Skip to Main Content
Table 2: 

Summary of model AUC scores in historical prediction of slang emergence (1960s-2000s). The non-contrastive SBERT baseline and the proposed full model (with contrastive embedding, CSE, and syntactic prior, SSP) are compared using collaborative-filtering Prototype. Models were trained and tested incrementally through time (test sizes shown) and trained initially on 20,899 Green’s Dictionary definitions prior to the 1960s.

Decade# TestBaselineSBERT+CSE+SSP
1960s 2010 67.5 77.4 
1970s 1757 66.3 77.9 
1980s 1655 66.3 78.6 
1990s 1605 66.2 75.4 
2000s 1374 65.9 77.0 
Decade# TestBaselineSBERT+CSE+SSP
1960s 2010 67.5 77.4 
1970s 1757 66.3 77.9 
1980s 1655 66.3 78.6 
1990s 1605 66.2 75.4 
2000s 1374 65.9 77.0 
Close Modal

or Create an Account

Close Modal
Close Modal