Problem

In TermCollections it is recommended from a terminology point of view to have a full form of a term and the short form within the same entry, as definitions, context, pictures etc. will be the same for both. However, when such an entry is used for DeepL glossary (or for LLM), only one term will be written into the glossary we create as we have rules on how to process synonyms. So either the full form and short form must be kept in separate entries or the glossary could pair wrong forms (short with full e.g.) and not all available information will go to DeepL.

Currently

We have terms collection like:

1 entry with 2 term in it:

English:
ML - admitted term (short)
Machine Learning - preferable term (full)

German:
ML - admitted term (short)
Maschinelles Lernen - preferable term (full)

And it will result in only 1 pair for DeepL or LLM:

Machine Learning - Maschinelles Lernen

That is because these terms have best usage type (administrative status), short terms are skipped here even though they are "good enough" to be used.

And in worse scenario both forms could have same administrative status. Then pair will be picked randomly from user point of view.

Of course then pairs may be picked wrong.

Solution

Introduce a new rule for the glossary creation. When the attribute "term type" is set and matches for source and target language write this pair to the glossary. 

So matching source-target pairs with set attribute will have higher prio then the current synonym logic. User must be aware of this when they use attribute "term type" and this must be stated in the important release notes.

We wanna have more pairs in glossaries that are passed to third-party tools in case if terms have "term type" attribute.

Rules to find new term pairs within one entry:

1. Terms without "term type" paired as before depending on administrative status.
2. Terms with deprecated or superseded administrative status excluded from vocabulary.
3. We have 1 term pair with short type and 1 with full type -> send them both.
4. We have short|full term type in source and some terms of same type in target:
   1. We can find 1 best target term based on administrative status -> use it as pair
   2. We can't find 1 term based on administrative status -> use one of best{}
5. We have short|full term type in source and no same type in target -> use term depending on best administrative status.
6. Entry has 2 terms with "term type" of same source language and only 1 in target:
   1. target has "term type" -> only 1 pair with appropriate term type formed 
   2. target doesn't have "term type" -> make pair per each source term
7. Entry has 2 terms with "term type" of same source language and some in target without "term type" -> make pair per each source term with best target.

Examples:

Same rules should be applied for finding best term in TermCollection pre-translation.