Tri-level Joint Natural Language Understanding for Multi-turn Conversational Datasets
This work addresses the challenge of more natural conversational environments for whom incremental improvements in joint NLU models, though it is incremental as it builds on existing joint approaches.
The authors tackled the problem of natural language understanding in multi-turn conversations by introducing a tri-level joint model that adds domain information and enables explicit semantic exchange between all levels, resulting in state-of-the-art performance in slot filling and intent detection on multi-turn datasets.
Natural language understanding typically maps single utterances to a dual level semantic frame, sentence level intent and slot labels at the word level. The best performing models force explicit interaction between intent detection and slot filling. We present a novel tri-level joint natural language understanding approach, adding domain, and explicitly exchange semantic information between all levels. This approach enables the use of multi-turn datasets which are a more natural conversational environment than single utterance. We evaluate our model on two multi-turn datasets for which we are the first to conduct joint slot-filling and intent detection. Our model outperforms state-of-the-art joint models in slot filling and intent detection on multi-turn data sets. We provide an analysis of explicit interaction locations between the layers. We conclude that including domain information improves model performance.