Skip to Main Content
Table 6 
Fully supervised experimental results on the GeoQuery data set. For Jia and Liang (2016), we include two of their results: one is a standard neural sequence to sequence model; and the other is the same model trained with a data augmentation algorithm on the labeled data (reported in parentheses).
ModelsAccuracy
Zettlemoyer and Collins (2005) 79.3 
Zettlemoyer and Collins (2007) 86.1 
Kwiatkowksi et al. (2010) 87.9 
Kwiatkowski et al. (2011) 88.6 
Kwiatkowski et al. (2013) 88.0 
Zhao and Huang (2015) 88.9 
Liang, Jordan, and Klein (2011) 91.1 
  
Dong and Lapata (2016) 84.6 
Jia and Liang (2016) 85.0 (89.1) 
Rabinovich, Stern, and Klein (2017) 87.1 
  
TNSP, soft attention, top–down 86.8 
TNSP, soft structured attention, top–down 87.1 
TNSP, hard attention, top–down 85.3 
TNSP, binomial hard attention, top–down 85.5 
TNSP, soft attention, bottom–up 86.1 
TNSP, soft structured attention, bottom–up 86.8 
TNSP, hard attention, bottom–up 85.3 
TNSP, binomial hard attention, bottom–up 85.3 
ModelsAccuracy
Zettlemoyer and Collins (2005) 79.3 
Zettlemoyer and Collins (2007) 86.1 
Kwiatkowksi et al. (2010) 87.9 
Kwiatkowski et al. (2011) 88.6 
Kwiatkowski et al. (2013) 88.0 
Zhao and Huang (2015) 88.9 
Liang, Jordan, and Klein (2011) 91.1 
  
Dong and Lapata (2016) 84.6 
Jia and Liang (2016) 85.0 (89.1) 
Rabinovich, Stern, and Klein (2017) 87.1 
  
TNSP, soft attention, top–down 86.8 
TNSP, soft structured attention, top–down 87.1 
TNSP, hard attention, top–down 85.3 
TNSP, binomial hard attention, top–down 85.5 
TNSP, soft attention, bottom–up 86.1 
TNSP, soft structured attention, bottom–up 86.8 
TNSP, hard attention, bottom–up 85.3 
TNSP, binomial hard attention, bottom–up 85.3 
Close Modal

or Create an Account

Close Modal
Close Modal