ClovaCall: Korean Goal-Oriented Dialog Speech Corpus for Automatic Speech Recognition of Contact Centers

TL;DR

This paper introduces ClovaCall, a large-scale Korean call-based speech corpus for goal-oriented dialog ASR in contact centers, addressing the lack of recent Korean call corpora in this domain.

Contribution

The paper presents a new Korean call-based speech corpus with 60,000 pairs for contact center scenarios, and validates its usefulness with ASR experiments.

Findings

Effective for training Korean contact center ASR models

Large-scale dataset improves recognition accuracy

Open source release facilitates further research

Abstract

Automatic speech recognition (ASR) via call is essential for various applications, including AI for contact center (AICC) services. Despite the advancement of ASR, however, most publicly available call-based speech corpora such as Switchboard are old-fashioned. Also, most existing call corpora are in English and mainly focus on open domain dialog or general scenarios such as audiobooks. Here we introduce a new large-scale Korean call-based speech corpus under a goal-oriented dialog scenario from more than 11,000 people, i.e., ClovaCall corpus. ClovaCall includes approximately 60,000 pairs of a short sentence and its corresponding spoken utterance in a restaurant reservation domain. We validate the effectiveness of our dataset with intensive experiments using two standard ASR models. Furthermore, we release our ClovaCall dataset and baseline source codes to be available via https://github.com/ClovaAI/ClovaCall.