Skip to Main Content
Table 5 

BLEU score of the word-based utterance class LMs for different n-gram sizes and different number of slots included in the utterance class (most frequent first). Best performing parameters are in bold. The BLEU score is averaged over all cross-validation folds. See figures 12 and 13 for results using other parameter configurations.

System configurationn-gram sizeBLEU no slotBLEU 1 slotBLEU 2 slots
O&R deterministic .25 .06 .05 
O&R deterministic .28 .02 .01 
O&R deterministic .25 .01 .00 
O&R sampling .25 .03 .03 
O&R sampling .27 .02 .01 
O&R sampling .23 .01 .00 
Bagel n/a  .37  
System configurationn-gram sizeBLEU no slotBLEU 1 slotBLEU 2 slots
O&R deterministic .25 .06 .05 
O&R deterministic .28 .02 .01 
O&R deterministic .25 .01 .00 
O&R sampling .25 .03 .03 
O&R sampling .27 .02 .01 
O&R sampling .23 .01 .00 
Bagel n/a  .37  
Close Modal

or Create an Account

Close Modal
Close Modal