| Summary: | In this research note we discuss the two basic computational methods available for categorizing nonprofit organizations (NPOs) according to their field of activity based on textual information about these organizations: (1) rule-based categorization and (2) pattern recognition by using machine learning techniques. These methods provide a solution to the widespread research problem that quantitative data on the activities of NPOs are needed but not readily available from administrative data, and that manual categorization is not feasible for large samples. We explain both methods and report our experience in using them to categorize Austrian nonprofit associations on the basis of the International Classification of Non-Profit Organizations (ICNPO). Since we have found that rule-based categorization works much better for this task than machine learning, we provide detailed recommendations for implementing a rule-based approach. We address scholars with a background in data analytics as well as those without, by providing non-technical explanations as well as open-source sample code that is free to use and adapt.
|
|---|
