Himanshu Maurya

[Rachael Tatman @[email protected] @rctatman]

I've noticed a slight uptick recently of folks reaching out to me asking for help with NLP problems where the underlying problem is one I consider unsolvable (at least practically). 😣

Here are some questions you to ask yourself *before* you invest a lot of time.

[Graham Neubig @gneubig]

As an experienced NLP researcher, I think this is really good advice that anyone getting into the field should know.

Also as an experienced NLP researcher, problems that seem unsolvable and have no good data or evaluation are maybe the most fun 😃

[Himanshu Maurya @lordzuko2]

@johnhewtt I really enjoyed your talk and learnt something new, thanks !

[Science Magazine @ScienceMagazine]

“Requiring students to ante up conference funds ... without the hope of being reimbursed for months makes academia less welcoming for scientists who are financially disadvantaged.” This week’s Working Life. https://t.co/DdteD6wNTi

[Himanshu Maurya @lordzuko2]

@sjmielke @adveisner Thankyou @sjmielke

[Adina Williams @adinamwilliams]

Loving this pocket of papers on theoretical ling. & learning approaches!

@Joe_Pater: https://t.co/8XWs7aXCki (orig.)

@tallinzen: https://t.co/u2Xzn14Bi8 (psycholing)

@ChrisGPotts: https://t.co/WtwG3C2LjP (semantics)

Rawski & Heinz: https://t.co/nY1putPS97 (learning theory)

[Jeremy Howard @jeremyphoward]

The "Imagenet in 1 hour" paper from @priy2201 is overflowing with wonderful little tricks. Here's one of my favorites - something that many researchers are still not aware of, but really makes model training easier
https://t.co/uf8TkpCkIQ

[alexandra birch @alexandrabirch1]

I have a PhD scholarship (UK/EU) at @InfAtEd funded by a new EU project working on low-resource machine translation. Please contact me if interested.

[Tim Baldwin @eltimster]

@acl2018 videos are now up for all oral presentations: https://t.co/AVpgXjEYVX. Let me know if there are any issues. #NLProc

[François Chollet @fchollet]

Books are a powerful thing. Some stories latch onto your mind one day and stay with you forever

[Sabrina J. Mielke @sjmielke]

Does the best number of BPE merges for language modeling¹ depend on the language²? Yes!
______________
¹ Why use BPE for LMs? It's a surprisingly good open-vocab baseline, see https://t.co/SlrY7yTK0m .
² Ooh, multiple languages? Yup, this is follow-up to https://t.co/gAS5VLdZ7f !

[Graham Neubig @gneubig]

Our #EMNLP2018 paper (last one, I promise!) introduces a data set for "Machine Translation of Noisy Text" by @pmichelX. I'm really excited about this data set, a *natural* test-bed for noise-robust sequence-to-sequence models. https://t.co/poT1DdVmUU

[Will Barkis @willbarkis]

Interesting: "Logged off: Meet the teens who refuse to use social media" https://t.co/xKkVYYGmSp And, yes, I note the irony in posting this here -- though Twitter is not the same as FB, Instagram, etc.

[Himanshu Maurya @lordzuko2]

A point made.

[Katharine Birbalsingh @Miss_Snuffy]

I say this to parents at school. The fat tech cats make their billions off you giving your children the latest tech gadget while they fill their houses with books.

[Isabelle Augenstein @IAugenstein]

Periodic reminder: I maintain the BIG #NLProc Directory. Consider signing up, or use it as a resource for hiring, finding invited speakers and more!
https://t.co/oONBxkjg0S #Coling2018

[Himanshu Maurya @lordzuko2]

My first reference when I started learning, waited for each chapter to come along, now it's time to binge read the entire book. Thanks @iamtrask

[ZPER @ZPER_team]

ZPER Official Video :)
More information - https://t.co/EIjXJmcXzv

[Himanshu Maurya @lordzuko2]

I liked a @YouTube video https://t.co/ZeRXo7zVDF Training CMU Sphinx Speech Recognition