Problem

Currently we use format "Segment Nr: #" as context of segment update in t5memory when no descriptor is set for use.

Descriptor may be resname field in xliff files as example.

We also decided to use "-" as context in case if system setting "runtimeOptions.LanguageResources.t5memory.skipContext" set to true.

As such for some cases "Segment Nr: #" looks like good idea as it provides additional chances to get 103 match in case of work on same document.

In reality it leads to cases when t5memory has segment with context that represents state of segment in specific task from date yyyy-mm-01
Then when client translates similar package with same document in it and produces some fixes for same segment in document that is now placed at different position in task in t5memory will be 2nd segment with date yyyy-mm-02
And then by chance client translates yet another package with same document that is this time result in segment placed at same position in task as it was the first time.
As this is still 100 match from t5memory and same document and same context (number in task is same) as it was in 1st translation this segment will be pre-translated with match from date yyyy-mm-01 as it will get 103 match rate. And 2nd match will not be used simply because of context (Segment Nr)

Solution

Do not use Segment Nr. in task anymore as context. Use "-" as context for cases when no descriptor is set for segment.

Incorporate this change in TM filtering that is done on content conversion, import and reorganize of memories.