DocView Capture
DocView Capture measures how much work people put into field values after OCR, so you can see which operators, steps, Batch Setups and fields take the most keying, and how good OCR is for each field.
What is recorded
For every field value, DocView keeps what OCR or the extraction read. Each time a person changes the value, in any step (scan, quality check, classification, correction, indexing or verification), one effort record is kept with:
- the step the batch was in, the person and the time;
- the OCR value, the value before the change and the value after it;
- their lengths in characters;
- the edits the change took: the Levenshtein distance, the fewest single characters added, removed or replaced to turn the old value into the new one (correcting
INV-0O12toINV-0012is 1 edit; typingACME Ltdinto an empty field is 8); - how far the value now is from what OCR read;
- OCR's confidence in its value.
Values set by DocView Workers, the REST API or the system are not effort.
Reports
Under Reports:
- Operator effort on field values (People): per person and step, the batches, changes, fields changed, characters keyed, edits, edits per change and the OCR accuracy of the values they worked on.
- Effort per field (OCR against people) (Quality): per Batch Setup and field, how often people changed what OCR read, in which steps, the edits needed and the OCR accuracy. Fields at the top are the ones to improve (a better finder, zone, format or dictionary).
OCR accuracy is 100 % minus the edits from OCR to the final values over their length: 100 % means people never had to change what OCR read. Choose any period, Batch Setups, step and operator. You see only the Batch Setups you may see.
How long it is kept
The values themselves (what OCR read, and each value before and after a change) are kept only while their batch exists. When a batch is deleted for good, by hand or when an archived batch's time is up, its effort records keep only the statistics: step, person, field, time, lengths and edits. The reports need nothing more.
The statistics are kept for Settings › Data retention › Effort statistics (default 365 days; 0 keeps them). The retention job deletes older ones about every 10 minutes while a DocView Worker runs. Archived batches are deleted after 7 days by default (Delete archived batches after).
A separate effort database
By default effort records are kept in the DocView Capture database, saved together with the change. To keep them apart (for example on a reporting server), give an empty SQL Server database in the connection string DocViewEffort of the web app and of every DocView Worker:
"ConnectionStrings": { "DocViewEffort": "Server=reports01;Database=DocViewEffort;Integrated Security=True;TrustServerCertificate=True" }
The table is made the first time it is used. Records are then written to it in the background a few seconds after each change, so work never waits for it; a problem with that database is written to the log and does not stop any work.