Top Tweets for #MTWeekly
#MTWeekly 1⃣0⃣0⃣ https://t.co/U59mw1PeXx discusses new multilingual multimodal benchmark IGLUE 🥶 https://t.co/iAFC648C8r Quite surprising to see how bad the zero-shot xling transfer work in multimodal tasks. 🤷♂️ But it also means big room for improvements 📈
#MTWeekly 9⃣9⃣ https://t.co/6s4OQOgfOL reviews a recent report by @DeepMind on how evil 🦹♂️language models can be. In my post, I try to speculate on how things change when we consider multilingual models. 🌍🌏🌎
#MTWeekly 9⃣7⃣ https://t.co/bzLqRQ5suI comments on a pre-print on non-autoregressive multlingual MT https://t.co/iYk4sUirdo by @swetaagrawal20, @KreutzerJulia and @ColinCherry. It does not as good as I would hope 🤷♂️, but still good to know. 😉
In #MTWeekly 9⃣6⃣ https://t.co/DqADS25tWD, I share some ideas on the evaluation of non-autoregressive MT systems. 📐📏 If you are now reviewing a paper on non-autoregressive MT for @ReviewAcl, please give a thought to my comments... 🤓
#MTWeekly 9⃣5⃣ on Minimum Bayes Risk decoding https://t.co/kdKpKzxt5L, a paper by @markuseful and others at @GoogleAI https://t.co/xkMTK31qP2. TLDR: the better metric, the better MBR gets 📈. And there is obviously something wrong with the probability in encoder-decoder models.
#MTWeekly 9⃣1⃣ https://t.co/YkJekbhpez a paper on a recipe 🍳👨🍳 for zero-shot MT when the only things you have at home are: a pre-trained multilingual sentence representation, parallel data in several languages, and some computation resources to spare. https://t.co/JSXkqpLtt6
How comes large monolingual language models are surprisingly multilingual? I speculate and conspire on that in #MTWeekly 9⃣0⃣ https://t.co/WSKWirjjt6
#MTWeekly 8⃣9⃣ https://t.co/9Wm9TqTXk4 discusses a paper on BPE vocabulary size and memorization effects https://t.co/FQGICH8mO8 by @facebookai. It is not only parameter count but also sequence length that matters for Transformers. 🤔
In #MTWeekly 8⃣6⃣ https://t.co/ioTRINlpiG I show some stats on WMT submissions from 2018 to 2020. Nothing surprising there, but it is nice to see the stuff tabulated. 🤓📊
An outstanding paper from this year's ACL https://t.co/h7dRfeAOuU shows how low the evaluation standards in MT are and that it is not getting much better over time. In #MTWeekly 85, I speculate what the reasons might be. https://t.co/HZKXP83Rha
In #MTWeekly 8⃣4⃣ https://t.co/vSGnjjDxW8 I comment on a paper on non-autoregressive MT https://t.co/dxmRjPy9Oj that shows that separate losses for fluency and accuracy might work well even in the neural times.
#MTWeekly 7⃣8⃣ https://t.co/JBH5JMLmXS on how multilingual hate speech detections reveals how little cultural-neutral multilingual BERT is. I comment on paper https://t.co/O7IaWtMehc
#mtweekly 75 https://t.co/TalK7mQRKx on outbound translation. If you want users to trust your MT system, just show them the back-translation. Paper: https://t.co/UC023atBt1 by @zouharvi and the @BergamotProject #NLProc
#MTWeekly https://t.co/lDdmECglab about three architecture innovation that I think we will hear a lot about in the next year in #neuralempty and #NLProc. (Or in two years, I can laugh about how wrong my prediction was.)
#MTWeekly https://t.co/sPAE5i7mnz on non-autoregressive MT with latent classes. Why model relations between thousands of words, when 64 latent classes are enough? Paper: https://t.co/IUBnu2oawM
#MTWeekly https://t.co/hLlf9OtjWz on a trick how to change the source language of MT without parallel data. Paper https://t.co/8f5Z2zU3Mg by @surafelml, @negri_teo and @Turchi_Marco.
Last Seen Hashtags on Sotwe
BuyfromGroovy
Seen from Indonesia
monkeyappgirls
mature
Seen from Turkey
cnc
Seen from United Kingdom
คลิปหลุดในกลุ่ม
Seen from Thailand
حواي
Seen from United States
zentreart
Seen from United States
Gecekulubu
Seen from Turkey
incest() +filter:native_video
Seen from Netherlands
nsfwtwtً
Seen from Argentina
Most Popular Users
Elon Musk 
@elonmusk
240.3M followers
Barack Obama
@barackobama
119.2M followers
Donald J. Trump
@realdonaldtrump
111.6M followers
Cristiano Ronaldo
@cristiano
109.7M followers
Narendra Modi
@narendramodi
106.9M followers
Rihanna
@rihanna
97.5M followers
NASA
@nasa
92.1M followers
Justin Bieber
@justinbieber
90.7M followers
KATY PERRY
@katyperry
87.2M followers
Taylor Swift
@taylorswift13
81M followers
Lady Gaga
@ladygaga
72.6M followers
Kim Kardashian
@kimkardashian
69.6M followers
Virat Kohli
@imvkohli
69.2M followers
YouTube
@youtube
68.6M followers
Bill Gates
@billgates
63.6M followers
The Ellen Show
@theellenshow
62.5M followers
Neymar Jr
@neymarjr
61.9M followers
CNN
@cnn
61.9M followers
X
@x
60.9M followers
Selena Gomez
@selenagomez
60.3M followers