SHF: Medium: Collaborative Research: Computer-Aided Programming for Data Science

这个名为DataWizard的项目的目标是通过使用计算机辅助编程，极大地简化当前数据分析所需的工作。具体地说，这个项目旨在通过从非正式规范自动生成程序来实现数据收集、查询和整理任务的半自动化。因此，DataWizard项目将允许领域科学家专注于更有趣的数据分析和可视化任务，将数据科学的“繁重工作”留给计算机辅助编程工具。该项目还将推进自动化程序合成和自然语言处理的最新技术，并将这些技术应用于新兴的大数据分析领域。从技术角度来看，DataWizard项目的目标有三个方面。首先，该项目开发了新的实例编程和信息提取技术，以解决数据收集中出现的挑战，包括不同数据源的整合、层次数据和关系数据之间的转换以及从非结构化数据源中提取信息。其次，该项目探索了使用自然语言描述查询数据的新技术。特别是，该项目考虑了从关系数据库和noSQL数据库以及半结构化数据源（如XML和JSON）中提取数据。第三，本项目开发了新的程序合成方法，用于自动化数据分析中常见的数据整理、清理和输入任务。总的来说，这些技术使数据科学家更容易从混乱的数据中获得见解。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The goal of this project, named DataWizard, is to dramatically simplify the effort that is currently required for data analytics through the use of computer-aided programming. Specifically, this project aims to semi-automate data collection, querying, and wrangling tasks by automatically generating programs from informal specifications. As a result, the DataWizard project will allow domain scientists to focus on more interesting data analytics and visualization tasks, leaving the "grunt work" of data science to computer-aided programming tools. The project will also advance the state-of-the-art in automated program synthesis and natural language processing and apply these techniques to the burgeoning field of big data analytics. From a technical perspective, the goals of the DataWizard project are three-fold. First, this project develops novel programming-by-example and information extraction techniques to address challenges that arise in data collection, including consolidation of different data sources, transformations between hierarchical and relational data, and extraction of information from unstructured data sources. Second, this project explores new techniques for querying data using natural language descriptions. In particular, this project considers data extraction from relational and noSQL databases as well as semi-structured data sources, such as XML and JSON. Third, this project develops novel program synthesis methods for automating data wrangling, cleaning, and imputation tasks that commonly arise in data analytics. Overall, these techniques make it significantly easier for data scientists to gain insights from messy data.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.