TyDi QA is a new multilingual dataset for information-seeking question answering featuring 11 Typologically Diverse languages and over 200k QA pairs. Learn more and start experimenting with the data and code ↓ goo.gle/39cZVkv

New from Google Research: CANINE, a pre-trained tokenization-free language encoder. This frees us from a variety of pitfalls associated with tokenization, but also improves quality on TyDi QA, a multilingual question answering benchmark. arxiv.org/abs/2103.06874

Just when subword models (BPE, SentencePiece, supported by the great Tokenizers library github.com/huggingface/to…) have become the new standard practice in #NLProc, signs emerge that their days may be numbered

New paper to appear in #emnlp2021! arxiv.org/abs/2109.07020 We study the syntactic abilities of BERT by manipulating the training corpus and retraining BERT.

Want your image generation model to stop misspelling everything? Try giving it access to character-level input features! arxiv.org/abs/2212.10562

Presenting FRMT, a new dataset and evaluation benchmark for Few-Shot Region-Aware Machine Translation that seeks to drive research progress on equitably serving speakers of different language varieties. Learn more and see how current models fare → goo.gle/3IvnPwc

Our low-resource POS-tagging NAACL (final) and ACL (draft) papers are now online along with source code: github.com/dhgarrette/low… #nlproc

My #naacl2013 talk slides on Low-Resource POS-Tagging are now posted on cs.utexas.edu/~dhg/ (.key & .pdf). Paper, code, and data too.

for all you #nlproc folks, my favorite #naacl2013 papers are... nlpers.blogspot.com/2013/06/my-naa…

Daniel Marcu's (my advisor's) fav #naacl2013 papers (totally agree but that's biased :P): sdl.com/community/blog…

The video of my #naacl2013 talk on Learning a POS-Tagger from Two Hours of Annotation is now online: techtalks.tv/talks/learning…

My #NAACL2013 presentation on weakly-supervised tagging was nominated for a *Best Talk* award! Watch it here: techtalks.tv/talks/learning… #NLProc

New paper with Chris Dyer, Jason Baldridge, Noah A. Smith! Weakly-Supervised Bayesian Learning of a CCG Supertagger, cs.utexas.edu/~dhg/papers/ga… #nlproc

My paper on weakly-supervised Bayesian CCG parser learning with Chris Dyer, Jason Baldridge, and Noah A. Smith was accepted to #AAAI2015!

Extremely excited to announce that I've accepted a postdoc position with Luke Zettlemoyer at the University of Washington starting in May!

Congrats to Dan Garrette for successfully defending his Computer Science at UT Austin dissertation on semi supervised learning for tagging and parsing! #NLProc

New paper! A "supertag-context" CCG parsing model: prefer constituent labels that fit contexts dhgarrette.com/papers/garrett…

Nice paper from Jason Baldridge that we had missed…. x.com/jasonbaldridge…

@skilpat Don't tell them about how there's not even guaranteed parental leave... it might be too much for them to handle.

New code for unsupervised joint dual -- literal and normalized -- OCR transcription available here github.com/tberg12/ocular w/ Hannah Alpert-Abrams 🤖✊🔮 #dh