EarthCube Building Blocks: A Cognitive Computer Infrastructure for Geoscience

在这个时代，信息和数据的获取往往不是问题，而是有效处理和使用这些信息和数据的能力。在某些情况下，这些问题是由难以存储、传输和/或分析的海量单一数据集引起的。在其他情况下，首要问题是发现并聚合相关数据，这些数据在许多不同的位置和格式中广泛传播，例如在已发表的报纸、政府机构报告、电子表格和网站的表格、文本和数字中。地球科学目前缺乏一个网络基础设施，能够高效、廉价、高精度和准确地发现、提取和组织许多不同类型的数据，这些数据对推动科学研究和利用当前和过去在数据采集方面的投资至关重要。取而代之的是，有几十种孤立的、有时是多余的地学数据挖掘工作，它们使用人类作为主要机制来寻找数据，然后将它们输入结构化数据库。这种运作模式不仅成本高、速度慢，而且也是对人力资源和科学专门知识的低效利用。该项目开发了一个面向地学的训练有素的计算系统，该系统可以作为一个跨学科的工具，用于快速查找、提取和组织地学数据。与传统的数据处理系统不同，训练有素的系统使用统计或机器学习技术，为结构化程度低得多的复杂数据查询提供丰富的答案。更长期的愿景是建立一个经过EarthCube训练的计算系统，该系统可以帮助发现、提取和汇总数据，以及以一种帮助地球科学家解决新问题并更好地理解和模拟地球系统的方式处理、汇总和综合这些数据。该项目汇集了一支独特的跨学科团队，致力于构建、测试和运营EarthCube积木，将训练有素的计算系统和技术的力量带给更广泛的地球科学界。训练有素的计算系统为数据处理提供了一种全新的工具。

This is an era when access to information and data is often less of a problem than the abilityto efficiently process and use it. In some cases, these problems are caused by massive, monolithicdatasets that are difficult to store, transfer, and/or analyze. In other cases, the first-orderproblem is discovering and then aggregating relevant data that are widely disseminated inmany different locations and formats, such as in the tables, text, and figures of publishedpapers, government agency reports, spreadsheets, and websites. Geosciences currently lacksa cyberinfrastructure that can efficiently, cheaply, and with high precision and accuracyfind, extract, and organize many different types of data that are critical to advancing scienceand leveraging current and past investments in data acquisition. Instead, there are dozensof isolated, sometimes redundant, geosciences data mining efforts that use humans as the primarymechanism for finding data and then keystroking them into structured databases. This modeof operation is not only costly and slow, but it is also an inefficient use of human resourcesand scientific expertise. This project develops a geoscience-oriented trained computingsystem that can serve as a cross-disciplinary tool for rapidly finding, extracting, and organizinggeosciences data. Unlike traditional data processing systems, trained systems use statistical, or machine learning, techniques to provide rich answers to complex queries of data that are much less structured. The longer-term vision is to establish an EarthCube trained computing system that can aid in finding, extracting, and aggregating data, as well as in processing, summarizing, and synthesizing them in a way that helps geoscientists to tackle new problemsand better understand and model Earth systems. This project brings together a unique interdisciplinary team that is committed to building, testing, and operating an EarthCube Building Block that will bring the power of trained computing systemstechnologies to the broader geoscience community. Trained computing systems offer an entirely new breed of tools for data processing.