Robust Dialogue State Tracking with Weak Supervision and Sparse Data
Robust Dialogue State Tracking with Weak Supervision and Sparse Data
About this item
Full title
Author / Creator
Publisher
One Broadway, 12th Floor, Cambridge, Massachusetts 02142, USA: MIT Press
Journal title
Language
English
Formats
Publication information
Publisher
One Broadway, 12th Floor, Cambridge, Massachusetts 02142, USA: MIT Press
Subjects
More information
Scope and Contents
Contents
Generalizing dialogue state tracking (DST) to new data is especially challenging due to the strong reliance on abundant and fine-grained supervision during training. Sample sparsity, distributional shift, and the occurrence of new concepts and topics frequently lead to severe performance degradation during inference. In this paper we propose a trai...
Alternative Titles
Full title
Robust Dialogue State Tracking with Weak Supervision and Sparse Data
Authors, Artists and Contributors
Identifiers
Primary Identifiers
Record Identifier
TN_cdi_proquest_journals_2893948503
Permalink
https://devfeature-collection.sl.nsw.gov.au/record/TN_cdi_proquest_journals_2893948503
Other Identifiers
ISSN
2307-387X
E-ISSN
2307-387X
DOI
10.1162/tacl_a_00513