for a segment in TM unless the following configurations are made:
Adding a TM Approve workflow step as part of the workflows.
Note
This is an automatic step type that automatically approves all of the segments translated during the current workflow in the TM. The step is closed automatically once the segments are approved.
Within any step in a workflow, there is a checkbox option to approve TM. This automatically approves the segment's translation in TM once it's completed.
Note
This is typically used in a correction or review workflow step.
Manually (no configuration required) regardless of the current state of a project. Simply edit a project using the
icon and select actions->Approve TM for this project. This approves all of the translations currently in the project for entry into the TM.
Rules for overwriting TM entries
Segments that have a 100% leveraged or ICE match translation can optionally be edited. The new TM entry will be created if the translation for such a match is edited.
Tip
By default, translations populated by ICE and Leveraged TM entries are locked for translation. To change this setting go to Configuration->Settings->Translation->TM. In the first section; Matches General look for the Mark segments as locked when checkbox, and uncheck it in order to enable edits to such translations.
Any combination of the following criteria can be set for editing 100% leveraged or ICE matched translations:
Modify the existing TM record if the project segment has the same:
Modify the existing TM record if the project segment and Segment ID have the same:
Source (i.e. the segment source text is the same as the TM Entry).
Inlines (i.e. the segment source text has the same inlines as the TM Entry).
Context (i.e. the segment source text has the same context as the TM Entry).
Tags (i.e. the segment source text has the same tags as the TM Entry).
Segment ID (i.e. the segment source text has the same segment ID as the TM Entry).
Tip
To change this settings go to Configuration->Settings->Translation->TM. In the first section; Matches General and change the Modify the existing TM record if the project segment .. entries.
The TM is live, so whatever happens anywhere inside or outside the project can optionally be leveraged even as a full match within XTM Workbench (CAT) for the linguists.
This is done by opening the project editor's General Info Tab and choosing the customers whose translation memory is to be leveraged, even if these customers are outside the scope of the project.