FastNLP中实现的模型包括:
这里提供的摘要任务数据集包括:
其中公开数据集(CNN/DailyMail, Newsroom, arXiv, PubMed)预处理之后的下载地址:
未公开数据集(NYT, NYT50, DUC)数据处理部分脚本放置于data文件夹
SummarizationLoader: 用于读取处理好的jsonl格式数据集,返回以下field
BertSumLoader:用于读取作为 BertSum(Liu 2019) 输入的数据集,返回以下 field:
Model | ROUGE-1 | ROUGE-2 | ROUGE-L | Paper |
---|---|---|---|---|
LEAD 3 | 40.11 | 17.64 | 36.32 | our data pre-process |
ORACLE | 55.24 | 31.14 | 50.96 | our data pre-process |
LSTM + Sequence Labeling | 40.72 | 18.27 | 36.98 | |
Transformer + Sequence Labeling | 40.86 | 18.38 | 37.18 | |
LSTM + Pointer Network | - | - | - | |
Transformer + Pointer Network | - | - | - | |
BERTSUM | 42.71 | 19.76 | 39.03 | Fine-tune BERT for Extractive Summarization |
LSTM+PN+BERT+RL | - | - | - |
Still in Progress...