Announcing SWAG, a new natural language inference dataset, to appear at #emnlp2018. We present a general framework for collecting adversarial QA pairs at scale, minimizing bias. With Yonatan Bisk - not atttending any conferences, @royschwartz02, Yejin Choi. rowanzellers.com/swag/ arxiv.org/abs/1808.05326

New blog on how to manage stress as a researcher: If you really are an impostor, it's not a syndrome. theexclusive.org/2018/09/impost…

WFSAs are dead. Long live WFSAs! Our new EMNLP paper "Rational Recurrences" is out: arxiv.org/abs/1808.09357 Work by Hao Peng, @royschwartz02, me, and Noah A. Smith. We analyze several recent RNNs as finite-state automata with neural transition weights.

Deep Latent-Variable Models for Natural Language (Tutorial #emnlp2018 ): Slides and notebook now available (nlp.seas.harvard.edu/latent-nlp-tut…). Hope see everyone at the tutorial tomorrow morning.

Today, we're sharing insights on how to defend against Neural Fake News, AI-written disinformation that looks like real news.🕵️‍♀️🛡️ Joint with Ari Holtzman @hjrashkin Yonatan Bisk - not atttending any conferences Ali Farhadi Franzi Roesner & Yejin Choi at Allen School + Ai2. arxiv.org/abs/1905.12616 thread 1/4

Hey Everyone! UW NLP is excited to see everyone starting at @NAACLHLT tomorrow. Our students, postdocs, faculty, and collaborators all published some great research so we hope you’ll come talk to use at our tutorials, talks, posters and workshops! #NLProc

We've spent a few evenings last week building an interactive demo called *Write with Transformer* It lets you interact in a very intimate way with GPT-2, call, control, question the model... and I just can't stop playing with it! You can try it at transformer.huggingface.co

Sentence summarization with Information Bottleneck 🍾 and no supervision! 😎 Self-supervised and unsupervised approaches 🍾BottleSum🍾 to appear at @emnlp2019 arxiv.org/abs/1909.07405 With Ari Holtzman Jan Buys Yejin Choi at Allen School and Ai2

🍾Information Bottleneck🍾 in action at #emnlp2019! (1) Specializing embeddings for parsing (arxiv.org/abs/1910.00163) by Xiang & Jason Eisner (2) 🍾BottleSum🍾 unsupervised & self-supervised summarization (arxiv.org/abs/1909.07405) with Peter West Ari Holtzman Jan Buys

Blog post: announcing "Findings of EMNLP", a new ACL anthology journal. "What sets Findings apart from the main conference papers is that there is no requirement for high perceived impact, and accordingly solid work in untrendy areas... will be eligible." 2020.emnlp.org/blog/2020-04-1…

In the 90s, people tried to build small HMM LMs, but they were much worse than ngrams and RNNs. Recently, we got a bit obsessed with rebuilding an HMM LM in CUDA. We were able to scale to 65,000 states at ~RNN speed. HMMs are surprisingly good LMs. arxiv.org/abs/2011.04640

I have a position available for a full-time, funded 2-year Masters by research at the University of Cape Town. The project will be in the area of low-resource text generation. For more information see janmbuys.com/masters-positi…

The HPI Research School at the University of Cape Town offers Scholarships for doing a PhD in Computer Science. Applications for 2022 close on 15 August. sit.uct.ac.za/sit/research/h…

📢 Large-Scale MT for African Languages! 📢 Are you interested in building MT systems useful for hundreds of millions of people? To be specific, for 24 languages of Africa?🌍 Check out this WMT shared task: statmt.org/wmt22/large-sc… featuring both data and modeling tracks!

I will be at #EMNLP2022 EMNLP 2026 in person this week to present our Findings paper "Subword Segmental Language Modelling for Nguni Languages" (w/ Jan Buys). arxiv.org/pdf/2210.06525… 1/

Our paper "Subword Segmental Machine Translation: Unifying Segmentation and Target Sentence Generation" has been accepted to Findings of #ACL2023 ACL 2026 (w/ Jan Buys). arxiv.org/pdf/2305.07005… 1/

I am excited to start my sabbatical at HPI!

I am at #NAACL2024 to present our paper on the role of subwords in multilingual MT (w/ Jan Buys). aclanthology.org/2024.findings-… Stop by our poster (Wednesday 11am) to learn about the tradeoffs between tokenizers in inducing cross-lingual transfer across different linguistic contexts.