Book Image

Using OpenRefine

Book Image

Using OpenRefine

Overview of this book

Data today is like gold - but how can you manage your most valuable assets? Managing large datasets used to be a task for specialists, but the game has changed - data analysis is an open playing field. Messy data is now in your hands! With OpenRefine the task is a little easier, as it provides you with the necessary tools for cleaning and presenting even the most complex data. Once it's clean, that's when you can start finding value. Using OpenRefine takes you on a practical and actionable through this popular data transformation tool. Packed with cookbook style recipes that will help you properly get to grips with data, this book is an accessible tutorial for anyone that wants to maximize the value of their data. This book will teach you all the necessary skills to handle any large dataset and to turn it into high-quality data for the Web. After you learn how to analyze data and spot issues, we'll see how we can solve them to obtain a clean dataset. Messy and inconsistent data is recovered through advanced techniques such as automated clustering. We'll then show extract links from keyword and full-text fields using reconciliation and named-entity extraction. Using OpenRefine is more than a manual: it's a guide stuffed with tips and tricks to get the best out of your data.
Table of Contents (13 chapters)
Using OpenRefine
Credits
Foreword
About the Authors
About the Reviewers
www.PacktPub.com
Preface
Index

Recipe 4 – applying a text filter


In this recipe, you will learn about filters that allow you to search for values displaying some patterns.

When you want to find rows matching a certain string, it is easier to rely on a simple text filter than on cumbersome facets. Let's start with a simple example. Suppose you want to filter all titles relating to the United States. Navigate to Object Title | Text filter and watch the filter box open on the left, in the same tab where facets appear. Now type in USA. OpenRefine tells you that there are 1,866 matching rows. Select the case sensitive checkbox to eliminate happenstance matches, such as karakusa and Jerusalem, and we are down to 1,737 rows:

Still, we cannot be sure that there is no noise left in these matches; there could be occurrences of JERUSALEM in capital letters for instance. To get around this problem, we could try to add spaces to either side of USA, but at the risk of losing cases, such as [USA] or /USA, along with occurrences of the...