Crowdsourcing offers an invaluable toolkit for obtaining dynamic trends and insights from social media data analytics, enabling the capture of the wisdom of the crowds. The plethora of available platforms requires the appropriate definition of data schemas and techniques to allow for efficient knowledge extraction from unstructured social media user generated content and users’ multilevel interactions. The present work addresses such challenges by designing an effective and flexible document based data model that supports heterogeneous social media data integrations. This model is then exploited under a crowdsourced topics observatory that involves interactive visualization modules and advanced topic modelling methods. The proposed framework is implemented and demonstrated on a social innovation platform aiming to promote awareness on plastic waste revaluation and empower stakeholders of the plastics value chain.

Published in the IEEE Intelligent Informatics BULLETIN, October 2020, Vol. 20, No. 1, ISSN 1727-5997