Exploring Natural Language Processing, Image Processing, Machine Learning, and Us

描述(申请人提供)：大多数生物医学文本挖掘系统仅以文本信息为目标，不提供对其他重要数据的智能访问，例如数字。在生物医学文献中，数字通常比任何其他文献更能代表发现的“证据”。全文生物医学文章几乎总是包含图像，这是生物医学知识发现的关键内容。生物医学科学家需要获取图像来验证研究事实，并制定或测试新的研究假设。评估表明，文献中所报告的文字陈述经常是嘈杂的(即，包含“虚假事实”)。捕捉本质上是支持文本“事实”的实验性“证据”的图像将使生物医学信息系统、数据库和生物医学科学家受益。我们正在开发一个生物医学文献图形搜索引擎BioFigureSearch。我们在自然语言处理、图像处理、机器学习和用户界面方面开发创新的算法和模型。交付成果将是新的生物医学自然语言图形处理(BNLfP)算法和iBioFigureSearch，允许生物医学科学家有效地访问图形数据，以及将增强生物医学信息检索、摘要和问题回答的开源工具。我们将开发的bNLfP算法可以应用于或集成到其他生物医学文本挖掘系统中。

DESCRIPTION (provided by applicant): Most biomedical text mining systems target only text information and do not provide intelligent access to other important data such as Figures. More than any other documentation, figures usually represent the "evidence" of discovery in the biomedical literature. Full-text biomedical articles nearly always incorporate images that are the crucial content of biomedical knowledge discovery. Biomedical scientists need to access images to validate research facts and to formulate or to test novel research hypotheses. Evaluation has shown that textual statements reported in the literature are frequently noisy (i.e., contain "false facts"). Capturing images that are essentially experimental "evidence" to support the textual "fact" will benefit biomedical information systems, databases, and biomedical scientists. We are developing a biomedical literature figure search engine BioFigureSearch. We develop innovative algorithms and models in natural language processing, image processing, machine learning and user interfacing. The deliverables will be novel biomedical natural language figure processing (bNLfP) algorithms and iBioFigureSearch allowing biomedical scientists to access figure data effectively, and open-source tools that will enhance biomedical information retrieval, summarization, and question answering. The bNLfP algorithms we will be developing can be applied or integrated into other biomedical text-mining systems.