Slot filling is a critical subtask of Spoken language understanding (SLU) in task-oriented dialogue systems. This is a common scenario that different slot filling tasks from different but similar domains have overlapped sets of slots (shared slots). In this paper, we propose an effective deep multi-task learning with Cross Connected Layer (CCL) to capture this information. The experiments show that our proposed model outperforms some mainstream baselines on the Chinese E-commerce datasets. The significant improvement in the F1 socre of the shared slots proves that CCL can capture more information about shared slots.
CITATION STYLE
Kong, J., Cai, Y., Ren, D., & Li, Z. (2019). Deep Multi-task Learning with Cross Connected Layer for Slot Filling. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 11839 LNAI, pp. 308–317). Springer. https://doi.org/10.1007/978-3-030-32236-6_27
Mendeley helps you to discover research relevant for your work.