ΑΙhub.org
Microblogs from ACL 2021
The Joint Conference of the 59th annual meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (ACL-IJCNLP 2021) took place in August. We gathered the microblogged papers in different languages and sessions. We hope that you find this selection helpful!
English Microblogs
Computational Social Science and Ethics in NLP
Here is my first microblog post for #ACL2021NLP: "Structured Sentiment Analysis as Dependency Graph Parsing" by Jeremy Barnes (presenting 🎙️), Robin Kurtz, Stephan Oepen, Lilja Øvrelid and Erik Velldal #ACL2021EN (1/9) 🧵
— Agostina Calabrese (@agostina_cal) August 1, 2021
Here is my 2nd microblog post for #ACL2021NLP: "Understanding and Countering Stereotypes: A Computational Approach to the Stereotype Content Model" by Kathleen C. Fraser (presenting 🎙️), Isar Nejadgholi and Svetlana Kiritchenko #ACL2021EN (1/10) 🧵
— Agostina Calabrese (@agostina_cal) August 1, 2021
Here is my 3rd microblog post for #ACL2021NLP: "Automatic Fake News Detection: Are Models Learning to Reason?" by Casper Hansen (presenting 🎙️), Christian Hansen and Lucas Chaves Lima #ACL2021EN (1/8) 🧵
— Agostina Calabrese (@agostina_cal) August 2, 2021
Here is my 4th microblog post for #ACL2021NLP: "Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detection" by Bertie Vidgen (presenting 🎙️), Tristan Thrush, Zeerak Waseem and Douwe Kiela #ACL2021EN (1/10) 🧵
— Agostina Calabrese (@agostina_cal) August 2, 2021
Here is my 5th microblog post for #ACL2021NLP: "Detecting Harmful Memes and Their Targets" by Shraman Pramanick, Dimitar Dimitrov, Rituparna Mukherjee, Shivam Sharma (presenting 🎙️), Md. Shad Akhtar, Preslav Nakov and Tanmoy Chakraborty #ACL2021EN (1/9) 🧵
— Agostina Calabrese (@agostina_cal) August 3, 2021
Here is my 6th microblog post for #ACL2021NLP: '“I’ve Seen Things You People Wouldn’t Believe”: Hallucinating Entities in GuessWhat?!' by Alberto Testoni (@alberto_testoni, presenting 🎙️) and Raffaella Bernardi (@raffagbernardi) #ACL2021EN (1/9) 🧵
— Agostina Calabrese (@agostina_cal) August 3, 2021
Here is my 7th microblog post for #ACL2021NLP: "Meta-Learning to Compositionally Generalize" by Henry Conklin (@henryconklin, presenting 🎙️), Bailin Wang, Kenny Smith and Ivan Titov #ACL2021EN (1/10) 🧵
— Agostina Calabrese (@agostina_cal) August 3, 2021
Here is my 9th microblog post for #ACL2021NLP: "Factorising Meaning and Form for Intent-Preserving Paraphrasing" by Tom Hosking (@tomhosking, presenting 🎙️) and Mirella Lapata #ACL2021EN (1/10) 🧵
— Agostina Calabrese (@agostina_cal) August 5, 2021
We need better models to detect online hate speech – and building them is possible!
P. Röttger, B. Vidgen, D. Nguyen, Z. Waseem, H. Margetts, J. Pierrehumbert. HateCheck: Functional Tests for Hate Speech Detection Models
Paper: https://t.co/DQSmZ4RlpH
1/11#ACL2021NLP #ACL2021en— NoOverfitting Lab (@nooverfitting) August 2, 2021
#ACL2021NLP #acl2021en Towards Knowledge-Grounded Counter Narrative Generation for Hate Speech by Yi-Ling Chung, Serra Sinem Tekiro˘glu, and Marco Guerini.
ACL Anthology Link: https://t.co/sJkmsUAa9x
— Sourav Das (@SouravDme) August 3, 2021
First micro-blog for #ACL2021NLP Ethics in NLP (session:4E)
"Bad Seeds: Evaluating Lexical Methods for Bias Measurement" by @maria_antoniak and @dmimno #ACL2021EN
Paper: https://t.co/N5tGLXM8vj
Presentation: https://t.co/Gbt1DwOKNN 1/n— Nedjma Ousidhoum نجمة أوسيدهم (@nedjmaou) August 2, 2021
#ACL2021NLP Ethics in NLP (Session: 3I)
"Exposing the limits of Zero-shot Cross-lingual Hate Speech Detection" by @debora_nozza .
Paper: https://t.co/AJHEv5fqyf
Presentation: https://t.co/S1SyHt0E7D #ACL2021EN 1/7— Nedjma Ousidhoum نجمة أوسيدهم (@nedjmaou) August 3, 2021
#ACL2021NLP (Session 2K: COMPUTATIONAL SOCIAL SCIENCE AND CULTURAL ANALYSIS)
"Claim Matching Beyond English to Scale Global Fact-Checking" by @ashkankazemi @gvrkiran @DGaff_ @computermacgyve
Paper https://t.co/UEVhN7PtRs
Presentation https://t.co/Vl3YNDv8az #ACL2021EN 1/n— Nedjma Ousidhoum نجمة أوسيدهم (@nedjmaou) August 5, 2021
#ACL2021NLP Session:4E Ethics in NLP
"Men Are Elected, Women Are Married: Events Gender Bias on Wikipedia"
by @sunjiao123sun_ and @VioletNPeng
Paper https://t.co/wt9IwUGKIo
Presentation https://t.co/tgOaXjvOmZ#ACL2021EN 1/n— Nedjma Ousidhoum نجمة أوسيدهم (@nedjmaou) August 5, 2021
#ACL2021NLP Ethics in NLP (Session 4E)
"RedditBias: A Real-World Resource for Bias Evaluation and Debiasing of Conversational Language Models" by Soumya Barikeri, @anne_lauscher, @licwu, @gg42554
Paper: https://t.co/EY2Phi37JY
Presentation: https://t.co/5SNSgoGBW3 #ACL2021EN 1/8— Nedjma Ousidhoum نجمة أوسيدهم (@nedjmaou) August 2, 2021
Resources and evaluation and NLP Applications
How to shuffle text for "real coherence".
Ph.Laban, L.Dai, L.Bandarkar, M.A.Hearst. Can Transformer Models Measure Coherence In Text: Re-Thinking the Shuffle Test
Paper: https://t.co/Zs76svLKc5
/1#ACL2021NLP #ACL2021en— Jekaterina Novikova (@JNov21602962) August 4, 2021
Overcoming the problem of small data in suicide detection models
C.Yang, Y.Zhang, S.Muresan. Weakly-Supervised Methods for Suicide Risk Assessment: Role of Related Domains
Paper: https://t.co/JNjl9q4Atn
1/6#ACL2021NLP #ACL2021en— NoOverfitting Lab (@nooverfitting) August 3, 2021
Do we really need to aggregate the average scores of multiple systems on test instances to compare them properly?
M. Peyrard, W. Zhao, S. Eger, R. West. Better than Average: Paired Evaluation of NLP systems
Paper: https://t.co/B8ToZTYiSl
1/7#ACL2021NLP #ACL2021en— NoOverfitting Lab (@nooverfitting) August 2, 2021
Assessing the Representations of "Idiomaticity in Vector Models with a Noun Compound Dataset Labeled at Type and Token Levels" by Marcos Garcia, Marco Idiart, Carolina Scarton, Tiago Kramer Viera, Aline Villavicencio. 1/7 #ACL2021NLP #ACL2021EN
— William Chen (@chenwanch1) August 3, 2021
#ACL2021NLP Microblogging
Margherita Fanton, Helena Bonaldi, Serra Sinem Tekiroglu, Marco Guerini: "Human-in-the-Loop for Data Collection: a Multi-Target Counter Narrative Dataset to Fight Online Hate Speech"
QA session in 1 hour.@helena_bonaldi @serrasinem @m_guerini
(1/n)— Mubashara Akhtar (@akhtarmubashara) August 3, 2021
#ACL2021NLP Microblogging
Nikita Nangia, Saku Sugawara, Harsh Trivedi, Alex Warstadt, Clara Vania, Samuel R. Bowman:
"What Ingredients Make for an Effective Crowdsourcing Protocol for Difficult NLU Data Collection Tasks?"@meloncholist @a_stadt @claravania @sleepinyourhat
1/n— Mubashara Akhtar (@akhtarmubashara) August 3, 2021
Interpretability and Analysis of Models for NLP
How do we incorporate faithfulness into contextual embeddings representing causal graphs? A thread on “Learning Faithful Representations of Causal Graphs” by Ananth Balashankar & Lakshminarayanan Subramanian.
1/#ACL2021NLP #acl2021en— Renny P. Kusumawardani @KIMLI 2021 (@rennypradina) August 2, 2021
How do we effectively extend the structural probing of Hewitt and Manning (2019) to study more linguistic phenomena in LMs? A thread on “Introducing Orthogonal Constraint in Structural Probes” by Tomasz Limisiewicz and David Mareček.
1/7#ACL2021NLP #acl2021en pic.twitter.com/EZqeTJkbix— Renny P. Kusumawardani @KIMLI 2021 (@rennypradina) August 2, 2021
While previous approaches in probing can be thought of as attempting to understand the representation of sentences in LMs, Li et. al (2021) aims to understand whether these LMs contain a representation of the situation described by a discourse.
1/10#ACL2021NLP #acl2021en— Renny P. Kusumawardani @KIMLI 2021 (@rennypradina) August 4, 2021
#ACL2021NLP #ACL2021EN Starting the first microblog for @aclmeeting with the Findings paper-
Nadir Durrani, Hassan Sajjad, and Fahim Dalvi: How transfer learning impacts linguistic knowledge in deep NLP models? Presenter-Nadir Durrani (ND).
Paper@ https://t.co/bwUK5p3MIT
[1/6]— Shaily Bhatt (@shaily99) August 1, 2021
Dialog and interactive Systems
#ACL2021NLP #ACL2021EN Starting the first microblog for @aclmeeting with the Findings paper-
Dinesh Raghu, Atishya Jain, Mausam, and Sachindra Joshi: Constraint based Knowledge Base Distillation
in End-to-End Task Oriented Dialogs,
Paper@ https://t.co/Yh0bDFTTuN
[1/n]— Anant Khandelwal (@anantwall) August 2, 2021
Have you ever felt frustrated by the too-obviously-template responses of “I don’t know” given by dialogue systems? Shrivastava et al. offer a simple two-steps approach to generate better, more engaging, fallback responses by contextualizing them.
1/#ACL2021NLP #acl2021en pic.twitter.com/MnYDUY1Wu5— Renny P. Kusumawardani @KIMLI 2021 (@rennypradina) August 2, 2021
In a real-world setting, a dialogue system would inevitably encounter questions outside its scope. Successfully recognizing such out-of-domain (OOD) queries is thus essential. But what can you do if all you have is in-domain (IND) data?
1/7#ACL2021NLP #acl2021en— Renny P. Kusumawardani @KIMLI 2021 (@rennypradina) August 3, 2021
Q.Liu, L.Yu, L.Rimell, Ph.Blunsom. Pretraining the Noisy Channel Model for Task-Oriented Dialogue
Paper: https://t.co/J2eHwDutrR
/1#ACL2021NLP #ACL2021en— Jekaterina Novikova (@JNov21602962) August 4, 2021
Can you trust your dialogue system on its advice for feeding your dog?
H. Rashkin, D. Reitter, G. S. Tomar, D. Das. Increasing Faithfulness in Knowledge-Grounded Dialogue with Controllable Features
Paper: https://t.co/m1CO87frtt
1/10#ACL2021NLP #ACL2021en pic.twitter.com/fj92pdSvx4— Jekaterina Novikova (@JNov21602962) August 1, 2021
How to automatically evaluate the coherence of a dialogue system response wrt a context?https://t.co/DVYsU4Qq09, https://t.co/zA03c1NgYo, L.Huang, L.Lin, X.Liang. Towards Quantifiable Dialogue Coherence Evaluation
Paper: https://t.co/uK6U8mYYBW
1/5#ACL2021NLP #ACL2021en— Jekaterina Novikova (@JNov21602962) August 3, 2021
How to define and carefully test robustness.
J.Liu, R.Takanobu, J.Wen, D.Wan, https://t.co/pJ1SUPbTuy, W.Nie, https://t.co/4JSJpYnieE, W.Peng, M.Huang. Robustness Testing of Language Understanding in Task-Oriented Dialog
Paper: https://t.co/veUVHEIelt
1/6#ACL2021NLP #ACL2021en— Jekaterina Novikova (@JNov21602962) August 3, 2021
Machine Translation, QA, and Machine Learning for NLP
#ACL2021NLP #ACL2021EN #NLProc #multilinguality #translation #codeswitching #textgeneration
Ishan Tarunesh (presenter), Syamantak Kumar, Preethi Jyothi: From Machine Translation to Code-Switching: Generating High-Quality Code-Switched Text. Paper at https://t.co/hFOBQJ36Lf
[1/7]— Shaily Bhatt (@shaily99) August 2, 2021
#ACL2021NLP #ACL2021EN Kanwatchara, Horsuwan: Rational LAMOL: A Rationale-Based Lifelong Learning Framework#ACL2021NLP
— Yohannes Biadgligne Ejigu (@BiadgligneEjigu) August 1, 2021
This thread is about "Diverse Pretrained Context Encodings Improve Document Translation" by Domenic Donato, Lei Yu, Chris Dyer. 1/5#ACL2021NLP #ACL2021EN #ACL2021FRhttps://t.co/NUGClRjugv pic.twitter.com/B4YpbcE1Ak
— Kayo Yin (@kayo_yin) August 2, 2021
This thread is about "Modeling Bilingual Conversational Characteristics for Neural Chat Translation" by Yunlong Liang, Fandong Meng, Yufeng Chen, Jinan Xu, Jie Zhou. 1/4#ACL2021NLP #ACL2021EN #ACL2021FRhttps://t.co/KjrJtRk2ph pic.twitter.com/MHQefFMfKD
— Kayo Yin (@kayo_yin) August 3, 2021
#ACL2021NLP (Session 14E: QUESTION ANSWERING #4)
"Question Answering Over Temporal Knowledge Graphs"
by Apoorv Saxena, Soumen Chakrabarti, Partha Talukdar
Paper: https://t.co/W0GfTMoIUb
Presentation: https://t.co/XB2MPsuu98 #ACL2021EN 1/n— Nedjma Ousidhoum نجمة أوسيدهم (@nedjmaou) August 4, 2021
#ACL2021NLP (Session12C: QUESTION ANSWERING #3)
"Towards more equitable question answering systems: How much more data do you need?" by Arnab Debnath, Navid Rajabi, Fardina Fathmiul Alam, @anas_ant
Paperhttps://t.co/DJ77jDnseL
Presentationhttps://t.co/3CT6Sz2WzX #ACL2021EN 1/n— Nedjma Ousidhoum نجمة أوسيدهم (@nedjmaou) August 3, 2021
Indonesian microblogs
Apakah ada alternatif lain untuk menginterpretasi Inter-rater Reliability?
"Cross-replication Reliability – An Empirical Approach to Interpreting Inter-rater Reliability" oleh Ka Wong, Praveen Paritosh, Lora Aroyo. https://t.co/mhGa9iQNbr#ACL2021NLP #ACL2021ID (1/n)
— Rifki Afina Putri (@rifkiaputri) August 4, 2021
He is very intelligent, she is very beautiful? 🤔
Saya akan mengulas makalah berjudul "He is very intelligent, she is very beautiful? On Mitigating Social Biases in Language Modelling and Generation" oleh Aparna Garimella, … (cont)#ACL2021NLP #ACL2021ID (1/n)
— Rifki Afina Putri (@rifkiaputri) August 3, 2021
🚨 New Indonesian Dataset Alert 🚨
Kali ini saya akan mengulas makalah "IndoCollex: A Testbed for Morphological Transformation of Indonesian Word Colloquialism" oleh Haryo Akbarianto Wibowo, Made Nindyatama Nityasya, Afra Feyza Akyürek, … (cont)#ACL2021NLP #ACL2021ID (1/n)
— Rifki Afina Putri (@rifkiaputri) August 3, 2021
Berikut ulasan utk makalah "What Ingredients Make for an Effective Crowdsourcing Protocol for Difficult NLU Data Collection Tasks?" oleh Nikita Nangia, Saku Sugawara, Harsh Trivedi, Alex Warstadt, Clara Vania, Samuel R Bowman. https://t.co/gqEJnSPQum#ACL2021NLP #ACL2021ID (1/n)
— Rifki Afina Putri (@rifkiaputri) August 2, 2021
Saya akan mengulas makalah berjudul "What’s in the Box? An Analysis of Undesirable Content in the Common Crawl Corpus" oleh Alexandra Luccioni dan Joseph Viviano.
Makalah dapat diakses di: https://t.co/Aftyh3dbJb#ACL2021NLP #ACL2021ID (1/n)
— Rifki Afina Putri (@rifkiaputri) August 2, 2021
Jika pendekatan2 sebelumnya thd probing dpt dipandang sbg usaha utk memahami representasi kalimat pd LM, Li dkk. (2021) bertujuan utk memahami apakah LM mengandung representasi dr sebuah situasi yg digambarkan oleh sebuah percakapan.
1/10#ACL2021NLP #acl2021id— Renny P. Kusumawardani @KIMLI 2021 (@rennypradina) August 5, 2021
Bagaimana mengintegrasikan konsep faithfulness (kesetiaan, ketepatan, kebenaran) ke dlm embeddings kontekstual representasi graf kausal?
🧵tentang “Learning Faithful Representations of Causal Graphs”, Ananth Balashankar & Lakshminarayanan Subramanian.
1/9#ACL2021NLP #acl2021id— Renny P. Kusumawardani @KIMLI 2021 (@rennypradina) August 5, 2021
Pada kondisi sehari-hari, sebuah sistem dialog pasti akan menjumpai pertanyaan di luar domainnya. Krn itu, adalah penting utk dpt mengenali pertanyaan out-of-domain (OOD). Namun, bagaimana jika yg tersedia hanyalah data in-domain (IND)?
1/7#ACL2021NLP #acl2021id— Renny P. Kusumawardani @KIMLI 2021 (@rennypradina) August 5, 2021
Pernah kesal dengan jawaban “tidak tahu” standar ala sistem dialog? Shrivastava dkk. menawarkan pendekatan dua-langkah sederhana untuk menghasilkan respon yang lebih baik, lebih menarik, dan sesuai konteks.
1/#ACL2021NLP #acl2021id pic.twitter.com/VJMwYZc3LQ— Renny P. Kusumawardani @KIMLI 2021 (@rennypradina) August 5, 2021
“Introducing Orthogonal Constraint in Structural Probes” oleh Tomasz Limisiewicz dan David Mareček membahas perluasan teknik Structural Probing oleh Hewitt dan Manning untuk mempelajari lebih banyak fenomena linguistik pada model bahasa (LM).
1/7#ACL2021NLP #acl2021id pic.twitter.com/eG6ARs462d— Renny P. Kusumawardani @KIMLI 2021 (@rennypradina) August 5, 2021
Italian microblogs
Questo è il mio primo microblog post per #ACL2021NLP: "Structured Sentiment Analysis as Dependency Graph Parsing" di Jeremy Barnes (🎙️), Robin Kurtz, Stephan Oepen, Lilja Øvrelid e Erik Velldal #ACL2021IT (1/9) 🧵
— Agostina Calabrese (@agostina_cal) August 1, 2021
Questo è il mio 3° microblog post per #ACL2021NLP: "Automatic Fake News Detection: Are Models Learning to Reason?" di Casper Hansen (🎙️), Christian Hansen e Lucas Chaves Lima #ACL2021IT (1/8) 🧵
— Agostina Calabrese (@agostina_cal) August 2, 2021
Questo è il mio 4° microblog post per #ACL2021NLP: "Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detection" di Bertie Vidgen (🎙️), Tristan Thrush, Zeerak Waseem e Douwe Kiela #ACL2021IT (1/10) 🧵
— Agostina Calabrese (@agostina_cal) August 2, 2021
Questo è il mio 5° microblog post per #ACL2021NLP: "Detecting Harmful Memes and Their Targets" di Shraman Pramanick, Dimitar Dimitrov, Rituparna Mukherjee, Shivam Sharma (🎙️), Md. Shad Akhtar, Preslav Nakov e Tanmoy Chakraborty #ACL2021IT (1/9) 🧵
— Agostina Calabrese (@agostina_cal) August 3, 2021
Questo è il mio 8° microblog post per #ACL2021NLP: "Exposing the limits of Zero-shot Cross-lingual Hate Speech Detection" di Debora Nozza (@debora_nozza, 🎙️) #ACL2021IT
English microblog post by @nedjmaou at: https://t.co/gd5lGsLlBq
(1/8) 🧵— Agostina Calabrese (@agostina_cal) August 4, 2021
French microblogs
"Bad Seeds: Evaluating Lexical Methods for Bias Measurement" par@maria_antoniak
et@dmimno
#ACL2021NLP (session : 4E sur l'Ethique en TAL.)
#ACL2021FR
Article : https://t.co/N5tGLXM8vj
Présentation : https://t.co/Gbt1DwOKNN 1/12— Nedjma Ousidhoum نجمة أوسيدهم (@nedjmaou) August 2, 2021
#ACL2021NLP L'éthique en TAL (Session : 3I)
"Exposing the limits of Zero-shot Cross-lingual Hate Speech Detection" par@debora_nozza
Article: https://t.co/AJHEv5fqyf
Présentation: https://t.co/S1SyHt0E7D #ACL2021FR 1/n— Nedjma Ousidhoum نجمة أوسيدهم (@nedjmaou) August 3, 2021
#ACL2021NLP L'Ethique en TAL(Session 4E)
"RedditBias: A Real-World Resource for Bias Evaluation and Debiasing of Conversational Language Models" par Soumya Barikeri,@anne_lauscher,@licwu,@gg42554
Article https://t.co/EY2Phi37JYhttps://t.co/5SNSgoGBW3 #ACL2021FR 1/9— Nedjma Ousidhoum نجمة أوسيدهم (@nedjmaou) August 2, 2021
Arabic microblogs
الأخلاقيات في معالجة اللغة الطبيعية (الجلسة: 4E)#ACL2021NLP
"Bad Seeds: Evaluating Lexical Methods for Bias Measurement" @maria_antoniak و @dmimno
#ACL2021AR
الورقة البحثية: https://t.co/N5tGLXM8vj
العرض: https://t.co/Gbt1DwOKNN 1/n— Nedjma Ousidhoum نجمة أوسيدهم (@nedjmaou) August 2, 2021
الأخلاقيات في معالجة اللغة الطبيعية (الجلسة: 4E)#ACL2021NLP
"RedditBias: A Real-World Resource for Bias Evaluation and Debiasing of Conversational Language Models" Soumya Barikeri, @anne_lauscher, @licwu, @gg42554https://t.co/EY2Phi37JYhttps://t.co/5SNSgoGBW3 #ACL2021AR 1/8— Nedjma Ousidhoum نجمة أوسيدهم (@nedjmaou) August 2, 2021
AIhub is supported by:
