Problem

The AutoQA error count is misleading for tasks where source text is editable. Current behaviour:

• "Forbidden in source text" is highlighted in pink, but is not counted if the target segment is empty
• If the target segment is populated but does not contain the target-language equivalent of the forbidden term, then it is not counted
• If the target-language equivalent (assumption: only if the highest available usage status is used) of the forbidden term is inserted into the target segment, the counter increases by 2 (assumption: once in the source column and once in the target column?)
• When the source term is corrected, the counter is decremented by 1 (although it was originally incremented twice)

Solution

• It is ok to match the pink highlights from the start, but AutoQA error should be counted only once per segment
• If the forbidden term in the source text is corrected, the counter should decrease by 1; this is already the case, but currently every error is counted twice whilst the counter only decreases by 1
• If the translation uses something completely different because it has nothing to do with the context of the forbidden term: The counter should then decrease by 1and, if possible with small effort, the pink highlighting should turn grey
• However, it should still be possible to filter the “Forbidden in source” category using “Show all” or “Show ignored errors” and find all segments with forbidden terms in orginal source text.