Skip to Main Content
Table 6: 

Top 10 SWDA dialog acts that benefit from punctuation and truecasing availability in pure dialog act segmentation. The columns denoted by (l) and (nl) represent numbers for dialog-level context XLNet lower and nolower models, respectively.

Mis-segmented dialog actsCountDSER (lc) [%]DSER (nlc) [%]Abs. gain [%]
Rhetorical-Questions 12 58.3 16.7 −41.7 
Uninterpretable 366 42.9 6.8 −36.1 
Hedge 23 39.1 4.3 −34.8 
Quotation 18 66.7 44.4 −22.2 
Other 15 40.0 20.0 −20.0 
Statement-non-opinion 1494 30.7 13.7 −17.0 
Agree-Accept 213 22.1 7.0 −15.0 
Statement-opinion 832 29.7 15.9 −13.8 
Declarative-Yes-No-Question 38 18.4 5.3 −13.2 
Open-Question 16 12.5 0.0 −12.5 
Mis-segmented dialog actsCountDSER (lc) [%]DSER (nlc) [%]Abs. gain [%]
Rhetorical-Questions 12 58.3 16.7 −41.7 
Uninterpretable 366 42.9 6.8 −36.1 
Hedge 23 39.1 4.3 −34.8 
Quotation 18 66.7 44.4 −22.2 
Other 15 40.0 20.0 −20.0 
Statement-non-opinion 1494 30.7 13.7 −17.0 
Agree-Accept 213 22.1 7.0 −15.0 
Statement-opinion 832 29.7 15.9 −13.8 
Declarative-Yes-No-Question 38 18.4 5.3 −13.2 
Open-Question 16 12.5 0.0 −12.5 
Close Modal

or Create an Account

Close Modal
Close Modal