Abstract
Dialog State Tracking (DST) is a core component in task-oriented dialogsystems. Existing approaches for DST usually fall into two categories, i.e, thepicklist-based and span-based. From one hand, the picklist-based methodsperform classifications for each slot over a candidate-value list, under thecondition that a pre-defined ontology is accessible. However, it is impracticalin industry since it is hard to get full access to the ontology. On the otherhand, the span-based methods track values for each slot through finding textspans in the dialog context. However, due to the diversity of valuedescriptions, it is hard to find a particular string in the dialog context. Tomitigate these issues, this paper proposes a Dual Strategy for DST (DS-DST) toborrow advantages from both the picklist-based and span-based methods, byclassifying over a picklist or finding values from a slot span. Empiricalresults show that DS-DST achieves the state-of-the-art scores in terms of jointaccuracy, i.e., 51.2% on the MultiWOZ 2.1 dataset, and 53.3% when the fullontology is accessible.