A Survey of the First Five Years of eRisk: Findings and Conclusions

Early risk detection is a challenging task in which multiple objectives are involved. For this task, getting timely predictions with a reasonable confidence level becomes critical. Over the last five years, eRisk shared-task campaigns have focused on stimulating research in extracting evidences of different types of risk from Internet data (e.g., diagnosis of a mental disorder), fostering the development of early risk detection systems (ERDS). In this chapter, we present an overview of the key aspects and stages involved in the design and development of ERDS. Despite the heterogeneity observed, common elements can be identified from the wealth of systems submitted since eRisk’s outset. In particular, we review and summarise what has been learnt from each team contribution, specially focusing on how they built their systems to tackle different early detection tasks. This chapter concludes with several findings and lessons that have been discovered and learnt along the way.

keywords: