- 1) Google 翻譯提供 “特定性別翻譯”,大大消除性別表述歧義
https://mp.weixin.qq.com/s/3b5PlwbTli1IgZZsRrat9g
- 2)清華大學(xué)自然語(yǔ)言處理組整理的必讀機(jī)翻論文list
https://github.com/THUNLP-MT/MT-Reading-List/blob/master/README.md#domain_adaptation
- 3)盤(pán)點(diǎn)2018年NLP令人激動(dòng)的10大想法
https://mp.weixin.qq.com/s/5F6h2F8L-OdFJxBY2QrWJg
- 4) 基于自編碼器的表征學(xué)習(xí):如何攻克半監(jiān)督和無(wú)監(jiān)督學(xué)習(xí)?
https://mp.weixin.qq.com/s/Dqz97_U5pw_4d9KFblJfLg
- 5)一招理解LSTM/GRU門(mén)控機(jī)制
https://mp.weixin.qq.com/s/oMeH8G4SP6xWqgF_An1lxA
- 6)聊聊Atman數(shù)據(jù)的高效利用(一)——數(shù)據(jù)清洗(去噪)
https://mp.weixin.qq.com/s/-OJGFzqUF2wVDTeONG1f7Q
- 7)聊聊Atman數(shù)據(jù)的高效利用(二)——數(shù)據(jù)增強(qiáng)
https://mp.weixin.qq.com/s/Ye-gsELO1pF6GHB4YQIPrg
- 8)【清華劉洋】244頁(yè)機(jī)器翻譯學(xué)術(shù)論文寫(xiě)作方法和技巧
https://mp.weixin.qq.com/s/6-Z3eBlybEfCMbITi17hjw
- 9)回顧 | Facebook AI Research研究員顧佳濤:面向低資源語(yǔ)言的多語(yǔ)神經(jīng)機(jī)器翻譯
https://mp.weixin.qq.com/s/KrjELh9RC7b1Gypn_Zzz-A
- 10)通過(guò)更快的訓(xùn)練和推理,將神經(jīng)機(jī)器翻譯推廣到更大的數(shù)據(jù)集
https://code.fb.com/ai-research/scaling-neural-machine-translation-to-bigger-data-sets-with-faster-training-and-inference/
- 11)在翻譯中發(fā)現(xiàn):通過(guò)深入學(xué)習(xí)從頭開(kāi)始構(gòu)建語(yǔ)言翻譯
https://blog.floydhub.com/language-translator/
- 12)Unsupervised machine translation: A novel approach to provide fast, accurate translations for more languages
https://code.fb.com/ai-research/unsupervised-machine-translation-a-novel-approach-to-provide-fast-accurate-translations-for-more-languages/
- 13)The Illustrated BERT, ELMo, and co. (How NLP Cracked Transfer Learning)
http://jalammar.github.io/illustrated-bert/
- 14)The Annotated Transformer
http://nlp.seas.harvard.edu/2018/04/03/attention.html
- 15)Natural Language Processing is Fun!
https://medium.com/@ageitgey/natural-language-processing-is-fun-9a0bff37854e
- 16) Meta-Learning 元學(xué)習(xí):學(xué)會(huì)快速學(xué)習(xí)
https://mp.weixin.qq.com/s/zJzAHaqBewItaXWt7lqk-Q
- 17)Understanding Back-Translation at Scale
https://mp.weixin.qq.com/s/2shtJx6A9lagGhsLQU3Igg
- 18) 注意力機(jī)制(Attention)最新綜述論文及相關(guān)源碼
https://mp.weixin.qq.com/s/WwBnPQweznxBtvpAjobUGg
- 19)囊括歐亞非大陸多種語(yǔ)言的25個(gè)平行語(yǔ)料庫(kù)數(shù)據(jù)集
https://mp.weixin.qq.com/s/kNX6LCsAxdJh_Nx0eBi7uw
-? 20)EMNLP 2018 最佳論文:Facebook 提升 11BLEU 的無(wú)監(jiān)督機(jī)器翻譯
https://mp.weixin.qq.com/s/HYIISWk3Ftan59CxCSrsMw
- 21)EMNLP 2018 | 用強(qiáng)化學(xué)習(xí)做神經(jīng)機(jī)器翻譯:中山大學(xué)&MSRA填補(bǔ)多項(xiàng)空白
https://mp.weixin.qq.com/s/GbrrHt8RFZvUeplUEmaTsg
- 22)EMNLP 2018 | 騰訊AI Lab提出翻譯改進(jìn)模型Transformer的3個(gè)優(yōu)化方法
https://mp.weixin.qq.com/s/EKzLQ9eYTR5l-GYtmb_3qQ
- 23) EMNLP 2018 | 為什么使用自注意力機(jī)制?
https://mp.weixin.qq.com/s/jRfOzKO6OlQLokIzipbqUQ
- 24) 機(jī)器翻譯新突破!“普適注意力”模型:概念簡(jiǎn)單參數(shù)少,性能大增
https://mp.weixin.qq.com/s/lZOIK5BRXZrmL_Z9crl6sA
- 25)專(zhuān)訪(fǎng) | 監(jiān)管機(jī)器翻譯質(zhì)量?且看阿里如何搭建翻譯質(zhì)量評(píng)估模型
https://mp.weixin.qq.com/s/T5ybwhxZF2bN2hIPxELe4A
- 26) 資源 | 讓手機(jī)神經(jīng)網(wǎng)絡(luò)速度翻倍:Facebook開(kāi)源高性能內(nèi)核庫(kù)QNNPACK
https://mp.weixin.qq.com/s/KuckBYOMbKYPDFabT63DsA
- 27) EMNLP 2018 | Google AI提出「透明注意力」機(jī)制,實(shí)現(xiàn)更深層NMT模型
https://mp.weixin.qq.com/s/h7sLwVXb_UI8jvJU-oe3Cg
- 28) 32分鐘訓(xùn)練神經(jīng)機(jī)器翻譯,速度提升45倍
https://mp.weixin.qq.com/s/mg-d1W5i9rzaLMNrvq0tSQ
- 29) EMNLP 2018 | 結(jié)合通用和專(zhuān)用NMT的優(yōu)勢(shì),CMU為NMT引入「語(yǔ)境參數(shù)生成器」
https://mp.weixin.qq.com/s/nTLxvrxnKRmHl8ZDWxI42Q
- 30) 入門(mén) | 無(wú)需雙語(yǔ)語(yǔ)料庫(kù)的無(wú)監(jiān)督式機(jī)器翻譯
https://mp.weixin.qq.com/s/wf4m4xxxJoEltWIglS-SSA