LSI_UNED at CLEF eHealth2021: Exploring the effects of transfer learning in negation detection and entity recognition in clinical texts

Hermenegildo Fabregat, Andrés Duque, Lourdes Araujo, Juan Martínez-Romo
2021 Conference and Labs of the Evaluation Forum  
This paper describes the approach presented by the LSI_UNED team in the Multilingual Information Extraction task (SpRadIE) of CLEF eHealth 2021. The proposed system is a deep learning stack designed for separately detecting negation hedge cues and other biomedical entities in the task. Transfer learning techniques are applied for studying whether pre-trained weights from a different negation detection task can be effectively incorporated into the model for improving a baseline system trained
more » ... y with the provided data. The system obtains promising results in the task, obtaining the second best F1 score, and the best precision score among all participant systems.
dblp:conf/clef/FabregatDAM21 fatcat:6pdhmdq6brezbkaoyp2oxlzavy