Some clarifications to the instructions were offered:
When Customizing Attributes, the wpos and feats attributes do not need to be displayed.
It saves time to do simultaneous consistency checking on the WordNet and Mikrokosmos concepts, in which case both should be set to display.
When annotations have been inconsistent, reference to Omega may be needed to decide which previous annotation to follow.
It isn't always clear what should count as carrying the same meaning.
On the one hand, the annotator can usually tell when a pair of words in the two translations stem from the same word/phrase in the original text and therefore should carry the same meaning.
On the other hand, it isn't necessarily the case that the two translations do convey the same meaning.
It would be handy to have:
Interleaved texts; this saves a lot of time, since changes should only be made to one copy of the text, even if two copies are being viewed.
A printout of the annotations that shows:
The original text
Words that were annotated, with both WN and Mk concepts selected
A modification to Tiamat that would allow annotators to see what they had chosen for a previous occurrence of a word (or perhaps asking whether the annotator would like to use the same set of concepts previously selected; in any case, they felt the annotator should have to make an explicit choice).
The list of word re-occurrences and WordNet synonyms was found helpful by one annotator, not so very helpful by another.
It takes 1.5-2 hours to do consistency checking on one text, pretty much evenly split between the pencil-and-paper preparation stage and the consistency-instilling stage in TrEd.
Most of the annotators made an effort to make consistent choices on word re-occurrences in the original task; most changes at this stage are coming from synonyms.