This paper briefly describes our research groups' efforts in tackling Task 1 (Early Detection of Signs of Self-Harm), and Task 2 (Measuring the Severity of the Signs of Depression) from the CLEF eRisk Track. Core to how we approached these problems was the use of BERT-based classifiers which were trained specifically for each task. Our results on both tasks indicate that this approach delivers high performance across a series of measures, particularly for Task 1, where our submissions obtained the best performance for precision, F1, latency-weighted F1 and ERDE at 5 and 50. This work suggests that BERT-based classifiers, when trained appropriately, can accurately infer which social media users are at risk of self-harming, with precision up to 91.3% for Task 1. Given these promising results, it will be interesting to further refine the training regime, classifier and early detection scoring mechanism, as well as apply the same approach to other related tasks (e.g., anorexia, depression, suicide).
|Title of host publication||Experimental IR Meets Multilinguality, Multimodality, and Interaction|
|Subtitle of host publication||12th International Conference of the CLEF Association, CLEF 2021, Virtual Event, September 21–24, 2021, Proceedings|
|Editors||K. Selçuk Candan, Bogdan Ionescu, Lorraine Goeuriot, Birger Larsen, Henning Mueller, Alexis Joly, Maria Maistro, Florina Piroi, Guglielmo Faggioli, Nicolo Ferro|
|Publication status||Published - 14 Sept 2021|
|Event||12th International Conference of the CLEF Association - Virtual Event|
Duration: 21 Sept 2021 → 24 Sept 2021
|Name||Lecture Notes in Computer Science|
|Conference||12th International Conference of the CLEF Association|
|Abbreviated title||CLEF 2021|
|Period||21/09/21 → 24/09/21|
Bibliographical noteCopyright © 2021 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). Acknowledgements & Funding: The first author would like to thank the following funding bod-
ies for their support: FEDER/Ministerio de Ciencia, Innovaci ́on y Universidades, Agencia Estatal de Investigaci ́on/Project (RTI2018-093336-B-C21), Conseller ́ıa de Educacion, Universidade e Formaci ́on Profesional and the European Regional DevelopmentFund (ERDF) (accreditation 2019–2022 ED431G-2019/04, ED431C 2018/29, ED431C2018/19). The second and third authors would like to thank the UKRI’s EPSRC Project Cumulative Revelations in Personal Data (Grant Number: EP/R033897/1) for theirsupport. They would also like to thank David Losada for arranging this collaboration.
- Social media
- Early detection