Publication | Open Access
MA-DST: Multi-Attention-Based Scalable Dialog State Tracking
48
Citations
30
References
2020
Year
EngineeringSpoken Language ProcessingSpoken Dialog SystemSpeech RecognitionNatural Language ProcessingComputational LinguisticsConversation AnalysisLanguage StudiesDialogue ManagementNlp TaskLinguisticsDialog AgentsDialog SystemsComputer ScienceSpeech CommunicationConversation HistoryCoreference ResolutionSpeech ProcessingDialog State Tracking
Task oriented dialog agents provide a natural language interface for users to complete their goal. Dialog State Tracking (DST), which is often a core component of these systems, tracks the system's understanding of the user's goal throughout the conversation. To enable accurate multi-domain DST, the model needs to encode dependencies between past utterances and slot semantics and understand the dialog context, including long-range cross-domain references. We introduce a novel architecture for this task to encode the conversation history and slot semantics more robustly by using attention mechanisms at multiple granularities. In particular, we use cross-attention to model relationships between the context and slots at different semantic levels and self-attention to resolve cross-domain coreferences. In addition, our proposed architecture does not rely on knowing the domain ontologies beforehand and can also be used in a zero-shot setting for new domains or unseen slot values. Our model improves the joint goal accuracy by 5% (absolute) in the full-data setting and by up to 2% (absolute) in the zero-shot setting over the present state-of-the-art on the MultiWoZ 2.1 dataset.
| Year | Citations | |
|---|---|---|
Page 1
Page 1