ADAPTIVE INFORMATION MONITORING AND EXTRACTION

说明(改编自摘要)： 在线生物医学信息的快速和持续扩散 资源已经需要创建软件工具来自动地 浏览这些网络，识别相关文档，并提取 希望从他们那里得到信息。设计、实施和评估这样的 软件系统是本项目的首要目标。更确切地说， 一个中心目标是创建一种可以针对特定情况进行个性化的系统 以及改变个体生物医学科学家的兴趣。机器学习 方法在我们的软件助手中扮演核心角色，使系统能够 适应个人用户。一种灵活的语言，科学家可以使用它 将他们的兴趣传达给机器学习算法将 发展起来的。机器学习方法通常存在的弱点是需要 用户提供的培训示例，但创建此类示例通常是一种 繁重的任务。因此，另一个核心目标是大大减少对 通过创建获取这些示例的技术来获得人类标记的训练数据 通过间接的方法。正在开发的系统将在 几个生物医学研究人员的实验室。

DESCRIPTION (adapted from the Abstract): The rapid and continuing proliferation of on-line biomedical information sources has necessitated the creation of software tools for automatically navigating these networks, identifying relevant documents, and extracting the desired information from them. Designing, implementing, and evaluating such a software system is the primary objective of this project. More specifically, a central aim is to create a system that can be personalized to the particular and changing interests of individual biomedical scientists. Machine leaming methods play a central role in our software assistant allowing the system to adapt to individual users. A flexible language with which scientists can communicate their interests to the machine leaming algorithms will be developed. Machine leaming methods typically have the weakness of requiring user-provided training examples, but creating such examples is usually a burdensome task. Hence, another central aim is to greatly reduce the need for human-labeled training data by creating techniques that obtain these examples by indirect methods. The system being developed will be field-tested in the laboratories of several biomedical researchers.