Automatically Find Equivalent Data in QC

[wsloand]

It would be very helpful to have a new feature in QC.

The way that I think this would work is that it would:

1. Look for data with both equal values and overlapping regions highlighted (any overlap-- maybe just "close" to be later defined).
2. It would then present a list of these data points to the QC user and ask them which they would prefer to keep (likely to be chosen based on field name).
3. If field names are identical in addition to the data value, then it will automatically be marked as matched.

After the initial matching and during the main part of QC:

1. Matched points are marked as QC complete at the same time (i.e. it matched point 1 is selected for QC complete, then matched point 2 is as well).
2. When the QC user clicks on a matched point one DE user, it goes to the group for the matched point in the other DE user. (To enable the QC user to find similar points from both users simultaneously.)

[roberttdev]

These extra steps sound like they will double the workload for a QC user to complete before marking the document complete. What did you want to do with this data.. just get a more accurate idea of how good each DE's work is? Do you predict the money you save from not compensating a DE user for unmatched points will cover all of the extra work for the QC, in order to not lose money on the endeavor?

Although, if QC's are compensated on points completed rather than time, and matched points aren't compensated, you technically wouldn't be paying them more.. but they would take longer to complete each document.

[wsloand]

I'm actually thinking that this will be a time saver for QC. The logic is that many documents will be much more complex than the current set that we're working with. In those documents, finding the same location for two DE users will be difficult.

Also, I think that it will increase the quality of the document if there is only one value that is a point. (It will prevent DE1 and DE2's copies of the same information from going in twice.)

The first pass will only be to find the automatic stuff that should go together, and it should be quick (<10% of the time for QC; probably <5%).

[roberttdev]

I'm confused.. if the highlights overlap, why is it hard to find both of them? Clicking on each highlight should reorient the UI to that point.. I think I'm missing something.

And currently, the QC doesn't do anything to acknowledge the "other" DE's points, other than the overall rating.. that's why I'm thinking the matching is extra work on his/her part.

[roberttdev]

From our discussion, it was decided:

It would be helpful to have the system try and do simple title/value matches of points with overlapping locations, in order to use the results as an extra flag to the QC as to abnormal data to pay special attention to.

At the end of DE, the system will do through each point, try to do a full title/value match (case-insensitive) for any overlapping points and store a match reference if one is found.

During QC, in the DE views, the row for each point will containing an icon indicating whether or not the point was successfully matched.

[wsloand]

Modification to the last paragraph:

During QC, in the DE views, the row for each point will containing an icon indicating whether or not the point was fully matched (title and value, best), matches value only, or value does not match (worst).