Skip to Main Content
Table 3 
Results for models of different context representations: Char(acter), W(ord)P(iece), W(ord)F(eature). For the contextual models we specify the tokenization level of the sentential context. *: difference is statistically significant (p < .05 on the McNemar test) in comparison with the sliding window baseline. For speed, e.g., “1.4x” means 1.4 times faster.
 Accuracy (Error)Speed
AllSemiotic classSentence
English (Standard): 
 Sliding window 99.79% (0.21%) 98.20% (1.80%) 97.99% (2.01%) 1.0x 
 Context (Char) 99.79% (0.21%) 98.20% (1.80%) 97.87% (2.13%) 1.3x 
 Context (WP) 99.79% (0.21%) 98.35% (1.65%) 97.76% (2.24%) 1.4x 
 Context (WF) 99.84% (0.16%)98.30% (1.70%) 98.20% (1.80%) 1.4x 
  
Russian (Standard): 
 Sliding window 99.64% (0.36%) 97.31% (2.69%) 95.61% (4.39%) 1.0x 
 Context (Char) 99.65% (0.35%) 97.26% (2.74%) 95.70% (4.30%) 1.8x 
 Context (WP) 99.61% (0.39%) 96.94% (3.06%)* 95.26% (4.74%) 1.8x 
 Context (WF) 99.62% (0.38%) 97.01% (2.99%)* 95.46% (4.54%) 2.0x 
 Accuracy (Error)Speed
AllSemiotic classSentence
English (Standard): 
 Sliding window 99.79% (0.21%) 98.20% (1.80%) 97.99% (2.01%) 1.0x 
 Context (Char) 99.79% (0.21%) 98.20% (1.80%) 97.87% (2.13%) 1.3x 
 Context (WP) 99.79% (0.21%) 98.35% (1.65%) 97.76% (2.24%) 1.4x 
 Context (WF) 99.84% (0.16%)98.30% (1.70%) 98.20% (1.80%) 1.4x 
  
Russian (Standard): 
 Sliding window 99.64% (0.36%) 97.31% (2.69%) 95.61% (4.39%) 1.0x 
 Context (Char) 99.65% (0.35%) 97.26% (2.74%) 95.70% (4.30%) 1.8x 
 Context (WP) 99.61% (0.39%) 96.94% (3.06%)* 95.26% (4.74%) 1.8x 
 Context (WF) 99.62% (0.38%) 97.01% (2.99%)* 95.46% (4.54%) 2.0x 
Close Modal

or Create an Account

Close Modal
Close Modal