English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Generalising dialogue state tracking (DST) to new data is especiallychallenging due to the strong reliance on abundant and fine-grained supervisionduring training. Sample sparsity, distributional shift and the occurrence ofnew concepts and topics frequently lead to severe performance degradationduring inference. In this paper we propose a training strategy to buildextractive DST models without the need for fine-grained manual span labels. Twonovel input-level dropout methods mitigate the negative impact of samplesparsity. We propose a new model architecture with a unified encoder thatsupports value as well as slot independence by leveraging the attentionmechanism. We combine the strengths of triple copy strategy DST and valuematching to benefit from complementary predictions without violating theprinciple of ontology independence. Our experiments demonstrate that anextractive DST model can be trained without manual span labels. Ourarchitecture and training strategies improve robustness towards samplesparsity, new concepts and topics, leading to state-of-the-art performance on arange of benchmarks. We further highlight our model's ability to effectivelylearn from non-dialogue data.

Robust Dialogue State Tracking with Weak Supervision and Sparse Data