MIT Press

Skip to Main Content

Table 2:

Accuracy of diagnostic classifier on predicting word class, with standard errors across 100 random train-test splits. ‘subs.’ marks in-vocabulary subset evaluation, not comparable with the other results.

	German	Italian
Random	50.0	50.0
Autoencoder	65.1 (± 0.22)	82.8 (± 0.26)
LSTM	89.0 (± 0.14)	95.0 (± 0.10)
RNN	82.0 (± 0.64)	91.9 (± 0.24)
WordNLM	53.5 (± 0.18)	62.5 (± 0.26)
WordNLM_subs.	97.4 (± 0.05)	96.0 (± 0.06)

	German	Italian
Random	50.0	50.0
Autoencoder	65.1 (± 0.22)	82.8 (± 0.26)
LSTM	89.0 (± 0.14)	95.0 (± 0.10)
RNN	82.0 (± 0.64)	91.9 (± 0.24)
WordNLM	53.5 (± 0.18)	62.5 (± 0.26)
WordNLM_subs.	97.4 (± 0.05)	96.0 (± 0.06)