Analysis of unstructured data: the Complidon case
In 2017/2018, Epiconcept supported Santé Publique France in its Complidon survey. The objective of this anonymous online study was to assess the opinion and behaviour of blood donors in relation to the selection criteria.
The challenge: to process the volume of free texts
The survey was a success: 110,000 blood donors participated. To get the most out of it, the processing of open-ended responses was essential. However, their volume was massive and difficult to compile by man.
To solve this problem, Santé Publique France chose to use "Feedback Explorer". Designed by Epiconcept, this tool uses Spark, an Open Source distributed computing framework. The natural language processing algorithms developed in this way make it possible to aggregate responses into semantic clusters that facilitate thematic grouping and analysis.