Table 1:

Model . | Input . | Sequence Encoder . | Output . |
---|---|---|---|

ELMo | CNN | LSTM | Sampled Softmax |

ELMo-C (ours) | FastText_{cc} | LSTM w/ LN | Cont w/ FastText_{cc} |

ELMo-A | FastText_{cc} | LSTM w/ LN | Adaptive Softmax |

ELMo-Sub | Subword | LSTM w/ LN | Softmax |

ELMo-C_{OneB} | FastText_{OneB} | LSTM w/ LN | Cont w/ FastText_{OneB} |

ELMo-C_{Rnd} | FastText_{cc} | LSTM w/ LN | Cont w/ Random Embedding |

ELMo-C_{CNN} | Trained CNN | LSTM w/ LN | Cont w/ Trained CNN |

ELMo-C_{CNN-CC} | Trained CNN | LSTM w/ LN | Cont w/ FastText_{cc} |

ELMo-C_{CC-CNN} | FastText_{cc} | LSTM w/ LN | Cont w/ Trained CNN |

Model . | Input . | Sequence Encoder . | Output . |
---|---|---|---|

ELMo | CNN | LSTM | Sampled Softmax |

ELMo-C (ours) | FastText_{cc} | LSTM w/ LN | Cont w/ FastText_{cc} |

ELMo-A | FastText_{cc} | LSTM w/ LN | Adaptive Softmax |

ELMo-Sub | Subword | LSTM w/ LN | Softmax |

ELMo-C_{OneB} | FastText_{OneB} | LSTM w/ LN | Cont w/ FastText_{OneB} |

ELMo-C_{Rnd} | FastText_{cc} | LSTM w/ LN | Cont w/ Random Embedding |

ELMo-C_{CNN} | Trained CNN | LSTM w/ LN | Cont w/ Trained CNN |

ELMo-C_{CNN-CC} | Trained CNN | LSTM w/ LN | Cont w/ FastText_{cc} |

ELMo-C_{CC-CNN} | FastText_{cc} | LSTM w/ LN | Cont w/ Trained CNN |

This site uses cookies. By continuing to use our website, you are agreeing to our privacy policy.