WeChat Mini Program
Old Version Features

ReactXT: Understanding Molecular "Reaction-Ship" Via Reaction-Contextualized Molecule-Text Pretraining

Annual Meeting of the Association for Computational Linguistics(2024)

National University of Singapore

Cited 7|Views31
Abstract
Molecule-text modeling, which aims to facilitate molecule-relevant tasks witha textual interface and textual knowledge, is an emerging research direction.Beyond single molecules, studying reaction-text modeling holds promise forhelping the synthesis of new materials and drugs. However, previous worksmostly neglect reaction-text modeling: they primarily focus on modelingindividual molecule-text pairs or learning chemical reactions without texts incontext. Additionally, one key task of reaction-text modeling – experimentalprocedure prediction – is less explored due to the absence of an open-sourcedataset. The task is to predict step-by-step actions of conducting chemicalexperiments and is crucial to automating chemical synthesis. To resolve thechallenges above, we propose a new pretraining method, ReactXT, forreaction-text modeling, and a new dataset, OpenExp, for experimental procedureprediction. Specifically, ReactXT features three types of input contexts toincrementally pretrain LMs. Each of the three input contexts corresponds to apretraining task to improve the text-based understanding of either reactions orsingle molecules. ReactXT demonstrates consistent improvements in experimentalprocedure prediction and molecule captioning and offers competitive results inretrosynthesis. Our code is available at https://github.com/syr-cn/ReactXT.
More
Translated text
PDF
Bibtex
AI Read Science
Must-Reading Tree
Example
Generate MRT to find the research sequence of this paper
Data Disclaimer
The page data are from open Internet sources, cooperative publishers and automatic analysis results through AI technology. We do not make any commitments and guarantees for the validity, accuracy, correctness, reliability, completeness and timeliness of the page data. If you have any questions, please contact us by email: report@aminer.cn
Chat Paper
Summary is being generated by the instructions you defined