Terminology matching
Pootle can help translators with terminology. Terminology can be specified to be global per language, and can be overridden per project for each language. A project called “terminology” (with any full name) can contain any files that will be used for terminology matching. Alternatively a file with the name “pootle-terminology.po” can be put in the directory of the project, in which case the global one (in the terminology project) will not be used. Matching is done in real time. Note that overriding the terminology project with a file called “pootle-terminology.po” does not work with GNU-style projects (where all the files are in one directory and have names according to the language code).
Ideally, the source term should be the shortest, simplest form of a word. Therefore cat, dog, house are good, but cats, dogged and housing are bad. Context indicators are allowed in the source text, in brackets after the term, but keep them short, eg “file (noun)”, “view (verb)”, etc.
The latest version of Pootle has the ability to insert the target term into the text when you click on it. The ideal is therefore that the target term be something that you'd like the translator to be able to insert… but strictly speaking the target text can be anything, including an definition.
If the terminology PO file has translator comments, it will be displayed as a tooltip in Pootle.
What does it do?
If our glossary has an entry: file→lêer, and we translate a sentence like “The file was not found”, we can suggest the glossary entry “file→lêer” as relevant to the translation, even if we don't have any TM entry that is related to the complete sentence that is available for translation.
Say our glossary has an entry “category→kategorie” and we translate a sentence like: “Please enter the categories for this photo”, we can suggest the glossary entry “category→kategorie”, even though the letters “category” doesn't occur anywhere in the original string.
Limits
Currently a single term entry can be up to 30 characters long (including context information), and the first 500 characters of each translation are scanned. Terms can consist of many words, but consider making them as general or simple as possible for maximum impact.
If these limits prove too restrictive, feel free to point out use cases where this is not sufficient.
