The auto-assembling function in CafeTran can produce the translation of a segment based on the matches in the translation memories and glossaries. The whole process is automatic and the results vary depending on the factors such as the size and the quality of translation memories, a language pair or the subject of the document being translated. To improve the quality of the matches, CafeTran enables the translator to have some control over the matching process by creating segment patterns which can be used in the auto-assembling process. The segment pattern has the following exemplary form:

• All the leaves are {1} and the sky is {2}. - a source segment in the translation memory.
• Todas las hojas son color {1} y el cielo es {2}. - a target segment in the translation memory.

Then CafeTran will be able to use terms in glossaries or fragments in translation memories to replace the variable parts with the found entry, creating a complete translation automatically. In our source segment example, the leaves and the sky can have any colors changing them in various contexts. Still, CafeTran will provide the correct translation if the separate matches for the colors exist either in a glossary or a translation memory. Note that the order of variable parts can be switched so you might create the following patterns as well:

• All the leaves are {1} and the sky is {2}.
• El cielo es {2} y todas las hojas son color {1}.

The feature also lets the user set the default exact match for the pattern such as:

• All the leaves are {1=brown} and the sky is {2=gray}.
• Todas las hojas son color {1=café} y el cielo es {2=gris}.

so when the default segment is found in the source document, CafeTran will use the default translation for it.

Segment patterns are added to translation memories via the "New fragment" dialog. If the current source segment matches the pattern, the translation is inserted in the target segment pane and displayed in the auto-assembling panel (F1) as a 100% match.