Automated Chemical Structure Extraction

自动化学结构提取

• 批准号: 2285004
• 金额: --
• 依托单位: University of Nottingham
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2019
• 资助国家: 英国
• 起止时间: 2019 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2285004
• 关键词: Automated; Chemical; Structure; Extraction

Automating chemical structure extraction from literature repositories such as, pdf documents, journals, images, into machine a machine-readable format. This is designed to extrapolate a wealth of data from an otherwise manual resource that can be used in large quantities to tackle ongoing unoptimized processes within the chemical industry. It also presents the potential for an expansive solution to the currently unsolved reliability problem of converting molecular structures to different formats using Image Recognition. By automating such an untapped resource; this opens opportunities for more data driven optimisation to help in the need for optimising chemical reaction/mechanism steps.Proposed solution and methodologyUse image recognition to autonomously segment images from journal articles and other resources to collect a wealth of machine-readable instances of molecules. The current proposed methodology is to use an end to end deep learning pipeline to define the stages of segmentation to conversion.

自动从文献库中提取化学结构，如pdf文档，期刊，图像，转换为机器可读格式。这旨在从其他手动资源中推断大量数据，这些数据可以大量用于解决化学工业中正在进行的未优化过程。它还提出了一个扩展的解决方案，目前尚未解决的可靠性问题的分子结构转换为不同的格式，使用图像识别的潜力。通过自动化这样一个未开发的资源，这为更多的数据驱动优化提供了机会，以帮助优化化学反应/机制步骤。建议的解决方案和方法使用图像识别自动分割期刊文章和其他资源中的图像，以收集大量的机器可读的分子实例。目前提出的方法是使用端到端深度学习管道来定义分割到转换的阶段。