Depending on the material to be interpreted, using a specific add-on interpretation engine— either alone or in combination with FORMS’ internal engine— may provide better recognition.
When you use an external interpretation engine in combination with FORMS’ own engine, the two engines give you somewhat different results. You can let the engines "vote" on the results.
Test the voting options if you are dissatisfied with the interpretation of any single interpretation engine. This might be the case if forms have poor image quality, lots of handwritten text, or material in which the accuracy of interpretation is critical.
You control how differing interpretation results are handled by the system. This lets you:
Take advantage of the fact that different interpreters usually make different mistakes. You use each engine’s advantages and avoid its disadvantages.
Make adjustments based on how important a perfect interpretation result is for each specific field.
When you use an external interpretation engine in combination with FORMS’ own engine, these voting options are available:
|
Select best match |
If the two engines submit the same first alternative, the interpretation status is set to Certain. If they differ, and if the confidence level of one of them is high enough, the alternative with highest confidence selected. This choice results in the highest rate of interpretation. |
|
Security voting |
If the results from the two engines differ, the character is set to Uninterpretable and is taken up in Verify. This option has the lowest substitution rate. |
|
Optimized voting |
A result is chosen based on knowledge of the typical advantages and problems with the engines involved. The engines from Open Text and A2iA have been analyzed this way, so there are routines in place for optimized voting between these engines and FORMS. With respect to interpretation vs. low substitution rate, this choice lies somewhere between Select best match and Security voting. This option is only available when you use Open Text or A2iA interpretation engines. |