BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Unsupervised attention-guided atom-mapping - Philippe Schwaller\, 
 IBM Research
DTSTART:20200831T160000Z
DTEND:20200831T163000Z
UID:TALK150703@talks.cam.ac.uk
CONTACT:Bingqing Cheng
DESCRIPTION:Language models called transformers have recently revolutioniz
 ed natural language processing and show great potential when applied to te
 xt-based representations of chemical reactions. The patterns in chemical r
 eactions are learned by predicting masked parts of reaction SMILES. The pr
 etrained models can then be specialized on a task like reaction classifica
 tion [1]and yield predictions [2]\, where they reach unprecedented accurac
 ies. Not only can specific outputs of the transformer models serve as fing
 erprints to map the chemical reaction space without the need of knowing th
 e reaction center or distinguishing between reactants and reagents\, but t
 hey can also be used to recover the rearrangement between reactant and pro
 duct atoms [3]. By opening the black-box using detailed visual analysis\, 
 we discovered that the transformer models learned atom-mapping without sup
 ervision. Atom-mapping is necessary for making chemical reaction data bett
 er machine-accessible and crucial for graph- and template-based reaction p
 rediction and synthesis planning approaches. Here\, we present an attentio
 n-guided reaction mapper that shows remarkable performance in terms of spe
 ed and accuracy\, even for strongly imbalanced reactions as typically foun
 d in patents. This work is the first demonstration of knowledge extraction
  from a self-supervised language model with a direct practical application
  in the chemical reaction domain.\n\nReferences:\n[1] Mapping the Space of
  Chemical Reactions using Attention-Based Neural Networks\nP Schwaller\, D
  Probst\, AC Vaucher\, VH Nair\, D Kreutter\, T Laino\, JL Reymond\nhttp:/
 /dx.doi.org/10.26434/chemrxiv.9897365\n\n[2] Prediction of Chemical Reacti
 on Yields using Deep Learning\nP Schwaller\, AC Vaucher\, T Laino\, JL Rey
 mond\nhttp://dx.doi.org/10.26434/chemrxiv.12758474\n\n[3] RXNMapper: Unsup
 ervised Attention-Guided Atom-Mapping.\nP Schwaller\, B Hoover\, JL Reymon
 d\, H Strobelt\, T Laino\nhttp://dx.doi.org/10.26434/chemrxiv.9897365
LOCATION:virtual ZOOM meeting ID: 263 591 6003\, https://zoom.us/j/2635916
 003
END:VEVENT
END:VCALENDAR