Dialect-Aware Modeling for End-to-End Japanese Dialect Speech Recognition

摘要

In this paper, we present a novel model for building end-to-end Japanese-dialect automatic speech recognition (ASR) system. It is known that ASR systems modeling for the standard Japanese language is not suitable for recognizing Japanese dialects, which include accents and vocabulary different from standard Japanese. Therefore, we aim to produce dialect-specific end-to-end ASR systems for Japanese. Since it is difficult to collect a massive amount of speech-to-text paired data for each Japanese dialect, we utilize both dialect data and standard Japanese language data for constructing the dialect-specific end-to-end ASR systems. One primitive approach is a multi-condition modeling that simply merges the dialect data with the standardlanguage data. However, this simple multi-condition modeling causes inadequate dialect-specific characteristics to be captured because of a mismatch between the dialects and standard language. Thus, to produce reliable dialect-specific end-to-end ASR systems, we propose the dialect-aware modeling that utilizes dialect labels as auxiliary features. The main strength of the proposed method is that it effectively utilizes both dialect and standard-language data while capturing adequate dialect-specific characteristics. In our experiments using a home-made database of Japanese dialects, the proposed dialect-aware modeling out-performedthe simple multi-condition modeling and achieved an error reduction of 19.2%.