Jiang Guo

ACL announcement: "The ACL Executive Committee has voted to significantly change ACL's approach to protecting anonymous peer review. The change is effective immediately. (1/4) #NLPRoc

2014->2024, 100M->1T, same people, pretty similar model. It's such a crazy story.

anon policy survey is out: tinyurl.com/aclarxivpolicy

The authors of MolScribe, which uses AI for molecular structure recognition, now present RxnScribe, a model which goes a step further by extracting reaction diagrams & predicting reactions in new diagrams. Paper: arxiv.org/abs/2305.11845 Code: github.com/thomas0809/Rxn…

Great tech talk on subtleties of LLM hallucinations by @johnschulman2 : where they come from, how to mitigate them, remaining open problems. youtube.com/watch?v=hhiLw5…

Molecules are commonly depicted as 2D images in literature w/ significant variation in drawing styles/conventions, making it difficult for machines to recognize their structures. #JameelClinic researchers @Yujie_Qian @jiangfeng1124 @ZhengkaiTu @zli11010 w/ PIs @cwcoley &...🧵 1/4

RIP, Octavian.

The second of the #acl2022nlp 25-year test of time awards goes to Michael Collins for his work on Statistical Parsing published at EACL 1997 #NLProc aclanthology.org/P97-1003/

@jwmay @cwcoley @dan2097 I agree it's ambiguous here, but it really refers to the reaction extraction system Dan described in his thesis, which is the combination of ChemTagger+OPSIN, and a set of rules for chemical role assignment (github.com/dan2097/patent…)

@cwcoley @dan2097 corrects me, the project was funded specifically for patents :-). Parsing IUPAC names is of course pretty useful in general and not just patents.

@cwcoley Did a/o ever let you know this is wrong: "OPSIN identifies products by a set of rules based on the tagging and parsing outputs of CT. It was developed specifically for processing patent literature, which is highly different from journal articles in terms of language use"

Yesterday, we "officially" introduced the Open Reaction Database to the community with a perspective in @J_A_C_S. Thanks to all of our terrific collaborators, advisors, and contributors for making this a reality! pubs.acs.org/doi/full/10.10… & docs.open-reaction-database.org/en/latest/over…

"When Attention Meets Fast Recurrence: Training Language Models with Reduced Compute" by Tao Lei @taolei15949106 - Outstanding Paper at EMNLP arxiv.org/abs/2102.12459 (Tao's work is always must read. Combines algorithmic cleverness with practical engineering and experiments.)

Automated #Chemical_Reaction Extraction from #Scientific_Literature pubs.acs.org/doi/10.1021/ac… @jiangfeng1124 @gao_hanyu @Varal7 @cwcoley @BarzilayRegina #ASAP

What if we train a classifier on one data environment and apply it on another environment? In #ICML2021, we show that its prediction results are informative of the unstable correlations between the input and the output. This allows us to effectively train a stable classifier.

Congrats to Regina Barzilay, Jameel Clinic AI Faculty Lead, on winning the The Wallace H. Coulter Lectureship Award. This award recognizes an individual who has made an impact on education, practice, research in lab medicine or patient care. Read more: aacc.org/community/meri…

Google Translate can't recover lost languages. But MIT's new ML model can, thanks to machine learning. Details: bit.ly/DecipherLang #ML #NLP

We are excited to announce the AI Cures Conference: Data-driven Clinical Solutions for COVID-19, which will take place online on September 29th. Learn more and register here, spots are limited: aicures.mit.edu/conference

On Friday, August 21 at 8pm ET, Greater Boston’s beloved Coolidge Corner Theater (@thecoolidge) is hosting a virtual premiere of From Controversy to Cure, a documentary film chronicling the biotech boom in Cambridge. Learn more and watch it here: bit.ly/31fqWCU

AICures team releases top SARS-CoV-2 antiviral molecules from the @broadinstitute Library. Read about our new adaptive invariance algorithm here: bit.ly/2Y29Irf

We invite you to attend the MIT Covid-19 challenge event, Beat the Pandemic, a series of virtual hackathons. In this 48-hour virtual event, help tackle the most critical unmet needs caused by Covid-19. covid19challenge.mit.edu

In hopes of developing a treatment for Covid-19, a team of chemists has designed a drug candidate that may block coronaviruses’ ability to enter human cells. It binds to the protein that the viruses use to enter cells, potentially disarming it. mitsha.re/7Unc50yZEX0