@@ -95,13 +95,16 @@ For more pretrained model selection, please refer to [PretrainedModels](./paddle
...
@@ -95,13 +95,16 @@ For more pretrained model selection, please refer to [PretrainedModels](./paddle
-[Models API](./docs/models.md)
-[Models API](./docs/models.md)
## Tutorials
## Tutorials
Please refer to our official AI Studio account for more interactive tutorials: [PaddleNLP on AI Studio](https://aistudio.baidu.com/aistudio/personalcenter/thirdview/574995)
Please refer to our official AI Studio account for more interactive tutorials: [PaddleNLP on AI Studio](https://aistudio.baidu.com/aistudio/personalcenter/thirdview/574995)
*[What's Seq2Vec?](https://aistudio.baidu.com/aistudio/projectdetail/1294333) shows how to use LSTM to do sentiment analysis.
*[What's Seq2Vec?](https://aistudio.baidu.com/aistudio/projectdetail/1283423) shows how to use LSTM to do sentiment analysis.
*[Sentiment Analysis with ERNIE](https://aistudio.baidu.com/aistudio/projectdetail/1283423) shows how to exploit the pretrained ERNIE to make sentiment analysis better.
*[Sentiment Analysis with ERNIE](https://aistudio.baidu.com/aistudio/projectdetail/1294333) shows how to exploit the pretrained ERNIE to make sentiment analysis better.
*[Waybill Information Extraction with BiGRU-CRF Model](https://aistudio.baidu.com/aistudio/projectdetail/1317771) shows how to make use of bigru and crf to do information extraction.
*[Waybill Information Extraction with BiGRU-CRF Model](https://aistudio.baidu.com/aistudio/projectdetail/1317771) shows how to make use of bigru and crf to do information extraction.
| Electra 预训练&GLUE下游任务 | [Electra(Pre-training Text Encoders as Discriminators Rather Than Generator)](https://github.com/PaddlePaddle/models/tree/develop/PaddleNLP/examples/electra) |ELECTRA模型新一种模型预训练的框架,采用generator和discriminator的结合方式,相对于BERT来说能提升计算效率,同时缓解BERT训练和预测不一致的问题。|
| Electra 预训练&GLUE下游任务 | [Electra(Efficiently Learning an Encoder that Classifies Token Replacements Accurately)](./language_model/electra) |ELECTRA 创新性地引入GAN的思想对BERT预训练过程进行了改进,在和BERT具有相同的模型参数、预训练计算量一样的情况下,ELECTRA GLUE得分明显好。同时相比GPT、ELMo,在GLUE得分略好时,ELECTRA预训练模型只需要很少的参数和计算量。|
| [ERNIE-GEN(An Enhanced Multi-Flow Pre-training and Fine-tuning Framework for Natural Language Generation)](https://github.com/PaddlePaddle/models/tree/develop/PaddleNLP/examples/text_generation/ernie-gen) |ERNIE-GEN是百度发布的生成式预训练模型,通过Global-Attention的方式解决训练和预测曝光偏差的问题,同时使用Multi-Flow Attention机制来分别进行Global和Context信息的交互,同时通过片段生成的方式来增加语义相关性。|
| [ERNIE-GEN(An Enhanced Multi-Flow Pre-training and Fine-tuning Framework for Natural Language Generation)](./text_generation/ernie-gen) |ERNIE-GEN是百度发布的生成式预训练模型,通过Global-Attention的方式解决训练和预测曝光偏差的问题,同时使用Multi-Flow Attention机制来分别进行Global和Context信息的交互,同时通过片段生成的方式来增加语义相关性。|
| [PLATO-2](./dialogue/plato-2) | 百度自研领先的开放域对话预训练模型。[PLATO-2: Towards Building an Open-Domain Chatbot via Curriculum Learning](https://arxiv.org/abs/2006.16779) |
[Microsoft Research Asia Chinese Word-Segmentation Data Set](https://www.microsoft.com/en-us/download/details.aspx?id=52531)
[The third international Chinese language processing bakeoff: Word segmentation and named entity recognition](https://faculty.washington.edu/levow/papers/sighan06.pdf)