@inproceedings{lee2019contextual,
author = {Lee, Sungjin and Shalyminov, Igor},
title = {Contextual Out of domain Utterance Handling with Counterfeit Data Augmentation},
organization = {IEEE},
booktitle = {International Conference on Acoustics, Speech, and Signal Processing 2019},
year = {2019},
month = {May},
abstract = {Neural dialog models often lack robustness to anomalous user input and produce inappropriate responses which leads to frustrating user experience. Although there are a set of prior approaches to out-of-domain (OOD) utterance detection, they share a few restrictions: they rely on OOD data or multiple sub-domains, and their OOD detection is context-independent which leads to suboptimal performance in dialog. The goal of this paper is to propose a novel OOD detection method that does not require OOD data by utilizing counterfeit OOD turns in the context of a dialog. For the sake of fostering further research, we also release new dialog datasets which are 3 publicly available dialog corpora augmented with OOD turns in a controllable way. Our method outperforms state-of-the-art dialog models equipped with a conventional OOD detection mechanism by a large margin in the presence of OOD utterances.},
url = {http://approjects.co.za/?big=en-us/research/publication/contextual-out-of-domain-utterance-handling-with-counterfeit-data-augmentation/},
}