Intelligent Test Algorithm for English Writing Using English Semantic and Neural Networks

Table 3

Accuracy of part of speech tagging in WSJ corpus with different network structures.

Model	POS
	Dev	Test
	Acc. (%)	Acc. (%)
BLSTM + Residual	96.12	96.21
BLSTM + Residual + 2Cost	96.46	96.48
BLSTM + Residual + 2Cost + BN	96.68	96.73,
BLSTM + CRNN + Residual + 2Cost + BN	97.53	97.60