Direct Output Connection for a High-Rank Language Model

Benchmark Model Rank Results
constituency-parsing-on-penn-treebankLSTM Encoder-Decoder + LSTM-LM#15F1 score: 94.47
language-modelling-on-penn-treebank-wordAWD-LSTM-DOC x5#8Test perplexity: 47.17Validation perplexity: 48.63Params: 185M
language-modelling-on-penn-treebank-wordAWD-LSTM-DOC#15Test perplexity: 52.38Validation perplexity: 54.12Params: 23M
language-modelling-on-wikitext-2AWD-LSTM-DOC x5#19Test perplexity: 53.09Validation perplexity: 54.19
language-modelling-on-wikitext-2AWD-LSTM-DOC#23Test perplexity: 58.03Validation perplexity: 60.29