Greetings from the east coast of Canada. I've been looking for information on how to improve "match quality" and haven't found much. What I'm referring to is the number out of "1" that you see in the log as the software attempts to find the most suitable layout. I have certain one's that always return close to 1, something like 0.97 but others are much lower like 0.27. Results are always strong however with generally 98% or better confidence level once matched and recognized. I have seen an improvement after making a change to the "master columns to detect rows" on one occasion, but other than that I've had no luck. Is this something I need to be concerned about since I'm getting strong results? Are there standard settings one should check or attempt to improve that will have a more significant impact?
Improving "Match Quality" [FlexiCap 11 Dist]
Was this article helpful?
0 out of 0 found this helpful
Comments
2 comments
Recognition result(Confedence level) is calculated after all text fields are recognized and rules evaluated.Its based on percent of nonsuspicious symbols and succeeded rules.
Matching quality is a diffrent parameter, it is the value of resulting hypothesis. This value depends on which elements were found or not and their null hypothesis value.
For example if you have two elements with null hyptohesis quality equal to 0.97 if they are both not found the resulting hypothesis would be 0.97*0.97.
So if you need to improve matching quality:
1.Improve image quality
2.Adjust Flexilayout project- check which elements reduce your resulting hypothesis
Regards
Timur
Post is closed for comments.