自然语言处理导航
NLTK教程:
- https://yq.aliyun.com/articles/419335?spm=a2c4e.11153940.blogcont419331.17.299a7d08tykCiH
- https://yiyibooks.cn/yiyi/nltk_python/index.html
jieba教程:
tensorflow教程:
文本分类:https://developers.google.cn/machine-learning/guides/text-classification/
word2vec:https://www.leiphone.com/news/201706/PamWKpfRFEI42McI.html
Machine Learning Repository:(可下载机器学习中的数据集)
Ubuntu Dialogue Corpus :http://dataset.cs.mcgill.ca/ubuntu-corpus-1.0/
基于深度学习的命名实体识别详解:https://blog.csdn.net/u012392084/article/details/78010047
NLP底层技术之句法分析:https://blog.csdn.net/qq_28031525/article/details/79187080#1pcfg
NLP参考资源
自然语言处理(Natural Language Processing)是深度学习的主要应用领域之一。
1. 教程
CS224d: Deep Learning for Natural Language Processing
http://cs224d.stanford.edu/
CS224d课程的课件
http://web.stanford.edu/class/cs224n/syllabus.html
CMU的NLP教程。该网页下方还有美国其他高校的NLP课程的链接。
http://demo.clab.cs.cmu.edu/NLP/
北京大学的NLP教程,特色:中文处理。缺点:传统方法居多,深度学习未涉及。
http://ccl.pku.edu.cn/alcourse/nlp/
COMS W4705: Natural Language Processing
http://www.cs.columbia.edu/~cs4705/
初学者如何查阅自然语言处理(NLP)领域学术资料
https://mp.weixin.qq.com/s/TSc4E8lKwgc-EvzP8OlJeg
揭开知识库问答KB-QA的面纱(知识图谱方面的系列专栏)
https://zhuanlan.zhihu.com/kb-qa
《语音与语言处理》第三版,NLP和语音合成方面的专著
http://web.stanford.edu/~jurafsky/slp3/ed3book.pdf
CIPS ATT 2017 文本分析和自然语言课程PPT
https://mp.weixin.qq.com/s/5KhTWdOk-b84DXmoVr68-A
CMU NN for NLP
http://phontron.com/class/nn4nlp2017/assets/slides/
CMU Machine Translation and Sequence to Sequence Models
http://phontron.com/class/mtandseq2seq2017/
Oxford Deep NLP 2017 course
https://github.com/oxford-cs-deepnlp-2017/lectures
2. 书籍
《Natural Language Processing with Python》,Steven Bird、Ewan Klein、Edward Loper著。这本书的作者们创建了著名的NLTK工具库。
http://ccl.pku.edu.cn/alcourse/nlp/LectureNotes/Natural%20Language%20Processing%20with%20Python.pdf
注:
Steven Bird,爱丁堡大学博士,墨尔本大学副教授。
http://www.stevenbird.net/about.html
Ewan Klein,苏格兰人,哥伦比亚大学博士(1978年),爱丁堡大学教授。
Edward Loper,宾夕法尼亚大学博士。
推荐5本经典自然语言处理书籍
https://mp.weixin.qq.com/s/0HmsMytif3INqAX1Si5ukA
3. 网站
一个自然语言处理爱好者的群体博客。包括52nlp、rickjin、liwei等国内外华人大牛.
http://www.52nlp.cn/
实战课程:自己动手做聊天机器人
http://www.shareditor.com/bloglistbytag/?tagname=%E8%87%AA%E5%B7%B1%E5%8A%A8%E6%89%8B%E5%81%9A%E8%81%8A%E5%A4%A9%E6%9C%BA%E5%99%A8%E4%BA%BA
北京大学计算机科学技术研究所语言计算与互联网挖掘研究
http://www.icst.pku.edu.cn/lcwm/
NLP深度学习方面的代码库
https://github.com/rockingdingo/deepnlp
NLP专家李维的blog
https://liweinlp.com/
一个NLP方面的blog
http://www.shuang0420.com/
一个DL+ML+NLP的blog
http://www.cnblogs.com/Determined22/
一个NLP方面的blog
http://www.cnblogs.com/robert-dlut/
一个NLP方面的blog
https://blog.csdn.net/wangxinginnlp
4. 工具
Natural Language Toolkit(NLTK)
官网:http://www.nltk.org/
可使用nltk.download()下载相关nltk官方提供的各种资源。
参考:
http://www.cnblogs.com/baiboy/p/nltk3.html
OpenNLP
http://opennlp.apache.org/
FudanNLP
https://github.com/FudanNLP/fnlp
Stanford CoreNLP
http://stanfordnlp.github.io/CoreNLP/
THUCTC
THUCTC(THU Chinese Text Classification)是由清华大学自然语言处理实验室推出的中文文本分类工具包。
http://thuctc.thunlp.org/
gensim
gensim是Python语言的计算文本相似度的程序包。
http://radimrehurek.com/gensim/index.html
安装指令:
pip install –upgrade gensim
GitHub 地址:
https://github.com/RaRe-Technologies/gensim
参考学习:
情感分析的新方法——基于Word2Vec /Doc2Vec/Python
http://www.open-open.com/lib/view/open1444351655682.html
Gensim Word2vec使用教程
http://blog.csdn.net/Star_Bob/article/details/47808499
GloVe
GloVe:Global Vectors for Word Representation
https://nlp.stanford.edu/projects/glove/
textsum
textsum是一个基于深度学习的文本自动摘要工具。
代码:
https://github.com/tensorflow/models/tree/master/textsum
参考:
http://www.jiqizhixin.com/article/1449
谷歌开源新的TensorFlow文本自动摘要代码:
TensorFlow文本摘要生成 – 基于注意力的序列到序列模型
http://blog.csdn.net/tensorflowshizhan/article/details/69230070
jieba
https://github.com/fxsjy/jieba
NLPIR:NLPIR汉语分词系统(又名ICTCLAS2013),是中科院张华平博士的作品。
http://ictclas.nlpir.org/
参考:
这个网页对于NLP的大多数功能进行了可视化的展示。NLP入门必看。
http://ictclas.nlpir.org/nlpir/
snownlp
https://github.com/isnowfy/snownlp
HanLP:HanLP是一个目前留学日本的中国学生的作品
http://hanlp.linrunsoft.com/
作者blog:
http://www.hankcs.com/
Github:
https://github.com/hankcs/HanLP/
从作者的名气来说,HanLP无疑是最低的,性能也不见得有多好。然而对于初学者来说,这却是最适合的工具。这主要体现在以下几个方面:
1.中文处理能力。NLTK和OpenNLP对中文支持非常差,这里不光是中文分词的问题,有些NLP算法需要一定的语言模型数据,但浏览NLTK官方的模型库,基本找不到中文模型数据。
2.jieba、IK之类的功能太单一,多数局限在中文分词方面领域。gensim、THUCTC专注于NLP的某一方面,也不是通用工具。
3.NLPIR和Stanford CoreNLP算是功能最强的工具包了。前者的问题在于收费不开源,后者的问题在于缺少中文文档。FudanNLP的相关文档较少,文档友好度不如HanLP。
4.HanLP在主页上提供了相关算法的blog,便于初学者快速掌握相关概念。其词典是明文发布,便于用户修改。HanLP执行时,会将明文词典以特定结构缓存,以提高执行效率。
注:不要以为中文有分词问题,就比别的语言复杂,英文还有词根问题呢。。。每种语言都不简单。
AllenNLP
AllenNLP是 Allen AI实验室的作品,采用深度学习技术,基于PyTorch开发。
http://allennlp.org/
Allen AI实验室由微软联合创始人Paul G. Allen投资创立。
http://allenai.org/
python版的汉字转拼音软件
https://github.com/mozillazg/python-pinyin
Java分布式中文分词组件-word分词
https://github.com/ysc/word
jena是一个语义网络、知识图谱相关的软件
http://jena.apache.org/
NLPchina
NLPchina(中国自然语言处理开源组织)旗下有许多好用的工具。
http://www.nlpcn.org/
Github:
https://github.com/NLPchina
Ansj
Ansj是一个NLPchina旗下的开源的Java中文分词工具,基于中科院的ictclas中文分词算法,比其他常用的开源分词工具(如mmseg4j)的分词准确率更高。
https://github.com/NLPchina/ansj_seg
Word2VEC_java
word2vec java版本的一个实现。
https://github.com/NLPchina/Word2VEC_java
doc2vec java版本的一个实现,基于Word2VEC_java。
https://github.com/yao8839836/doc2vec_java
ansj_fast_lda
LDA算法的Java包。
https://github.com/NLPchina/ansj_fast_lda
nlp-lang
这个项目是一个基本包.封装了大多数nlp项目中常用工具
https://github.com/NLPchina/nlp-lang
词性标注
ICTPOS3.0汉语词性标记集
http://jacoxu.com/ictpos3-0%E6%B1%89%E8%AF%AD%E8%AF%8D%E6%80%A7%E6%A0%87%E8%AE%B0%E9%9B%86/
Word Hashing
Word Hashing是非常重要的一个trick,以英文单词来说,比如good,他可以写成#good#,然后按tri-grams来进行分解为#go goo ood od#,再将这个tri-grams灌入到bag-of-word中,这种方式可以非常有效的解决vocabulary太大的问题(因为在真实的web search中vocabulary就是异常的大),另外也不会出现oov问题,因此英文单词才26个,3个字母的组合都是有限的,很容易枚举光。
那么问题就来了,这样两个不同的单词会不会产出相同的tri-grams,paper里面做了统计,说了这个冲突的概率非常的低,500K个word可以降到30k维,冲突的概率为0.0044%。
但是在中文场景下,这个Word Hashing估计没有这么有效了:
词汇共现
http://sewm.pku.edu.cn/TianwangLiterature/SEWM/2005(5)/%5b%b3%c2%c1%88,%20et%20al.,2005%5d/050929.pdf
词汇共现是指词汇在文档集中共同出现。以一个词为中心,可以找到一组经常与之搭配出现的词,作为它的共现词汇集。
词汇共现的其中一种用例:
有若干关键词,比如:水果、天气、风,有若干描述词,比如,很甜、晴朗、很大,然后现在要找出他们之间的搭配,在这个例子里,我们最终要找到:水果很甜、天气晴朗、风很大。
关键词提取
主要三种方法:1.基于统计特征,如TF-IDF;2.基于词图模型,如TextRank;3.基于主题模型,如LDA。
自然语言理解
Natural language understanding(NLU)属于NLP的一个分支,属于人工智能的一个部分,用来解决机器理解人类语言的问题,属于人工智能的核心难题。
http://www.shuang0420.com/2017/04/27/NLP%E7%AC%94%E8%AE%B0%20-%20NLU%E4%B9%8B%E6%84%8F%E5%9B%BE%E5%88%86%E7%B1%BB/
论文
《Distant Supervision for relation extraction without labeled data》
《Using Recurrent Neural Networks for Slot Filling in Spoken Language Understanding》
《Convolutional Neural Networks for Sentence Classification》
知识图谱参考资源
知识图谱构建技术综述
https://wenku.baidu.com/view/38ad3ef7e109581b6bd97f19227916888586b959.html
知识图谱技术综述
https://wenku.baidu.com/view/e69a3619fe00bed5b9f3f90f76c66137ee064f15.html
知识图谱技术原理介绍
https://wenku.baidu.com/view/b3858227c5da50e2534d7f08.html
基于知识图谱的问答系统关键技术研究
https://mp.weixin.qq.com/s/JLYegFP7kEg6n34crgP09g
什么是知识图谱?
https://mp.weixin.qq.com/s/XgKvh63wgEe-CR9bchp03Q
当知识图谱遇上聊天机器人
https://mp.weixin.qq.com/s/iqFXvhvYfOejaeNAhXxJEg
知识图谱前沿技术课程实录
https://mp.weixin.qq.com/s/U-dlYhnaR8OQw2UKYKUWKQ
阿里知识图谱首次曝光:每天千万级拦截量,亿级别全量智能审核
https://mp.weixin.qq.com/s/MZE_SXsNg6Yt4dz2fmB1sA
东南大学漆桂林:知识图谱的应用
https://mp.weixin.qq.com/s/WIro7pk7kboMvdwpZOSdQA
东南大学高桓:知识图谱表示学习
https://mp.weixin.qq.com/s/z1hhG4GaBQXPHHt9UGZPnA
复旦肖仰华:基于知识图谱的问答系统
https://mp.weixin.qq.com/s/JZYH_m1eS93KRjkWA82GoA
多源信息表示学习在知识图谱中的应用
https://mp.weixin.qq.com/s/cEmtOAtfP2gSBlaPfGXb3w
如何构建知识图谱
https://mp.weixin.qq.com/s/cL1aKdu8ig8-ocOPirXk2w
中文通用百科知识图谱(CN-DBpedia)
https://mp.weixin.qq.com/s/Nh7XJOLNBDdpibopVG4MrQ
原文链接:
https://blog.csdn.net/antkillerfarm/article/details/78082564
转自:https://blog.csdn.net/zouyu1746430162/article/details/79591670
自然语言处理(NLP)是计算机科学,人工智能,语言学关注计算机和人类(自然)语言之间的相互作用的领域。本文作者为NLP初学者整理了一份庞大的自然语言处理领域的概览。选取的参考文献与资料都侧重于最新的深度学习研究成果。这些资源能为想要深入钻研一个NLP任务的人们提供一个良好的开端。
指代消解
论文自动评分
- 论文:Automatic Text Scoring Using Neural Networks(使用神经网络的自动文本评分):https://arxiv.org/abs/1606.04289
- 论文:A Neural Approach to Automated Essay Scoring(一种自动将论文评分的神经学方法):http://www.aclweb.org/old_anthology/D/D16/D16-1193.pdf
- 挑战:Kaggle:The Hewlett Foundation: Automated Essay Scoring(Kaggle:The Hewlett Foundation:论文自动评分系统):https://www.kaggle.com/c/asap-aes
- 项目:Enhanced AI Scoring Engine(增强的人工智能得分引擎):https://github.com/edx/ease
自动语音识别
- 维基百科: 语言识别:https://en.wikipedia.org/wiki/Speech_recognition
- 论文:DeepSpeech 2: End-to-End Speech Recognition in English and Mandarin(深度语音2:用英语和普通话进行端对端语音识别):https://arxiv.org/abs/1512.02595
- 论文:WaveNet:A Generative Model for Raw Audio(WaveNet:原始音频的生成模型):https://arxiv.org/abs/1609.03499
- 项目:A TensorFlow implementation of Baidu’s Deep Speech architecture(百度深度语音架构的一个TensorFlow实现:https://github.com/mozilla/DeepSpeech
- 项目:Speech-to-Text-WaveNet: End-to-end sentence level English speech recognition using DeepMind’s WaveNet(Speech-to-Text-WaveNet: 使用DeepMind的WaveNet,对端到端句子的英语水平语音识别):https://github.com/buriburisuri/speech-to-text-wavenet
- 挑战:The 5th CHiME Speech Separation and Recognition Challenge(第五届CHiME语音的分离和识别挑战):http://spandh.dcs.shef.ac.uk/chime_challenge/
- 资料:The 5thCHiME Speech Separation and Recognition Challenge(第五届CHiME语音的分离和识别挑战):http://spandh.dcs.shef.ac.uk/chime_challenge/download.html
- 资料:CSTRVCTK Corpus :http://homepages.inf.ed.ac.uk/jyamagis/page3/page58/page58.html
- 资料:LibriSpeech ASR corpus:http://www.openslr.org/12/
- 资料:Switchboard-1 Telephone Speech Corpus:https://catalog.ldc.upenn.edu/ldc97s62
- 资料:TED-LIUM Corpus:http://www-lium.univ-lemans.fr/en/content/ted-lium-corpus
自动摘要
- 维基百科:自动摘要:https://en.wikipedia.org/wiki/Automatic_summarization
- 书籍:Automatic Text Summarization(自动本文摘要):https://www.amazon.com/Automatic-Text-Summarization-Juan-Manuel-Torres-Moreno/dp/1848216688/ref=sr_1_1?s=books&ie=UTF8&qid=1507782304&sr=1-1&keywords=Automatic+Text+Summarization
- 论文:Text Summarization Using Neural Networks(使用神经网络进行文本摘要):http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.823.8025&rep=rep1&type=pdf
- 论文:Ranking with Recursive Neural Networks and Its Application to Multi-DocumentSummarization(使用递归神经网络及其应用程序对多文档摘要进行排序):https://www.aaai.org/ocs/index.php/AAAI/AAAI15/paper/viewFile/9414/9520
- 资料:Text Analytics Conferences(文本分析会议):https://tac.nist.gov/data/index.html
- 资料:Document Understanding Conferences(文书理解会议):http://www-nlpir.nist.gov/projects/duc/data.html
共指消解
- 信息:共指消解:https://nlp.stanford.edu/projects/coref.shtml
- 论文:Deep Reinforcement Learning for Mention-Ranking Coreference Models(对Mention-Ranking的共指模型进行深度强化学习:https://arxiv.org/abs/1609.08667
- 论文:Improving Coreference Resolution by Learning Entity-Level Distributed Representations(通过学习实体级分布式表示来改善相关的解决方案):https://arxiv.org/abs/1606.01323
- 挑战:CoNLL 2012 Shared Task: Modeling Multilingual Unrestricted Coreference in OntoNotes(CoNLL 2012共享任务:在OntoNotes中对多语言的不受限制的共指进行建模):http://conll.cemantix.org/2012/task-description.html
- 挑战:CoNLL 2011 Shared Task: Modeling Unrestricted Coreference in OntoNotes(CoNLL 2011共享任务:在OntoNotes中对多语言的不受限制的共指进行建模):http://conll.cemantix.org/2011/task-description.html
语法错误校正
- 论文:Neural Network Translation Models for Grammatical Error Correction(语法错误校正的神经网络翻译模型):https://arxiv.org/abs/1606.00189
- 挑战:CoNLL 2013 Shared Task: Grammatical Error Correction(CoNLL 2013共享任务:语法错误校正):http://www.comp.nus.edu.sg/~nlp/conll13st.html
- 挑战:CoNLL 2014Shared Task: Grammatical Error Correction(CoNLL 2014共享任务:语法错误校正):http://www.comp.nus.edu.sg/~nlp/conll14st.html
- 资料:NUSNon-commercial research/trial corpus license:http://www.comp.nus.edu.sg/~nlp/conll14st/nucle_license.pdf
- 资料:Lang-8 Learner Corpora:http://cl.naist.jp/nldata/lang-8/
- 资料:Cornell Movie–Dialogs Corpus:http://www.cs.cornell.edu/~cristian/Cornell_Movie-Dialogs_Corpus.html
- 项目:Deep Text Corrector(深度文本校正器):https://github.com/atpaino/deep-text-corrector
- 产品:deep grammar:http://deepgrammar.com/
字素转换到音素
- 论文:Grapheme-to-Phoneme Models for (Almost) Any Language(适合(几乎)任何语言的字素到音素的模型):https://pdfs.semanticscholar.org/b9c8/fef9b6f16b92c6859f6106524fdb053e9577.pdf
- 论文:Polyglot Neural Language Models: A Case Study in Cross-Lingual Phonetic Representation Learning(多语言神经语言模型:跨语语音表达学习的案例研究):https://arxiv.org/pdf/1605.03832.pdf
- 论文:Multi task Sequence-to-Sequence Models for Grapheme-to-Phoneme Conversion(多任务序列到序列的字素到音素转换的模型):https://pdfs.semanticscholar.org/26d0/09959fa2b2e18cddb5783493738a1c1ede2f.pdf
- 项目:Sequence-to-Sequence G2P toolkit(序列到序列G2P工具包):https://github.com/cmusphinx/g2p-seq2seq
- 资料:Multilingual Pronunciation Data(多语种发音数据):https://drive.google.com/drive/folders/0B7R_gATfZJ2aWkpSWHpXUklWUmM
语种识别
- 维基百科: 语种识别:https://en.wikipedia.org/wiki/Language_identification
- 论文:AUTOMATIC LANGUAGE IDENTIFICATION USING DEEP NEURAL NETWORKS(使用深度神经网络的自动语言识别):https://repositorio.uam.es/bitstream/handle/10486/666848/automatic_lopez-moreno_ICASSP_2014_ps.pdf?sequence=1
- 挑战: 2015 Language Recognition Evaluation(2015语言识别评估):https://www.nist.gov/itl/iad/mig/2015-language-recognition-evaluation
语言建模
- 维基百科:语言模型:https://en.wikipedia.org/wiki/Language_model
- 工具包: KenLM Language Model Toolkit(KenLM语言模型工具包):http://kheafield.com/code/kenlm/
- 论文:Distributed Representations of Words and Phrases and their Compositionality(词汇和短语的分布表示及其组合性):http://papers.nips.cc/paper/5021-distributed-representations-of-words-and-phrases-and-their-compositionality.pdf
- 论文:Character-Aware Neural Language Models(Character-Aware神经语言模型):https://www.aaai.org/ocs/index.php/AAAI/AAAI16/paper/viewFile/12489/12017
- 资料: Penn Treebank :https://github.com/townie/PTB-dataset-from-Tomas-Mikolov-s-webpage/tree/master/data
词形还原
- 维基百科:词形还原:https://en.wikipedia.org/wiki/Lemmatisation
- 工具包:WordNet Lemmatizer:http://www.nltk.org/api/nltk.stem.html#nltk.stem.wordnet.WordNetLemmatizer.lemmatize
- 资料:Treebank-3:https://catalog.ldc.upenn.edu/ldc99t42
唇语辨别
- 维基百科:唇读法:https://en.wikipedia.org/wiki/Lip_reading
- 论文:Lip Reading Sentences in the Wild (在野外读懂唇语):https://arxiv.org/abs/1611.05358
- 论文:3D Convolutional Neural Networks for Cross Audio-Visual Matching Recognition(交叉视听匹配识别的3D卷积神经网络):https://arxiv.org/abs/1706.05739
- 项目: Lip Reading – Cross Audio-Visual Recognition using 3D Convolutional Neural Networks(唇读法—使用3D卷积神经网络的交叉视听识别:https://github.com/astorfi/lip-reading-deeplearning
- 资料: The GRID audiovisual sentence corpus:http://spandh.dcs.shef.ac.uk/gridcorpus/
机器翻译
- 论文:Neural Machine Translation by Jointly Learning to Align and Translate(通过共同学习来调整和翻译神经机器翻译):https://arxiv.org/abs/1409.0473
- 论文:Neural Machine Translation in Linear Tim(在线性时间中的神经机器翻译):https://arxiv.org/abs/1610.10099
- 挑战: ACL2014 NINTH WORKSHOP ON STATISTICAL MACHINE TRANSLATION(ACL2014第九届统计机器翻译研讨会):http://www.statmt.org/wmt14/translation-task.html#download
- 资料:OpenSubtitles2016:http://opus.lingfil.uu.se/OpenSubtitles2016.php
- 资料: WIT3:Web Inventory of Transcribed and Translated Talks:https://wit3.fbk.eu/
- 资料: The QCRI Educational Domain (QED) Corpus:http://alt.qcri.org/resources/qedcorpus/
命名实体识别
- 维基百科:命名实体识别:https://en.wikipedia.org/wiki/Named-entity_recognition
- 论文:Neural Architectures for Named Entity Recognition(命名实体识别的神经结构):https://arxiv.org/abs/1603.01360
- 项目: OSU Twitter NLP Tool:https://github.com/aritter/twitter_nlp
- 挑战: Named Entity Recognition in Twitter(在推特上被命名的实体识别):https://noisy-text.github.io/2016/ner-shared-task.html
- 资料:CoNLL-2002 NER corpus:https://github.com/teropa/nlp/tree/master/resources/corpora/conll2002
- 资料:CoNLL-2003 NER corpus:https://github.com/synalp/NER/tree/master/corpus/CoNLL-2003
释义检测
- 论文:Dynamic Pooling and Unfolding Recursive Autoencoders for Paraphrase Detection(动态池和展开递归自动编码器的释义检测):http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.650.7199&rep=rep1&type=pdf
- 项目:Paralex: Paraphrase-Driven Learning for Open Question Answering(Paralex:释义驱动学习的开放问答):http://knowitall.cs.washington.edu/paralex/
- 资料:Microsoft Research Paraphrase Corpus:https://www.microsoft.com/en-us/download/details.aspx?id=52398
- 资料:Microsoft Research Video Description Corpus :https://www.microsoft.com/en-us/download/details.aspx?id=52422&from=http%3A%2F%2Fresearch.microsoft.com%2Fen-us%2Fdownloads%2F38cf15fd-b8df-477e-a4e4-a4680caa75af%2F
- 资料: Pascal Dataset:http://nlp.cs.illinois.edu/HockenmaierGroup/pascal-sentences/index.html
- 资料:Flicker Dataset:http://nlp.cs.illinois.edu/HockenmaierGroup/8k-pictures.html
- 资料: TheSICK data set:http://clic.cimec.unitn.it/composes/sick.html
- 资料: PPDB:The Paraphrase Database:http://www.cis.upenn.edu/~ccb/ppdb/
- 资料:WikiAnswers Paraphrase Corpus:http://knowitall.cs.washington.edu/paralex/wikianswers-paraphrases-1.0.tar.gz
语法分析
- 维基百科:语法分析:https://en.wikipedia.org/wiki/Parsing
- 工具包:The Stanford Parser: A statistical parser:https://nlp.stanford.edu/software/lex-parser.shtml
- 工具包: spaCyparser:https://spacy.io/docs/usage/dependency-parse
- 论文:A fastand accurate dependency parser using neural networks(快速而准确地使用神经网络的依赖解析器):http://www.aclweb.org/anthology/D14-1082
- 挑战:CoNLL2017 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies(CoNLL2017共享任务:从原始文本到通用依赖项的多语言解析):http://universaldependencies.org/conll17/
- 挑战:CoNLL2016 Shared Task: Multilingual Shallow Discourse Parsing(CoNLL2016共享任务:多语言的浅会话解析):http://www.cs.brandeis.edu/~clp/conll16st/
词性标记
- 维基百科:词性标记:https://en.wikipedia.org/wiki/Part-of-speech_tagging
- 论文:Unsupervised Part-Of-Speech Tagging with Anchor Hidden Markov Models(有Anchor Hidden Markov模型的非监督性的词性标记):https://transacl.org/ojs/index.php/tacl/article/viewFile/837/192
- 资料:Treebank-3:https://catalog.ldc.upenn.edu/ldc99t42
- 工具包:nltk.tag package:http://www.nltk.org/api/nltk.tag.html
拼音与中文转换
- 论文:Neural Network Language Model for Chinese Pinyin Input Method Engine(中文拼音输入法引擎的神经网络语言模型):http://aclweb.org/anthology/Y15-1052
- 项目:Neural Chinese Transliterator:https://github.com/Kyubyong/neural_chinese_transliterator
问答系统
- 维基百科:问答系统:https://en.wikipedia.org/wiki/Question_answering
- 论文:Ask Me Anything: Dynamic Memory Networks for Natural Language Processing(自然语言处理的动态内存网络):http://www.thespermwhale.com/jaseweston/ram/papers/paper_21.pdf
- 论文:Dynamic Memory Networks for Visual and Textual Question Answering(用于视觉和文本的问答系统的动态记忆网络):http://proceedings.mlr.press/v48/xiong16.pdf
- 挑战:TREC Question Answering Task(TREC问答系统任务):http://trec.nist.gov/data/qamain.html
- 挑战:SemEval-2017 Task 3: Community Question Answering:http://alt.qcri.org/semeval2017/task3/
- 资料:MSMARCO: Microsoft MAchine Reading COmprehension Dataset(MSMARCO:微软机器阅读理解数据集)http://www.msmarco.org/
- 资料:Maluuba NewsQA:https://github.com/Maluuba/newsqa
- 资料:SQuAD:100,000+ Questions for Machine Comprehension of Text(SQuAD:100,000+个文本的机器理解的问题):https://rajpurkar.github.io/SQuAD-explorer/
- 资料:Graph Questions: A Characteristic-rich Question Answering Dataset(图形问题:一个特征丰富的问题回答数据集):https://github.com/ysu1989/GraphQuestions
- 资料: Story Cloze Test and ROC Stories Corpora:http://cs.rochester.edu/nlp/rocstories/
- 资料:Microsoft Research WikiQA Corpus:https://www.microsoft.com/en-us/download/details.aspx?id=52419&from=http%3A%2F%2Fresearch.microsoft.com%2Fen-us%2Fdownloads%2F4495da01-db8c-4041-a7f6-7984a4f6a905%2Fdefault.aspx
- 资料:DeepMind Q&A Dataset:http://cs.nyu.edu/~kcho/DMQA/
- 资料: QASent:http://cs.stanford.edu/people/mengqiu/data/qg-emnlp07-data.tgz
关系提取
- 维基百科:关系提取:https://en.wikipedia.org/wiki/Relationship_extraction
- 论文:A deep learning approach for relationship extraction from interaction context in social manufacturing paradigm(一种从社会生产范例的互动情境中提取关系深度学习的方法):http://www.sciencedirect.com/science/article/pii/S0950705116001210
语义角色标记
- 维基百科:语义角色标记:https://en.wikipedia.org/wiki/Semantic_role_labeling
- 书籍:Semantic Role Labeling(语义角色标记):https://www.amazon.com/Semantic-Labeling-Synthesis-Lectures-Technologies/dp/1598298313/ref=sr_1_1?s=books&ie=UTF8&qid=1507776173&sr=1-1&keywords=Semantic+Role+Labeling
- 论文:End-to-end Learning of Semantic Role Labeling Using Recurrent Neural Networks(使用循环神经网络对语义角色标签进行端到端学习):http://www.aclweb.org/anthology/P/P15/P15-1109.pdf
- 论文:Neural Semantic Role Labeling with Dependency Path Embeddings(有着依赖路径嵌入的神经语义角色标记):https://arxiv.org/abs/1605.07515
- 挑战:CoNLL-2005 Shared Task: Semantic Role Labeling(CoNLL-2005共享任务:语义角色标记):http://www.cs.upc.edu/~srlconll/st05/st05.html
- 挑战:CoNLL-2004 Shared Task: Semantic Role Labeling(CoNLL-2004共享任务:语义角色标记):http://www.cs.upc.edu/~srlconll/st04/st04.html
- 工具包:Illinois Semantic Role Labeler(SRL):http://cogcomp.org/page/software_view/SRL
- 资料:CoNLL-2005 Shared Task: Semantic Role Labeling(CoNLL-2005共享任务:语义角色标记):http://www.cs.upc.edu/~srlconll/soft.html
语句边界消歧
- 维基百科:语句边界消歧:https://en.wikipedia.org/wiki/Sentence_boundary_disambiguation
- 论文:A Quantitative and Qualitative Evaluation of Sentence Boundary Detection for theClinical Domain(对临床领域的语句边界检测进行定量和定性的评估):https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5001746/
- 工具包: NLTK Tokenizers:http://www.nltk.org/_modules/nltk/tokenize.html
- 资料: The British National Corpus:http://www.natcorp.ox.ac.uk/
- 资料:Switchboard-1 Telephone Speech Corpus:https://catalog.ldc.upenn.edu/ldc97s62
情绪分析
- 维基百科:情绪分析:https://en.wikipedia.org/wiki/Sentiment_analysis
- 信息:Awesome Sentiment Analysis(了不起的情绪分析):https://github.com/xiamx/awesome-sentiment-analysis
- 挑战:Kaggle: UMICH SI650 – Sentiment Classification(Kaggle: UMICH SI650 – 情绪分类):https://www.kaggle.com/c/si650winter11#description
- 挑战:SemEval-2017 Task 4: Sentiment Analysis in Twitter(SemEval-2017任务4:推特上的情绪分析):http://alt.qcri.org/semeval2017/task4/
- 项目:SenticNet:http://sentic.net/about/
- 资料:Multi-Domain Sentiment Dataset(version2.0):http://www.cs.jhu.edu/~mdredze/datasets/sentiment/
- 资料:Stanford Sentiment Treebank:https://nlp.stanford.edu/sentiment/code.html
- 资料:Twitter Sentiment Corpus:http://www.sananalytics.com/lab/twitter-sentiment/
- 资料:Twitter Sentiment Analysis Training Corpus:http://thinknook.com/twitter-sentiment-analysis-training-corpus-dataset-2012-09-22/
源分离
- 维基百科:源分离:https://en.wikipedia.org/wiki/Source_separation
- 论文:From Blind to Guided Audio Source Separation(从盲目到有指导性的音频源分离):https://hal-univ-rennes1.archives-ouvertes.fr/hal-00922378/document
- 论文:Joint Optimization of Masks and Deep Recurrent Neural Networks for Monaural Source Separation (对单声道分离的掩膜和深层循环神经网络的联合优化):https://arxiv.org/abs/1502.04149
- 挑战:Signal Separation Evaluation Campaign(信号分离评估活动):https://sisec.inria.fr/
- 挑战: CHiME Speech Separation and Recognition Challenge(CHiME语音分离和识别的挑战):http://spandh.dcs.shef.ac.uk/chime_challenge/
说话者识别
- 维基百科:说话者识别:https://en.wikipedia.org/wiki/Speaker_recognition
- 论文:A NOVEL SCHEME FOR SPEAKER RECOGNITION USING A PHONETICALLY-AWARE DEEP NEURAL NETWORK(一种使用语音识别的深度神经网络的新方案):https://pdfs.semanticscholar.org/204a/ff8e21791c0a4113a3f75d0e6424a003c321.pdf
- 论文:DEEP NEURAL NETWORKS FOR SMALL FOOTPRINT TEXT-DEPENDENT SPEAKER VERIFICATION(深度神经网络,用于小范围的文本依赖的说话者验证):https://static.googleusercontent.com/media/research.google.com/en//pubs/archive/41939.pdf
- 挑战: NIST Speaker Recognition Evaluation(NIST说话者识别评价):https://www.nist.gov/itl/iad/mig/speaker-recognition
语音分段
- 维基百科:语音分段:https://en.wikipedia.org/wiki/Speech_segmentation
- 论文:Word Segmentation by 8-Month-Olds: When Speech Cues Count More Than Statistics(8个月大婴儿的单词分段:当语音提示比统计数字更重要时):http://www.utm.toronto.edu/infant-child-centre/sites/files/infant-child-centre/public/shared/elizabeth-johnson/Johnson_Jusczyk.pdf
- 论文:Unsupervised Word Segmentation and Lexicon Discovery Using Acoustic Word Embeddings(不受监督的单词分割和使用声学词嵌入的词汇发现):https://arxiv.org/abs/1603.02845
- 资料:CALLHOME Spanish Speech:https://catalog.ldc.upenn.edu/ldc96s35
语音合成
- 维基百科:语音合成:https://en.wikipedia.org/wiki/Speech_synthesis
- 论文:WaveNet:A Generative Model for Raw Audio(WaveNet:原始音频的生成模型):https://arxiv.org/abs/1609.03499
- 论文:Tacotron:Towards End-to-End Speech Synthesis(Tacotron:对端到端的语音合成):https://arxiv.org/abs/1703.10135
- 资料: The World English Bible:https://github.com/Kyubyong/tacotron
- 资料: LJ Speech Dataset:https://github.com/keithito/tacotron
- 资料: Lessac Data:http://www.cstr.ed.ac.uk/projects/blizzard/2011/lessac_blizzard2011/
- 挑战:Blizzard Challenge 2017:https://synsig.org/index.php/Blizzard_Challenge_2017
- 项目: The Festvox project:http://www.festvox.org/index.html
- 工具包:Merlin: The Neural Network (NN) based Speech Synthesis System(Merlin:基于神经网络的语音合成系统):https://github.com/CSTR-Edinburgh/merlin
语音增强
- 维基百科:语音增强:https://en.wikipedia.org/wiki/Speech_enhancement
- 书籍: Speech enhancement: theory and practice(语音增强:理论与实践):https://www.amazon.com/Speech-Enhancement-Theory-Practice-Second/dp/1466504218/ref=sr_1_1?ie=UTF8&qid=1507874199&sr=8-1&keywords=Speech+enhancement%3A+theory+and+practice
- 论文 An Experimental Study on Speech Enhancement Based on Deep Neural Network(一项基于深度神经网络的语音增强实验):http://staff.ustc.edu.cn/~jundu/Speech%20signal%20processing/publications/SPL2014_Xu.pdf
- 论文: A Regression Approach to Speech Enhancement Based on Deep Neural Networks(一种基于深度神经网络的语音增强的回归方法):https://www.researchgate.net/profile/Yong_Xu63/publication/272436458_A_Regression_Approach_to_Speech_Enhancement_Based_on_Deep_Neural_Networks/links/57fdfdda08aeaf819a5bdd97.pdf
- 论文:Speech Enhancement Based on Deep Denoising Autoencoder(基于深度降噪自编码的语音增强):https://www.researchgate.net/profile/Yu_Tsao/publication/283600839_Speech_enhancement_based_on_deep_denoising_Auto-Encoder/links/577b486108ae213761c9c7f8/Speech-enhancement-based-on-deep-denoising-Auto-Encoder.pdf
词干提取
- 维基百科:词干提取:https://en.wikipedia.org/wiki/Stemming
- 论文: A BACKPROPAGATION NEURAL NETWORK TO IMPROVE ARABIC STEMMING(一个反向传播的神经网络,用来改善阿拉伯语的词干提取):http://www.jatit.org/volumes/Vol82No3/7Vol82No3.pdf
- 工具包: NLTK Stemmers:http://www.nltk.org/howto/stem.html
术语提取
- 维基百科:术语提取:https://en.wikipedia.org/wiki/Terminology_extraction
- 论文: Neural Attention Models for Sequence Classification: Analysis and Application to KeyTerm Extraction and Dialogue Act Detection(序列分类的神经提示模型:分析和应用于关键词提取和对话法检测):https://arxiv.org/pdf/1604.00077.pdf
文本简化
- 维基百科:文本简化:https://en.wikipedia.org/wiki/Text_simplification
- 论文:Aligning Sentences from Standard Wikipedia to Simple Wikipedia(调整句子,从标准的维基百科到简单的维基百科):https://ssli.ee.washington.edu/~hannaneh/papers/simplification.pdf
- 论文:Problems in Current Text Simplification Research: New Data Can Help(当前文本简化研究中的问题:可提供帮助的新数据):https://pdfs.semanticscholar.org/2b8d/a013966c0c5e020ebc842d49d8ed166c8783.pdf
- 资料:Newsela Data:https://newsela.com/data/
文本蕴涵
- 维基百科:文本蕴含:https://en.wikipedia.org/wiki/Textual_entailment
- 项目:Textual Entailment with TensorFlow(文本蕴含与TensorFlow):https://github.com/Steven-Hewitt/Entailment-with-Tensorflow
- 竞赛:SemEval-2013 Task 7: The Joint Student Response Analysis and 8th Recognizing Textual Entailment Challenge(SemEval-2013任务7:联合学生反应分析和第8届认知文本蕴含挑战):https://www.cs.york.ac.uk/semeval-2013/task7.html
音译
- 维基百科:音译:https://en.wikipedia.org/wiki/Transliteration
- 论文:A Deep Learning Approach to Machine Transliteration(一个机器音译的深度学习方法):https://pdfs.semanticscholar.org/54f1/23122b8dd1f1d3067cf348cfea1276914377.pdf
- 项目:Neural Japanese Transliteration—can you do better than SwiftKey™ Keyboard?(神经日语音译:你能比SwiftKey键盘做得更好吗?):https://github.com/Kyubyong/neural_japanese_transliterator
词嵌入
- 维基百科:词嵌入:https://en.wikipedia.org/wiki/Word_embedding
- 工具包:Gensim: word2vec:https://radimrehurek.com/gensim/models/word2vec.html
- 工具包:fastText:https://github.com/facebookresearch/fastText
- 工具包:GloVe:Global Vectors for Word Representation:https://nlp.stanford.edu/projects/glove/
- 信息:Where to get a pretrained model?(哪里能够获得一个预先训练的模型?):https://github.com/3Top/word2vec-api
- 项目:Pre-trained word vectors of 30+ languages(30多种语言的预先训练的词向量):https://github.com/Kyubyong/wordvectors
- 项目:Polyglot: Distributed word representations for multilingual NLP(Polyglot:多语言NLP的分布式词汇表征):https://sites.google.com/site/rmyeid/projects/polyglot
词汇预测
- 信息:What is Word Prediction?(什么是词汇预测?):http://www2.edc.org/ncip/library/wp/what_is.htm
- 论文: The prediction of character based on recurrent neural network language model(基于循环神经网络语言模型的字符预测):http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=7960065
- 论文: An Embedded Deep Learning based Word Prediction(一个基于深度学习的词汇预测):https://arxiv.org/abs/1707.01662
- 论文:Evaluating Word Prediction: Framing Keystroke Savings(评估单词预测:框击键保存):http://aclweb.org/anthology/P08-2066
- 资料:An Embedded Deep Learning based Word Prediction(一个基于深度学习的词汇预测):https://github.com/Meinwerk/WordPrediction/master.zip
- 项目: Word Prediction using Convolutional Neural Networks—can you do better than iPhone™ Keyboard?(使用卷积神经网络的词汇预测——你能比iPhone键盘做得更好吗?):https://github.com/Kyubyong/word_prediction
词分割
- 论文: Neural Word Segmentation Learning for Chinese(中文的神经词分割学习):https://arxiv.org/abs/1606.04300
- 项目:Convolutional neural network for Chinese word segmentation(中文的词分割的卷积神经网络):https://github.com/chqiwang/convseg
- 工具包:Stanford Word Segmenter:https://nlp.stanford.edu/software/segmenter.html
- 工具包: NLTK Tokenizers:http://www.nltk.org/_modules/nltk/tokenize.html
词义消歧
- 维基百科:词义消歧:https://en.wikipedia.org/wiki/Word-sense_disambiguation
- 论文:Train-O-Matic: Large-Scale Supervised Word Sense Disambiguation in Multiple Languages without Manual Training Data(Train-O-Matic:在没有人工训练数据的情况下,在多种语言中大规模的监督词义消歧):http://www.aclweb.org/anthology/D17-1008
- 资料:Train-O-Matic Data:http://trainomatic.org/data/train-o-matic-data.zip
- 资料:BabelNet:http://babelnet.org/