Transfer Learning Models for Health-Related Twitter Data

These models were created for the Shared Tasks of the Social Media for Health Research and Applications Workshop at ACL 2019. For more detail on the tasks see: https://healthlanguageprocessing.org/smm4h/challenge/ They were all based on first fine-tuning pre-trained language models using transfer learning methods and then training these models to perform the task at hand. For more details on the methodology please see our paper: Dirkson, A.R. & Verberne, S. Transfer Learning for Health-Related Twitter Data. Proceedings of the ACL 2019 SMM4H workshop. Pre-print available at: http://liacs.leidenuniv.nl/~dirksonar/wp-content/uploads/2019/07/SharedTasksSMM4H2019_TMRLeiden_Preprint.pdf本系列模型专为2019年国际计算语言学协会（Association for Computational Linguistics, ACL）年会下设的健康研究与应用社交媒体工作坊（Social Media for Health Research and Applications Workshop）的共享任务赛道开发而成。若需了解该共享任务的更多细节，请访问：https://healthlanguageprocessing.org/smm4h/challenge/。所有模型的构建流程均统一为：首先通过迁移学习（transfer learning）方法对预训练语言模型（pre-trained language models）进行微调（fine-tuning），随后针对当前任务完成模型的针对性训练。若需了解模型构建方法的详细细节，请参阅以下论文：Dirkson, A.R. 与 Verberne, S. 撰写的《面向健康相关推特（Twitter）数据的迁移学习》（Transfer Learning for Health-Related Twitter Data），发表于ACL 2019 SMM4H工作坊会议论文集，该论文的预印本可通过以下链接获取：http://liacs.leidenuniv.nl/~dirksonar/wp-content/uploads/2019/07/SharedTasksSMM4H2019_TMRLeiden_Preprint.pdf