Abstract
Having the ability to predict request alternatives (reqalts) user acts accurately is essential to tracking dialog state, especially when users are allowed to change their mind during the dialog. These can be detected reasonably well using ASR n-grams and additional features derived from dialog context on the Dialog State Tracking Challenge corpus. We are somewhat surprised at the high detection F1 score (90 %), and find that with transcripts we can obtain almost perfect accuracy (F1 \(=\) 99 %). There may be a wording bias introduced during the data collection process, which implies that the task may not generalize beyond the corpus.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Since the system does not know how long an ongoing dialog will last, we apply this non-linear transformation to approximate the end of the dialog.
- 2.
The corpus is available from DSTC 2 and 3 website at http://camdial.org/~mh521/dstc/. Labelled system log files (no audio) for the dialogs are provided.
- 3.
The dataset gives both system’s live ASR hypotheses and off-line batch ASR hypotheses for each user utterance. We only use the top live ASR hypothesis in all conducted experiments.
- 4.
We assume this is how the research conductor collected the dataset.
- 5.
4.3 % of the time the user issued a reqalts act when the task description does not require a goal change.
References
Henderson M, Thomson B, Williams J (2013) Dialog state tracking challenge 2 & 3
Ma Y (2013) User goal change model for spoken dialog state tracking. NAACL HLT SRW, p 91
Manning C, Klein D (2003) Optimization, maxent models, and conditional estimation without magic. In: Proceedings of the 2003 conference of the north american chapter of the association for computational linguistics on human language technology: tutorials-volume 5, pp 8–8. Association for Computational Linguistics
Young S, Gašić M, Keizer S, Mairesse F, Schatzmann J, Thomson B, Yu K (2010) The hidden information state model: a practical framework for pomdp-based spoken dialogue management. Comput Speech Lang 24(2):150–174
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2016 Springer International Publishing Switzerland
About this chapter
Cite this chapter
Ma, Y., Fosler-Lussier, E. (2016). Detecting ‘Request Alternatives’ User Dialog Acts from Dialog Context. In: Rudnicky, A., Raux, A., Lane, I., Misu, T. (eds) Situated Dialog in Speech-Based Human-Computer Interaction. Signals and Communication Technology. Springer, Cham. https://doi.org/10.1007/978-3-319-21834-2_9
Download citation
DOI: https://doi.org/10.1007/978-3-319-21834-2_9
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-21833-5
Online ISBN: 978-3-319-21834-2
eBook Packages: EngineeringEngineering (R0)