Building the European Social Innovation Database with Natural Language Processing and Machine Learning

Area of Science:

• Social Sciences
• Computer Science
• Data Science

Background:

• Social innovation, defined as new products, services, or models meeting social needs and fostering collaboration, lacks comprehensive data for research.
• Existing data limitations hinder a thorough understanding and analysis of social innovation initiatives globally.

Purpose of the Study:

• To address the data gap in social innovation research.
• To create a comprehensive and accessible database of social innovation projects worldwide.
• To leverage advanced computational methods for data collection and characterization.

Main Methods:

• Developed the European Social Innovation Database (ESID) through large-scale collection of unstructured website text.
• Employed advanced machine learning techniques to extract features like social innovation dimensions, project locations, summaries, and topics.
• Achieved high model performance with an F1 score of up to 0.90.

Main Results:

• ESID currently contains 11,468 social innovation projects from 159 countries.
• The database includes extracted features such as project summaries, locations, and thematic classifications.
• Data is freely available and accessible via a web-based application.

Conclusions:

• ESID provides a valuable, data-driven resource for social science research on social innovation.
• The database facilitates a deeper understanding of global social innovation trends and impacts.
• Future plans include database expansion, variable addition, and regular data updates.