Towards Textual Out-of-Domain Detection Without In-Domain Labels
| Autor: | Di Jin, Shuyang Gao, Seokhwan Kim, Yang Liu, Dilek Hakkani-Tur |
|---|---|
| Rok vydání: | 2022 |
| Předmět: |
FOS: Computer and information sciences
Computer Science - Machine Learning Computational Mathematics Computer Science - Computation and Language Artificial Intelligence (cs.AI) Acoustics and Ultrasonics Computer Science - Artificial Intelligence Computer Science (miscellaneous) ComputerApplications_COMPUTERSINOTHERSYSTEMS Electrical and Electronic Engineering Computation and Language (cs.CL) Machine Learning (cs.LG) |
| Zdroj: | IEEE/ACM Transactions on Audio, Speech, and Language Processing. 30:1386-1395 |
| ISSN: | 2329-9304 2329-9290 |
| DOI: | 10.1109/taslp.2022.3162081 |
| Popis: | In many real-world settings, machine learning models need to identify user inputs that are out-of-domain (OOD) so as to avoid performing wrong actions. This work focuses on a challenging case of OOD detection, where no labels for in-domain data are accessible (e.g., no intent labels for the intent classification task). To this end, we first evaluate different language model based approaches that predict likelihood for a sequence of tokens. Furthermore, we propose a novel representation learning based method by combining unsupervised clustering and contrastive learning so that better data representations for OOD detection can be learned. Through extensive experiments, we demonstrate that this method can significantly outperform likelihood-based methods and can be even competitive to the state-of-the-art supervised approaches with label information. Comment: Accepted by IEEE/ACM Transactions on Audio Speech and Language |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|