Continually Adaptive Machine Learning Platform for Personalized Biomedical Literature Curation and Exploration

用于个性化生物医学文献管理和探索的持续自适应机器学习平台

• 批准号: 10660315
• 负责人: TIMOTHY W CLARK
• 金额: $ 33.48万
• 依托单位: UNIVERSITY OF VIRGINIA
• 依托单位国家: 美国
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-08-01 至 2027-04-30
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10660315
• 关键词: Acceleration; Adoption; Artificial Intelligence; Biomedical Research; Clinical; Communities; Complex; Custom; Data; Development; Dimensions; Environment; Evaluation; Expert Systems; Foundations; Goals; Individual; Knowledge; Laboratories; Learning; Literature; Machine Learning; Measures; Methods; Mission; Modeling; Morphologic artifacts; Navigation System; Printing; Process; Records; Reporting; Research; Research Personnel; Resources; Semantics; Services; Strategic Planning; Structure; System; Techniques; Time; Training; United States National Library of Medicine; Update; adaptive learning; deep learning; design; digital; flexibility; improved; interest; journal article; knowledge base; knowledge curation; learning strategy; natural language; next generation; novel; outreach; public repository; search engine; tool; web-based tool

Project Summary: This proposal develops a novel, continually adaptive learning, and web-based tool which automatically and continually integrates and curates biomedical literature and knowledge. It allows researchers and curators to continually integrate individual lab-generated articles with large-scale public repositories, thereby enabling biomedical users to perform custom and context-based integrative literature analysis and exploration. More specifically, the proposed framework proposes a novel knowledge-enriched representation learning model that continually interleaves the available biomedical corpora with the biomedical knowledge from domain-expert curated knowledge bases to generate a knowledge-enriched representation needed by curators to annotate biomedical journal articles, preprints, and clinical records with high accuracy and learning efficiency. Furthermore, the proposed framework builds an evolving multi-dimensional knowledge network that incorporates fresh knowledge into the knowledge network via dynamic updates. Such an up-to-date knowledge network facilitates perpetual exploration of biomedical literature and allows researchers and curators to fulfill their personalized information navigation needs quickly. Finally, the proposed web-based tool enables users to flexibly integrate their own/personalized articles and lab reports into the massive existing corpora and knowledge bases, query the up-to-date biomedical information landscapes, and drill down into aspects relevant to them in order to fulfill their specific analysis goals. Currently, there is no automatic or systematic method to continually integrate individual lab-generated articles with large-scale public repositories that enables biomedical researchers and curators to perform integrative analysis based on their own context and obtain biologically meaningful results. Thus, the proposed research is an important step to expedite the goal of automatically and continually curating entities from large-scale biomedical corpora, integrating valuable information from the community curated platforms, and designing an advanced navigation system that allows users to perform knowledge exploration for their specific information needs. Dissemination activities through tools will help promote the adoption of the proposed system into real-world laboratories and research environments, and beyond. Moreover, such a form of integration promotes great outreach and demonstrates our commitment to the FAIR principles, the ability to Find, Access, Interoperate, and Reuse digital content.

项目概要： 该建议开发了一种新颖的、持续自适应的学习和基于网络的工具， 不断整合和策划生物医学文献和知识。它允许研究人员和馆长 不断地将实验室生成的文章与大规模的公共知识库集成，从而使 生物医学用户执行定制和基于上下文的综合文献分析和探索。更 具体而言，所提出的框架提出了一种新的知识丰富的表示学习模型， 不断地将现有的生物医学语料库与来自领域专家的生物医学知识交织在一起， 策展的知识库，以生成策展者注释所需的知识丰富的表示 生物医学期刊文章，预印本和临床记录，具有高准确性和学习效率。 此外，所提出的框架建立了一个不断发展的多维知识网络， 通过动态更新将新知识注入知识网络。这样一个最新的知识网络 促进生物医学文献的永久探索，并允许研究人员和策展人履行他们的职责。 个性化信息导航需求迅速。最后，建议的基于Web的工具使用户能够灵活地 将他们自己的/个性化的文章和实验室报告整合到大量现有的语料库和知识库中， 查询最新的生物医学信息景观，并深入研究与之相关的方面，以便 实现其特定的分析目标。目前，还没有自动或系统的方法来持续集成 个人实验室生成的文章与大规模的公共知识库，使生物医学研究人员和 策展人根据自己的背景进行综合分析，并获得有生物学意义的结果。 因此，本研究对于加速自动化及持续策展的目标是重要的一步 实体从大型生物医学语料库，整合有价值的信息，从社区策划 平台，并设计一个先进的导航系统，允许用户进行知识探索， 具体的信息需求。通过各种工具开展的传播活动将有助于促进 建议的系统到现实世界的实验室和研究环境，以及超越。此外，这种形式 一体化促进了更大的推广，并表明了我们对公平原则的承诺， 查找、访问、互操作和重用数字内容。