🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Interesting ACL short paper showing that classic word embeddings become competitive with their contextualised counterparts as end-task labeled data increases, particularly where simpler language is used. arxiv.org/abs/2005.09117

Are you well-versed in any of the following technologies and would you like to help make virtual ACL 2020 a success? Call for Volunteers - see the blog post #acl2020nlp #NLProc CSS/JS, Python/Flask, Apache/NGINX, and GitHub CI and Actions.

#winlp2020 ACL 2025 decisions are out! Congratulations to the authors of the accepted papers. We look forward to having all of our submitters participate in the workshop. Thanks and best wishes for your hard work. Stay safe! #NLProc #acl2020nlp #nlp

Tip: make Spotlight and Finder searches far more predictable by prefixing queries with name:, forcing a filename search. This represents >90% of my searches. For the other 10% I issue a normal query.

Like to use Regular Expressions (RE) for building rule-based classifiers for cold-start problems but would like to improve them with labeled data? Very nifty paper EMNLP2020 that shows how to go from RE->RNN and back again. faculty.sist.shanghaitech.edu.cn/faculty/tukw/e…

📝I wrote a blog post that explains few-shot learning with Pattern-Exploiting Training (and how it can be used to outperform OpenAI's GPT-3 on SuperGLUE) in simple terms: timoschick.com/explanatory%20… Would be happy to hear your feedback😊#NLProc

Our paper was awarded at ECIR2021! It is an honour! Thank you for the amazing conference. 💪🏻💪🏻

If you've registered for #sigir2021 and are attending #ictir2021 today, you should consider joining Session 2B on Conversational Search. It will start in about 15 minutes. SIGIR 2025 ICTIR 2024

Join our #convsearch evaluation session this Friday 10AMCET🔍Alexandre Salle UFRGS presents simulation framework which received industry impact award @2021ecir joint work w Shervin Malmasi Oleg Rokhlenko Eugene Agichtein Amazon Science Emory University amazon.science/publications/s…

Interesting talk by Alexandre Salle on his work done while at Amazon Science on refinement, e.g. clarification questions, with user simulation for conv. search. The user simulator is called CoSearcher. github.com/alexandres/CoS…. SCAI #convsearch

Interesting new paper: arxiv.org/abs/2204.11447 1) A method for breaking an existing IR evaluation into interpolation and extrapolation regimes 2) Late-interaction models extrapolate far better than dense retrieval models 3) Focus more on pretraining and less on negative sampling

A wonderful post by @LaurentPoirrier with the solution to a challenge I sent him back in March: "Can you always solve Wordle with a fixed set 5 of initial guesses?" The answer is no, but with 6 you can. poirrier.ca/notes/wordle-f…

I had a macro for this in C++ which I always missed when coding in Python. Turns out it exists in Python since 2019! github.com/python/cpython…

The recording of the wonderful AACL-IJCNLP2022 debate on “is there more to NLP than deep learning?” is public now: underline.io/lecture/67442-…

The Office but with ATP players. This is gold 😂

Tired of PDFs with filenames like ed1d47.pdf and 2303.08774.pdf? I built a little tool that leverages GPT to rename PDF files. github.com/alexandres/pdf… Below-After pics: