Documentation Center

Translation memory penalties

A translation memory penalty is a number that indicates the loss of reliability in a translation match.

A translation match is a match between a source segment in a document and a source segment in the translation memory.

SDL Trados Studio displays matching segments in the Translation Results window. It also displays a figure indicating the reliability of the match in the segment status column of the Editor window.

If the source document segment is identical to the source segment in the translation memory, the match is exact and the match reliability figure is 100%.

However, you can specify penalties that apply to a match when there are slight differences. For example, you can specify a penalty that applies if the formatting differs between the translation memory source and the document source.

You specify these penalties in the Language Pairs settings: Translation Memory and Automated Translation > Penalties.

You can also specify other penalties as follows:

  • You can specify a penalty that applies to all translation from a specific translation memory. See below, Provider penalty.
  • You can apply a filter to specify a penalty. See below, Filter penalty.

When a penalty is applied, the Editor window display changes and:

  • The match in the segment status column is shown as a percentage: 100% minus the penalty.
  • The penalty is indicated by an icon beside the match percentage.
  • A description of the penalty is displayed when you hold the mouse over the penalty icon.

Penalties are applied before results are displayed. Penalties can reduce a match percentage so much that the match is not displayed.

PenaltyIconWhy the penalty is applied

Missing formatting

One of the source segments (either the translation memory or the document source) has formatting that is not in the other source segment.

Different formatting

The translation memory and the document source have different formatting.

Multiple translations

There is more than one exact match for the document source segment.

Auto-localization

SDL Trados Studio recognizes that the document source text is the same pattern as a date, time, number or measurement match in the translation memory but with different values.

Auto-substitution and Auto-localization of recognized tokens (dates, times, numbers or measurements) occurs when the matching source segment in the translation memory contains a recognized token that is different to that in the source segment in the document. When the translation memory is applied, Studio automatically substitutes the recognized token from the source segment in the document to the translation in the target segment. At the same time it applies the correct target language format for that element and applies the auto-localization penalty.

This penalty is only applied during auto-substitution. Specify if dates, times, numbers and measurements are auto-substituted on the Auto-Substitution page. Specify auto-localization patterns on the Dates and Times and Measurements page.

Text Replacement

SDL Trados Studio recognizes that the document source text has a match in the translation memory, but with a different variable, acronym or alphanumeric string.

SDL Trados Studio uses the translation memory segment to supply the translation for the document segment and applies the text replacement penalty.

You can define your variables in the Variables dialog box in the translation memory settings. This penalty is only applied during auto-substitution. Specify if variables, acronyms or alphanumeric strings are auto-substituted on the Auto-Substitution page.

Filter

A filter penalty has been added. For example, you can specify that translations older than two years have a penalty.

One match can have more than one filter penalty. The penalty information includes the name of the filter.

Specify filter penalties in Translation Memory and Automated Translation > Filters.

Provider

The translation memory that found this match has this penalty applied to all its matches.

Provider penalties allow you to rate some translation memories as more trustworthy than other translation memories.