Since in ONTRAM text is segmented paragraph by paragraph and the texts can come from different sources, e.g. from brochures or web content, it can happen that texts are segmented differently in the translation.
In ONTRAM parts of a text are recognized as 100% partial memory. If a large segment is entered into the translation, translated and then exported, the text is recognized as partial memory if at least 25% of the text matches and the matching text is at least 40 characters long. The percentages for fuzzy TM matches are not relevant for this function, since it is a 100% partial match of a larger segment.
Example: If the memory contains a translation for a segment with three sentences and a new segment with one of these sentences is requested, no match would be returned for this segment without the function Partial memory. By recognizing the high match of the sentence, the existing segment from the memory is now linked even if the quality of the overall hit does not meet the minimum quality.
The Detect partial memory feature is enabled by default in the group settings in the Translation Memory (TM) area.