MTTLADE: A multi-task transfer learning-based method for adverse drug events extraction
作者：
Highlights：
• We propose an end-to-end system for adverse drug events extraction, called MTTLADE.
• MTTLADE integrates a dual-task sequence labelling for dealing with multi-head issue.
• It combines multi-task and transfer learning for simultaneous learning process.
• We assess the effect of five pre-trained language models to the overall results.
• We study the generalizability and the effectiveness of MTTLADE on two datasets.
摘要
•We propose an end-to-end system for adverse drug events extraction, called MTTLADE.•MTTLADE integrates a dual-task sequence labelling for dealing with multi-head issue.•It combines multi-task and transfer learning for simultaneous learning process.•We assess the effect of five pre-trained language models to the overall results.•We study the generalizability and the effectiveness of MTTLADE on two datasets.
论文关键词：Adverse drug events,Transfer learning,Multi-task learning,Named entity recognition,Relation extraction,Natural language processing
论文评审过程：Received 8 September 2020, Revised 12 December 2020, Accepted 13 December 2020, Available online 8 February 2021, Version of Record 8 February 2021.
论文官网地址：https://doi.org/10.1016/j.ipm.2020.102473
