By default, alignment tools use punctuation as the main criterion for segmenting texts and aligning translation units.
Some systems also exploit differences in the length of two potentially matching segments; and if these differences are big, they try to make use of additional criteria.
In some cases, it is possible to "tune" the alignment tool by using such additional criteria, for example, to let the tool itself assign 1:2 or 2:1 relations between source and target segments. This is done by defining criteria which increase the probability that one segment in one language corresponds to two segments in the other language.
For example, if a document contains quite a lot of numbers, and if in one language certain numbers occur in one segment and in the other language the same numbers are distributed over two segments, this can increase the likelihood that this is a case of a 1:2 relation.
In this example the significance of numbers and some other criteria for binding together potential segments as translation units can be increased or decreased by just moving the sliders to the right or left.
Further possibilities for improving the performance of the alignment tool can include Structure Recognition in source and target files.
In this case, the alignment tool makes use of document formatting, such as styles specified in style sheets, paragraph numbering or even font sizes. This means that if two potentially matching segments share the same structural formatting, this increases the likelihood that they can be regarded as corresponding translation units.