Adds the Krypton recognition engine confidence score to the XML result.
|
Value |
String. Allowed value: SWI_krconf |
|
Default |
(empty) |
|
How to set |
In Management Station, set on the Natural Language Processing service. If not using Management Station, set in the NLP service configuration file (User-nlpsxx.properties, see Configuration roadmap). |
|
Usage |
Set this parameter when your client application uses Dragon Voice confidence scores to make runtime dialog decisions. |
|
Performance |
Trivial impact on CPU and disk space. |
The returned confidence score is identical to the value of KRCONF in the call logs. See KRrslt—Krypton recognition result.
Setting this parameter does not guarantee that SWI_krconf is always returned:
-
For Krypton-only recognition situations, SWI_krconf is only returned when the Krypton nbest list has values that meet the confidence criteria.
-
For semantic interpretation situations, SWI_krconf is only returned when NLE intepretations match literal intepretations from the Krypton recognition engine.
Explanation for semantic intepretation:
- Krypton delivers n-best interpretations to the Natural Language Engine. The list include literal values and confidence scores.
- NLE creates its own n-best interpretations enhanced with its semantic abilities.
- The lists are compared.
- The literal of each NLE nbest item is compared with the literal of every item in the Krypton nbest list.
- When the NLE literal matches the Krypton literal, the value of KRCONF (from the Krypton nbest item) is copied to SWI_krconf (for the matching NLE nbest item) in the results.
The matching literal interpretations can belong to different nbest items. In this example, the first nbest from NLE ("forecast") gets populated with the KRCONF with the second nbest from Krypton:
| nbest item | NLE literal | Krypton literal |
|---|---|---|
| 1 | forecast | fork rafts |
| 2 | floor crafts | forecast |
| 3 | force crafts | flour shaft |
