Speaker-Sensitive Dual Memory Networks for Multi-Turn Slot Tagging
This addresses slot tagging errors in multi-turn dialogs for personal assistants like Cortana, representing an incremental advance with a novel method for a known bottleneck.
The paper tackled the problem of multi-turn slot tagging errors due to missing contextual information by introducing Speaker-Sensitive Dual Memory Networks that encode utterances based on speaker roles, resulting in significant performance improvements over state-of-the-art models on real user data from Microsoft Cortana.
In multi-turn dialogs, natural language understanding models can introduce obvious errors by being blind to contextual information. To incorporate dialog history, we present a neural architecture with Speaker-Sensitive Dual Memory Networks which encode utterances differently depending on the speaker. This addresses the different extents of information available to the system - the system knows only the surface form of user utterances while it has the exact semantics of system output. We performed experiments on real user data from Microsoft Cortana, a commercial personal assistant. The result showed a significant performance improvement over the state-of-the-art slot tagging models using contextual information.