Sasha Sotudeh

@mattshumer_ @HyperWriteAI This is so hilarious lol. Liked the part on improvising “$25k not covering coffee” and the suggestion for the investor “to focus on whatever occupies their time”. Wondering if you kinda personalize the LLM response given user interests/preferences (not present within the prompt)?

LLaMA 3 70B is live on @HyperWriteAI! A custom version I've been working on for weeks. - writes more like a human than any other model - can access real-time info Crushing every other model, including GPT-4, in user preference tests. 🤯Seriously, look at the writing quality:

@kornraphop Feels good. In that case, we'll turn out to be a "supercomputer" tho😂

@sajastu Agree. I wish I could have a superpower that I can get results based on whatever ideas real quick, like, in 3 seconds per idea would be great 🤣

Spent 3 hours to come up with an equation that only gods know if it’s gonna work 🤪🤢 #phdlife #AcademicTwitter

I'll be presenting my poster on "TSTR: Intro-Guided Extended Summary Generation" tomorrow 2:30-4:00pm. If you're attending #NAACL2022 in-person, please stop by for discussions. I'll be more than glad to chat with you all on this work. arxiv.org/abs/2206.00847

@sepideh_ntzr @acm_ictir @MahmoodNeshati Congrats!

Our paper "A novel Approach for Mining Team Leaders in Community Question Answering" is accepted at @acm_ictir #ictir2021, w/ @MahmoodNeshati

@EYangTW Congrats, Eugene!

I am very excited to announce that I will be joining HLTCOE at JHU this September as a postdoc researcher!

part of scientific reading is to critically read beyond the text and considering what the paper/authors are not telling you / what they left out. it is true for work coming out of FAANG as it is to word coming from every other academic researcher.

We often cite papers using arXiv info w/o noting that they are already PUBLISHED in confs like @aclmeeting and @iclr_conf. These incorrect bib entries are so annoying!😠 We introduce Rebiber, a tool to fix them using @aclanthology and @dblp_org. 😆 Code: github.com/yuchenlin/rebi…

The extrinsic analysis: 1) our model improves consistently as the summary length increases; 2) our model adjusts the extraction probability of sentences toward salient sentences across diverse sections of the source document and pick those with higher confidence. (4/4)

In order to support this task, we additionally collect two extended summarization datasets: arXiv-Long, and PubMed-Long. The experimental results indicate that the multi-tasking model either outperforms or matches the performance of the prior baseline (i.e., BertSumExt). (3/4)

I'm excited to announce our new paper: "On Generating Extended Summaries of Long Documents" w/ @armancohan and Nazli Goharian. arXiv: arxiv.org/abs/2012.14136 Code and Datasets: github.com/Georgetown-IR-… (1/4)

If you're at #emnlp2020 this year, check out our paper on "Generating long summaries from scientific documents" at aclweb.org/anthology/2020…, and come to our presentation today 12:15-12:30 pm ET. At the SDP workshop.

According to our model, Taiwan will be virus free in 3 days! Check out your country! #Taiwan" target="_blank" rel="nofollow noopener">ir.cs.georgetown.edu/sidir/#Taiwan #COVID19 #Taiwan #virusfreeday

ML twitter: I am teaching a grad level course on “foundations of deep learning“ in the Fall. Looking for suggestions about important deep learning papers: old or new, published or on arxiv, well-cited or less-known, theoretical or empirical. Topics are flexible.

Help Negar fight cancer gofundme.com/f/help-negar-f…

which could significantly improve the summarization process. The extrinsic evaluation also showed that the system could generate summaries, over 80% of which are as good as human-written summaries.

Our paper accepted to #acl2020nlp went online at arxiv.org/abs/2005.00163. Therein, we showed the high importance of robust content selectors in summarizing clinical notes. Specifically, finding "domain knowledge alignments" between source and target...

@IR_oldie @xiuzhenzhang @rmit_csit Congrats! Do you have plans to release pre-print anytime sooner?

"Evaluation of Cross Domain Text Summarization" short paper accepted #sigir2020 with @xiuzhenzhang Shiwei Zhang. Based on the work of a brilliant @rmit_csit Honours undergraduate student Liam Scanlon. Well done Liam, your first paper.

@ABH878 Thanks! Absolutely, will get ready shortly!

@sajastu Congrats! Do you plan to release a preprint any time soon?

Our paper: "Attend to Medical Ontologies: Content Selection for Clinical Abstractive Summarization" w/ Nazli Goharian got accepted at #acl2020nlp. So excited as to being my first ACL! #NLProc