Exploiting out-of-vocabulary words for out-of-domain detection in dialog systems

Multi-domain dialog systems often encounter user requests for out-of-domain (OOD) service. This paper focuses on detecting these requests. The proposed OOD detection method is included in a multi-domain detection component naturally. This component consists of multiple in-domain verifiers: an in-domain verifier accepts a user utterance when it belongs to the domain and rejects the utterance otherwise. To detect OOD requests without using an actual OOD corpus, the in-domain verifiers exploit the occurrence of out-of-vocabulary words. In experiments, the proposed OOD detection method was more accurate than three baseline methods.