Multi-task Learning over Graph Structures
This work addresses the need for more flexible and interpretable multi-task learning methods in natural language processing, though it is incremental in its approach.
The paper tackles the problem of multi-task learning by introducing a graph-based framework that dynamically learns task relationships, outperforming baselines in text classification and sequence labeling tasks.
We present two architectures for multi-task learning with neural sequence models. Our approach allows the relationships between different tasks to be learned dynamically, rather than using an ad-hoc pre-defined structure as in previous work. We adopt the idea from message-passing graph neural networks and propose a general \textbf{graph multi-task learning} framework in which different tasks can communicate with each other in an effective and interpretable way. We conduct extensive experiments in text classification and sequence labeling to evaluate our approach on multi-task learning and transfer learning. The empirical results show that our models not only outperform competitive baselines but also learn interpretable and transferable patterns across tasks.