如何在2019年变成NLP专家

【导读】本文介绍了近期自然语言处理的一些论文,代码,博客及研究趋势等。

 Fastai

  • Lesson 4 Practical Deep Learning for Coders

https://course.fast.ai/videos/?lesson

它会教你在fastai,语言模型是如何实现的。

LSTM:

即使transfomer更为流行,你还是有必要学习一些LSTM相关的知识, 因为在某些时候你仍然可以使用它,并且它是第一个在序列数据上取得较好较好效果的模型。

  • LSTM原始论文

https://www.bioinf.jku.at/publications/older/2604.pdf

  • 详细解释了LSTM 模型的博客:Understanding LSTM Networks blog

https://colah.github.io/posts/2015-08-Understanding-LSTMs

AWD_LSTM

在LSTM的基础上增加了dropout等,克服原始LSTM的缺点。

  • 论文:

https://arxiv.org/pdf/1708.02182.pdf

  • Salesforce 官方实现:

https://github.com/salesforce/awd-lstm-lm

  • fastai 实现:

https://github.com/fastai/fastai/blob/master/fastai/text/models/awd_lstm.py

Pointer模型

  • 论文:

https://arxiv.org/pdf/1609.07843.pdf

  • 官方视频介绍:

https://www.youtube.com/watch?v=Ibt8ZpbX3D8

Improving Neural Language Models with a continuous cache论文:

https://openreview.net/pdf?id=B14E5qee

Attention

只要记得 Attention is not all you need.

  • CS224n 视频从 1:00:55 开始,解释了attention.

https://www.youtube.com/watch?v=XXtpJxZBa2c

  • Attention is all you need 论文,同时提出了transformer。

https://papers.nips.cc/paper/7181-attention-is-all-you-need.pdf

  • 官方视频介绍

https://www.youtube.com/watch?v=rBCqOTEfxvg

  • 谷歌博客:

https://ai.googleblog.com/2017/08/transformer-novel-neural-network.html

  • 另一版本的transformer:Transformer-XL: Attentive Language Models Beyond a Fixed Length Contex paper

https://arxiv.org/pdf/1901.02860.pdf

  • 谷歌官方博客Transformer-XL

https://ai.googleblog.com/2019/01/transformer-xl-unleashing-potential-of.html

  • Transformer-XL — Combining Transformers and RNNs Into a State-of-the-art Language Model

https://www.lyrn.ai/2019/01/16/transformer-xl-sota-language-model

  • Attention and Memory in Deep Learning and NLP blog

http://www.wildml.com/2016/01/attention-and-memory-in-deep-learning-and-nlp

  • Attention and Augmented Recurrent Neural Networks blog

   https://distill.pub/2016/augmented-rnns

  • Building the Mighty Transformer for Sequence Tagging in PyTorch: Part 1 blog

https://medium.com/@kolloldas/building-the-mighty-transformer-for-sequence-tagging-in-pytorch-part-i-a1815655cd8

  • Building the Mighty Transformer for Sequence Tagging in PyTorch: Part 2 blog

https://medium.com/@kolloldas/building-the-mighty-transformer-for-sequence-tagging-in-pytorch-part-ii-c85bf8fd145

多任务学习

  • An overview of Multi-Task Learning in deep neural networks

 https://arxiv.org/pdf/1706.05098.pdf

  • The Natural Language Decathlon: Multitask Learning as Question Answering

https://arxiv.org/abs/1806.08730

  • Multi-Task Deep Neural Networks for Natural Language Understanding

https://arxiv.org/pdf/1901.11504.pdf

PyTorch

  • Pytorch 处理文本的教程

https://pytorch.org/tutorials/#text

  • 最近的进展在

http://ruder.io/nlp-imagenet

ELMo

  • Deep Contextualized word representations论文

https://arxiv.org/abs/1802.05365

  • 视频介绍:

https://vimeo.com/277672840

ULMFit:

  • Universal Language Model Fine-tuning for Text Classification论文:

https://arxiv.org/abs/1801.06146

  • Jeremy Howard 的博客

http://nlp.fast.ai/classification/2018/05/15/introducting-ulmfit.html

OpenAI GPT

  • GPT1 论文:

https://s3-us-west-2.amazonaws.com/openai-assets/research-covers/language-unsupervised/language_understanding_paper.pdf

  • 博客:

https://openai.com/blog/language-unsupervised

  • 代码:

https://github.com/openai/finetune-transformer-lm

  • GPT2论文:

https://d4mucfpksywv.cloudfront.net/better-language-models/language_models_are_unsupervised_multitask_learners.pdf

  • 博客:

https://openai.com/blog/better-language-models

  • 代码:

https://github.com/openai/gpt-2

  • GPT2 视频:

   https://www.youtube.com/watch?v=T0I88NhR9M

BERT

  •  BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding论文:

https://arxiv.org/abs/1810.04805

  • 谷歌官方博客:

https://ai.googleblog.com/2018/11/open-sourcing-bert-state-of-art-pre.html

  • Dissecting BERT Part 1: The Encoder 博客

https://medium.com/dissecting-bert/dissecting-bert-part-1-d3c3d495cdb3

  • Understading BERT Part 2: BERT Specifics 博客

https://medium.com/dissecting-bert/dissecting-bert-part2-335ff2ed9c73

  • Dissecting BERT Appendix: The Decoder博客:

https://medium.com/dissecting-bert/dissecting-bert-appendix-the-decoder-3b86f66b0e5f


原文链接:

https://medium.com/@kushajreal/how-to-become-an-expert-in-nlp-in-2019-1-945f4e9073c0

-END-

专 · 知

专知,专业可信的人工智能知识分发,让认知协作更快更好!欢迎登录www.zhuanzhi.ai,注册登录专知,获取更多AI知识资料!

欢迎微信扫一扫加入专知人工智能知识星球群,获取最新AI专业干货知识教程视频资料和与专家交流咨询!

请加专知小助手微信(扫一扫如下二维码添加),加入专知人工智能主题群,咨询技术商务合作~

专知《深度学习:算法到实战》课程全部完成!540+位同学在学习,现在报名,限时优惠!网易云课堂人工智能畅销榜首位!

点击“阅读原文”,了解报名专知《深度学习:算法到实战》课程

展开全文
Top
微信扫码咨询专知VIP会员