You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
屠明 df6512794b cleaning 4 years ago
..
tests initial upload 4 years ago
__init__.py initial upload 4 years ago
__main__.py initial upload 4 years ago
configuration_auto.py initial upload 4 years ago
configuration_bert.py initial upload 4 years ago
configuration_distilbert.py initial upload 4 years ago
configuration_gpt2.py initial upload 4 years ago
configuration_openai.py initial upload 4 years ago
configuration_roberta.py initial upload 4 years ago
configuration_transfo_xl.py initial upload 4 years ago
configuration_utils.py initial upload 4 years ago
configuration_xlm.py initial upload 4 years ago
configuration_xlnet.py initial upload 4 years ago
convert_gpt2_checkpoint_to_pytorch.py initial upload 4 years ago
convert_openai_checkpoint_to_pytorch.py initial upload 4 years ago
convert_pytorch_checkpoint_to_tf.py initial upload 4 years ago
convert_roberta_checkpoint_to_pytorch.py initial upload 4 years ago
convert_tf_checkpoint_to_pytorch.py initial upload 4 years ago
convert_transfo_xl_checkpoint_to_pytorch.py initial upload 4 years ago
convert_xlm_checkpoint_to_pytorch.py initial upload 4 years ago
convert_xlnet_checkpoint_to_pytorch.py initial upload 4 years ago
file_utils.py initial upload 4 years ago
modeling_auto.py initial upload 4 years ago
modeling_bert.py initial upload 4 years ago
modeling_distilbert.py initial upload 4 years ago
modeling_gpt2.py initial upload 4 years ago
modeling_openai.py initial upload 4 years ago
modeling_roberta.py initial upload 4 years ago
modeling_transfo_xl.py initial upload 4 years ago
modeling_transfo_xl_utilities.py initial upload 4 years ago
modeling_utils.py initial upload 4 years ago
modeling_xlm.py initial upload 4 years ago
modeling_xlnet.py initial upload 4 years ago
optimization.py initial upload 4 years ago
sentGNN_DGL_BiDAF.py initial upload 4 years ago
tokenization_auto.py initial upload 4 years ago
tokenization_bert.py initial upload 4 years ago
tokenization_distilbert.py initial upload 4 years ago
tokenization_gpt2.py initial upload 4 years ago
tokenization_openai.py initial upload 4 years ago
tokenization_roberta.py initial upload 4 years ago
tokenization_transfo_xl.py initial upload 4 years ago
tokenization_utils.py initial upload 4 years ago
tokenization_xlm.py initial upload 4 years ago
tokenization_xlnet.py initial upload 4 years ago

该项目开源了一种可解释的、高效的多文档阅读理解算法SAE(Select, Answer and Explain)。该算法通过选择、回答和解释来对多个信息源进行推理,并通过提供支持证据来解释答案预测。具体地,SAE算法首先过滤掉与答案无关的文件, 从而减少干扰信息的数量。然后将选定与答案相关的文档输入到一个模型中, 预测答案和提供支持的句子。 该模型在答案预测的单词级别和支持句子预测的句子级别同时优化了多任务学习目标, 并通过注意力机制实现了这两个任务的交互。

Python Text