Google has updated and re-released open-source software for cleaning, analyzing and transforming data sets, now called Google Refine.
The software, originally called Freebase Gridworks, came with Metaweb, a company Google purchased in July
Google Refine is a collection of tools that could come in handy when wrangling useful information from a data set, particularly ones that have data inconsistencies.
This desktop application can, for instance, find all the variant spellings of a word in a data set and replace them with the appropriate term. This process, called normalization, is nothing new. But normalizing data usually requires writing code that is specific to one data set, noted Christopher Groskopf, a developer for the Chicago Tribune.
This desktop application can, for instance, find all the variant spellings of a word in a data set and replace them with the appropriate term. This process, called normalization, is nothing new. But normalizing data usually requires writing code that is specific to one data set, noted Christopher Groskopf, a developer for the Chicago Tribune.
The software contains a number of other tools as well. It includes an expression language that can be used to analyze a set of data. Filters can be used to isolate subsets of data, which then can be analyzed or changed through a set of transform commands.
The software works with plain text files, the data in which can be split into different columns by the use of commas. Results can exported back out in the JSON (JavaScript Object Notation) format, which can then be easily transformed into HTML tables or other formats.
The software can work with up to a few hundred thousand rows per data set, depending on the user's computer memory. And unlike most spreadsheet software, this software can interactively transform large subsets of data, the company asserted.
Google said this week that it has added several new features to the software, officially called Google Refine 2.0, including the ability to link records to other databases, and a number of new transformation commands and expressions.
The non-profit government watchdog organization ProPublica has used this software to aggregate data from seven different data sets to show how pharmaceutical companies pay doctors to recommend certain medications.
|
