MIT Press

Skip to Main Content

Table 7:

Accuracy of single-unit and full-hidden-state word-boundary diagnostic classifiers, trained and tested on balanced data requiring new-word generalization. Chance accuracy is at 50%.

	LSTM	LSTM	RNN	RNN
	single	full	single	full
English	77.5	90.0	65.9	76.8
German	80.8	79.7	67.0	75.8
Italian	75.5	82.9	71.4	75.9

	LSTM	LSTM	RNN	RNN
	single	full	single	full
English	77.5	90.0	65.9	76.8
German	80.8	79.7	67.0	75.8
Italian	75.5	82.9	71.4	75.9