ΑΙhub.org
EMNLP 2021 in tweets
The Conference on Empirical Methods in Natural Language Processing (EMNLP 2021) took place from the 7th to the 11th of November both in Punta Cana and online. If you did not have time to check the papers and the keynotes at the main conference, here are the livetweeted keynotes and papers sorted by language.
Keynotes
Where next? Towards multi-text consumption via three inspired research lines
✍️Live Notes of EMNLP 2021 #EMNLP2021 Keynote by Ido Dagan on 3 directions that #NLProc should pursue: https://t.co/LLeBjcffOP @emnlpmeeting
— Zhijing Jin (@ZhijingJin) November 7, 2021
The Language System in the Human Brain
At #EMNLP2021 Evelina Fedorenko makes a strong case to defuse criticism that neural language models cannot "think". Neither can the human language modules in the brain, she argues, based on human brain studies. #EMNLP2021livetweet pic.twitter.com/WXiUZkWZTy
— Zeta Alpha (@ZetaVector) November 8, 2021
LT4All!? Rethinking the Agenda
#EMNLP2021livetweet
Should language technology (LT) be one-size-fits-all? In his keynote speech, Prof. Steven Bird argues that we should *not* assume the approaches and assumptions for developing LT apply to all languages & communities.
1/7#emnlp2021enhttps://t.co/S5onMBSRu1— Renny P. Kusumawardani (@rennypradina) November 10, 2021
Papers
Brazilian Portuguese livetweets
Transformer Feed-Forward Layers Are Key-Value Memories
#EMNLP2021 #EMNLP2021livetweet
[PT-BR]
Para as pessoas interessadas em interpretação do que os modelos Transformers estão aprendendo, segue um trabalho bem interessando "Transformer Feed-Forward Layers Are Key-Value Memories", da Mor Geva et al.https://t.co/COXgTeN2cb— Gabriel Oliveira (@GOdS_briel) November 9, 2021
CIDEr-R: Robust Consensus-based Image Description Evaluation
#EMNLP2021livetweet
[PT-BR] Hoje é dia de fazer propaganda do meu trabalho 😀CIDEr-R: Robust Consensus-based Image Description Evaluation
✏️ Gabriel Oliveira (me), Esther Colombini (@esthercolombini), Sandra Avila(@sandraavilabr)
— Gabriel Oliveira (@GOdS_briel) November 11, 2021
CLIPScore: A Reference-free Evaluation Metric for Image Captioning
#EMNLP2021LiveTweet
[PT-BR] Para as pessoas interessadas em métodos de avaliação para tarefas multi-modal aqui vai um recomendação de leitura:CLIPScore: A Reference-free Evaluation Metric for Image Captioning
✏️ Jack Hessel(@jmhessel) et al.
— Gabriel Oliveira (@GOdS_briel) November 10, 2021
Machine-in-the-Loop Rewriting for Creative Image Captioning
#EMNLP2021LiveTweet
[PT-BR] Segue uma recomendação de leitura às pessoas interessadas design em conjunto com aprendizado de máquina:Machine-in-the-Loop Rewriting for Creative Image Captioning
✏️Vishakh Padmakumarl (@vishakh_pk) et al.
?️https://t.co/Zwz43NJKz6— Gabriel Oliveira (@GOdS_briel) November 10, 2021
English livetweets
Grammatical Profiling for Semantic Change Detection
Most works on semantic change use neural word representations(w2vBERT). How do features compare? Surprisingly well.@glnmario Andrey Kutuzov, Lidia Pivovarova#EMNLP2021livetweet #CoNLL2021 https://t.co/M5MPTvB4Fw pic.twitter.com/iEcdVrsiSW
— Leshem Choshen (@LChoshen) November 11, 2021
Stepmothers are mean and academics are pretentious: What do pretrained language models learn about you?
#EMNLP2021 Stepmothers are mean and academics are pretentious: What do pretrained language models learn about you?https://t.co/ARACENSLxr
Rochelle Choenni, Ekaterina Shutova, Robert van Rooi (No twitter?!) pic.twitter.com/lAwnnzI2JT— Leshem Choshen (@LChoshen) November 7, 2021
We Need to Talk About Train-dev-test Splits
#EMNLP2021 We Need to Talk About train-dev-test Splits@robvanderg
Tuning is done on the dev, so people use the test to compare many models and the test sets become unreliable. We should use another split. pic.twitter.com/dwsz4Hyjvl— Leshem Choshen (@LChoshen) November 8, 2021
AVodaDo: Strategy fr Adapting Vocabulary to Downstream Domain
#EMNLP2021 #EMNLP2021livetweet https://t.co/3Y79Hf4xwPhttps://t.co/vn0PU0wyGZ
Jimin Hong, Taehee Kim, Hyesu Lim, Jaegul Choo
AVocaDo: Strategy for Adapting Vocabulary to Downstream Domain
How to change the vocabulary to fit our data (with a pretrained model). pic.twitter.com/oFu1ihn02J— Leshem Choshen (@LChoshen) November 8, 2021
Effects of Parameter Norm Growth During Transformer Training: Inductive Bias from Gradient Descent
#EMNLP2021
William Merrill @yoavgo @royschwartzNLP @RamanujanVivek @nlpnoah https://t.co/MkZGAqcuiohttps://t.co/Uem2c0nixV pic.twitter.com/2xelzI37wr— Leshem Choshen (@LChoshen) November 7, 2021
HypMix: Hyperbolic Interpolative Data Augmentation
#EMNLP2021 #EMNLP2021livetweet https://t.co/HBmHH9qJvv@ramitsawhney Megh Thakkar @AGARvalaAgarwal Di Jin @Diyi_Yang Lucie Flek
HypMix: Hyperbolic Interpolative Data Augmentationhttps://t.co/IwrMoNfg7N pic.twitter.com/Hia0BqveuH— Leshem Choshen (@LChoshen) November 8, 2021
On Homophony and Rényi Entropy
#EMNLP2021 #EMNLP2021livetweet https://t.co/Cl1EQsMyey
On Homophony and Rényi Entropy@tpimentelms @clara__meister Simone Teufel @ryandcotterell pic.twitter.com/MdIDVtjnUd— Leshem Choshen (@LChoshen) November 9, 2021
The Effect of Efficient Messaging and Input Variability on Neural-Agent Iterated Language Learning
#EMNLP2021 #EMNLP2021livetweet https://t.co/G4eb9NIn4W
The Effect of Efficient Messaging and Input Variability on Neural-Agent Iterated Language Learninghttps://t.co/XgxRK7ioqz@ArijRiabi @AriannaBisazza @VerhoefTessa
Emerging language and linguistics,
gotta love this https://t.co/404mruzv1h— Leshem Choshen (@LChoshen) November 9, 2021
Competency Problems: On Finding and Removing Artifacts in Language Data
#EMNLP2021 Competency Problems: On Finding and Removing Artifacts in Language Datahttps://t.co/XpxixDEvnc
William Merrill, @JesseDodge , @nlpmattg , Matthew Peters, @alexisjross, @sameer_, @nlpnoah— Leshem Choshen (@LChoshen) November 7, 2021
Filling the Gaps in Ancient Akkadian Texts: A Masked Language Modelling Approach
#EMNLP2021 #EMNLP2021livetweet @LazarKoren Benny Saret, Asaf Yehudai, Wayne Horowitz, Nathan Wasserman, @GabiStanovsky https://t.co/xf6LZ1VE05 https://t.co/GUlSv6wNx9
— Leshem Choshen (@LChoshen) November 8, 2021
Coarse2Fine: Fine-Grained Text Classification on Coarsely-grained Annotated Data
#EMNLP2021 Coarse2Fine: Fine-grained Text Classification on Coarsely-grained Annotated Data@MekalaDheeraj, @VarunGangal, @shangjingbohttps://t.co/xR953Pga9a
Can we learn fine-grained classes with data on only coarse grained models? pic.twitter.com/nCtD2d3Apd
— Leshem Choshen (@LChoshen) November 7, 2021
Information-theoretic Characterization of Fusion
#EMNLP2021 #EMNLP2021livetweet https://t.co/K8OJ0hQNLjhttps://t.co/KHuMVZ3a25
Information-Theoretic Characterization of Fusion@neil_rathi @OlgaZamaraeva @ethanachiThe first quantitative measure of how morphological fusing a language is.
p.s. Kudo 1st author in high school! pic.twitter.com/jipozLYao2— Leshem Choshen (@LChoshen) November 9, 2021
AligNART: Non-autoregressive Neural Machine Translation by Learning to Estimate Alignment and Translate
#EMNLP2021 AligNART: Non-autoregressive Neural Machine Translation by Jointly Learning to Estimate Alignment and Translate
Jongyoon Song, @sungwon__kim , Sungroh Yoonhttps://t.co/YHrHPvBfYW pic.twitter.com/YBX7yi7PnC— Leshem Choshen (@LChoshen) November 7, 2021
IndoBERTweet: A Pretrained LM for Indonesian Twitter w/ Effective Domain-Specific Vocabulary Initialization
To give you a context of the significance of the following work, here is an article why in Indonesia simultaneously no-one, and everyone, speaks the national language, Bahasa Indonesia:https://t.co/MKdt6HMmCu
1/4#EMNLP2021livetweet #emnlp2021en— Renny P. Kusumawardani (@rennypradina) November 11, 2021
MindCraft: Theory of Mind Modeling for Situated Dialogue in Collaborative Tasks
#EMNLP2021 How do you model what people think they know about what others think? And, how dialogues influence such beliefs in the course of a collaboration? “MindCraft: Theory of Mind Modeling for Situated Dialogue in Collaborative Tasks” aims to answer these.
1/n#emnlp2021en pic.twitter.com/cQYrLzq3Io— Renny P. Kusumawardani (@rennypradina) November 7, 2021
Multimodal Pretraining Unmasked
#EMNLP2021LiveTweet ?
? Multimodal Pretraining Unmasked
✏️ E. Bugliarello (@ebugliarello), R. Cotterell (@ryandcotterell), N. Okazaki (@chokkanorg), D. Elliott (@delliott)
? https://t.co/pQYz4grx61
? https://t.co/bZMM48JfOU
/1 pic.twitter.com/CPgVI6bDit— Max (@mxmeij) November 10, 2021
HypMix: Hyperbolic Interpolative Data Augmentation
#EMNLP2021LiveTweet ?
? HypMix: Hyperbolic Interpolative Data Augmentation
✏️ Ramit Sawhney (@ramitsawhney), Megh Thakkar, Shivam Agarwal (@Shivamag12, Di Jin, Diyi Yang (@Diyi_Yang), Lucie Flek (@lucie_nlp)
? https://t.co/WmEuMlEBHs
? https://t.co/1lk8bGojG6
/1 pic.twitter.com/GFvGxhIHKZ— Max (@mxmeij) November 10, 2021
Vision-and-Language or Vision-for-Language? On Cross-Modal Influence in Multimodal Transformers
#EMNLP2021LiveTweet ?
? Vision-and-Language or Vision-for-Language? On Cross-Modal Influence in Multimodal Transformers
✏️ Stella Frank (@stellalalallets), Emanuele Bugliarello (@ebugliarello), Desmond Elliott (@delliott)
? https://t.co/Wj0OTkN07T
? https://t.co/dAGcjgmmWT
/1 pic.twitter.com/mjbNq4DvNH— Max (@mxmeij) November 10, 2021
COVR: A Test-Bed for Visually Grounded Compositional Generalization with Real Images
#EMNLP2021LiveTweet ?
? COVR: A Test-Bed for Visually Grounded Compositional Generalization with Real Images
✏️ Ben Bogin (@ben_bogin), Shivanshu Gupta, Matt Gardner (@nlpmattg), Jonathan Berant (@JonathanBerant)
? https://t.co/FTB5z72v8s
? https://t.co/uOrRMc8lLn
/1 pic.twitter.com/8hq5YO7LQe— Max (@mxmeij) November 9, 2021
On Pursuit of Designing Multi-model Transformer for Video Grounding
#EMNLP2021LiveTweet ?
? On Pursuit of Designing Multi-modal Transformer for Video Grounding
✏️ Meng Cao, Long Chen, Mike Zheng Shou, Can Zhang, Yuexian Zou
? https://t.co/r3lslpe5kH
? https://t.co/XpKL4zBO1I
/1 pic.twitter.com/LvGlxqR0RT— Max (@mxmeij) November 9, 2021
Inflate and Shrink: Enriching and Reducing Interactions for Fast Text-Image Retrieval
#EMNLP2021LiveTweet ?
? Inflate and Shrink: Enriching and Reducing Interactions for Fast Text-Image Retrieval
✏️ Haoliang Liu, Tan Yu (@taoyds), Ping Li
? https://t.co/edKFuvLo4n
? https://t.co/sghqJCyp3Y
/1 pic.twitter.com/XO2CqjwtuK— Max (@mxmeij) November 9, 2021
Robust Open-Vocabulary Translation from Visual Text Representations
#EMNLP2021LiveTweet ?
? Robust Open-Vocabulary Translation from Visual Text Representations
✏️ Elizabeth Salesky (@esalesk), David Etter, Matt Post (@mjpost)
? https://t.co/yMSTxXajrM
? https://t.co/fNaLOptVEv
/1 pic.twitter.com/nTJnlUMZZ8— Max (@mxmeij) November 8, 2021
Boosting Cross-lingual Transfer via Self-learning with Uncertainty Estimation
#EMNLP2021LiveTweet ?
? Boosting Cross-Lingual Transfer via Self-Learning with Uncertainty Estimation
✏️ Liyan Xu, Xuchao Zhang, Xujiang Zhao, Haifeng Chen, Feng Chen, Jinho D. Choi (@Jinho_D_Choi)
? https://t.co/JwxdHKwGjz
? https://t.co/ehrt2bAxlp
/1 pic.twitter.com/CHdKDJsej0— Max (@mxmeij) November 8, 2021
It Is Not As Good As You Think!
#EMNLP2021LiveTweet ?
? It Is Not As Good As You Think!
✏️ J. Zhao (@michellezhaozh1), P. Arthur (@_philip_arthur), G. Haffari (@haffari), T. Cohn (@trevorcohn), E. Shareghi (@EhsanShareghi)
? https://t.co/ta9id1iENw
? https://t.co/CRG5OKh66g
/1 pic.twitter.com/IN1aLpky6C— Max (@mxmeij) November 8, 2021
A Generative Framework for Simultaneous Machine Translation
#EMNLP2021LiveTweet ?
? A Generative Framework for Simultaneous Machine Translation
✏️ Yishu Miao, Phil Blunsom, Lucia Specia (@lspecia)
? https://t.co/OHjq2rAKiH
? https://t.co/27ujMOgf33
/1 pic.twitter.com/OPC5ksdPKA— Max (@mxmeij) November 8, 2021
Controlling Machine Translation for Multiple Attributes with Additive Interventions
#EMNLP2021LiveTweet ?
? Controlling Machine Translation for Multiple Attributes with Additive Interventions
✏️ Andrea Schioppa, Katja Filippova (@fajtak), Artem Sokolov, David Vilar
? https://t.co/3dAzLA8zPx
? https://t.co/c2qLWUOoKs
/1 pic.twitter.com/EUl4L9s0MW— Max (@mxmeij) November 8, 2021
BERT, mBERT, or BiBERT? A Study on Contextualized Embeddings for Neural Machine Translation
#EMNLP2021LiveTweet ?
? BERT, mBERT, or BiBERT? A Study on Contextualized Embeddings for Neural Machine Translation
✏️ Haoran Xu (@fe1ixxu), Benjamin Van Durme (@ben_vandurme), Kenton Murray (@kentonmurray)
? https://t.co/TcH8ghZpsM
? https://t.co/LqfEtftWGj
/1 pic.twitter.com/iCaag8tLzL— Max (@mxmeij) November 8, 2021
Multilingual Unsupervised Neural Machine Translation with Denoising Adapters
#EMNLP2021LiveTweet ?
? Multilingual Unsupervised Neural Machine Translation with Denoising Adapters
✏️ Ahmet Üstün (@ahmetustun89), Alexandre Berard , Laurent Besacier (@laurent_besacie), Matthias Gallé (@mgalle)
? https://t.co/9yVc5C0wHw
? https://t.co/9OPlZnKGLU
/1 pic.twitter.com/M32JF4JauN— Max (@mxmeij) November 8, 2021
Indonesian livetweets
Disentangling Representations of Text by Masking Transformers
#EMNLP2021 "Disentangling Representations of Text by Masking Transformers", by Xiongyi Zhang, Jan-Willem van de Meent, and Byron Wallace https://t.co/1vsqMJeSCw
— Iftitahu Nimah (@IftitahuNimah) November 7, 2021
Aligning Faithful Interpretations with their Social Attributes
#EMNLP2021 "Aligning Faithful Interpretations with their Social Attribution", by Alon Jacovi and Yoav Goldberg https://t.co/Q8ZjF6e3SP
— Iftitahu Nimah (@IftitahuNimah) November 7, 2021
How Does Counterfactually Augmented Data Impact Models for Social Computing Constructs?
#EMNLP2021 "How Does Counterfactually Augmented Data Impact Models for Social Computing Constructs?", by Indira Sen et al. https://t.co/QqVlsl8qfL
— Iftitahu Nimah (@IftitahuNimah) November 7, 2021
Idiosyncratic but not Arbitrary: Learning Idiolects in Online Registers Reveals Distinctive yet Consistent Individual Styles
#EMNLP2021 "Idiosyncratic but not Arbitrary: Learning Idiolects in Online Registers Reveals Distinctive yet Consistent Individual Styles", by Zhu and Jurgens https://t.co/wVTc3CTaHa
— Iftitahu Nimah (@IftitahuNimah) November 7, 2021
Multi-domain Multilingual Question Answering
Tertarik dengan Question Answering System?
Sedang berlangsung tutorial tentang Multi-Domain Multilingual Question Answering by @seb_ruder dan @aviaviavi__ https://t.co/2ANGZe8eve#EMNLP2021livetweet #emnlp2021id pic.twitter.com/7JNBWTL4UD
— Said al faraby (@SaidFaraby) November 11, 2021
IndoNLI: A Natural Language Inference Dataset for Indonesian
“IndoNLI: A Natural Language Inference Dataset for Indonesian” oleh @rmahendrarm, @AlhamFikri, @samuel_louvan, Fahrurrozi Rahman, & @claravania menyajikan dataset inferensi bahasa natural (NLI) pertama Bahasa Indonesia yg disusun oleh manusia.
1/4#EMNLP2021livetweet #emnlp2021id— Renny P. Kusumawardani (@rennypradina) November 11, 2021
MindCraft: Theory of Mind Modelling for Situated Dialogue in Collaborative Tasks
#EMNLP2021 Bgmn memodelkan apa yg orang pikirkan tentang pemikiran orang lain? Bgmn percakapan mempengaruhi pemikiran tsb di sepanjang kerjasama? “MindCraft: Theory of Mind Modeling for Situated Dialogue in Collaborative Tasks” berusaha menjawab pertanyaan2 tsb.
1/n#emnlp2021id pic.twitter.com/K3OBpFTubi— Renny P. Kusumawardani (@rennypradina) November 7, 2021
UNKs Everywhere: Adapting Multilingual Language Models to New Scripts
#EMNLP2021 #EMNLP2021livetweet "UNKs Everywhere: Adapting Multilingual Language Models to New Scripts", by Jonas Pfeiffer et al. https://t.co/SazkNKfVpf
— Iftitahu Nimah (@IftitahuNimah) November 9, 2021
Finally, here is an interesting selection by Iftitahu Nimah
List of #EMNLP2021 keynote talks + papers that I've been binge-watching during conference to get a quick grasp about the paper's main message (only few of them are live-tweeted):
/1
— Iftitahu Nimah (@IftitahuNimah) November 12, 2021
AIhub is supported by:
