Investigation of Automatic Speech Recognition Systems via the Multilingual Deep Neural Network Modeling Methods for a Very Low-Resource Language, Chaha

摘要

Automatic speech recognition (ASR) is vital for verylow-resource languages for mitigating the extinction trouble. Chaha is one ofthe low-resource languages, which suffers from the problem of resourceinsufficiency and some of its phonological, morphological, and orthographicfeatures challenge the development and initiatives in the area of ASR. Byconsidering these challenges, this study is the first endeavor, which analyzedthe characteristics of the language, prepared speech corpus, and developed differentASR systems. A small 3-hour read speech corpus was prepared and transcribed.Different basic and rounded phone unit-based speech recognizers were exploredusing multilingual deep neural network (DNN) modeling methods. The experimentalresults demonstrated that all the basic phone and rounded phone unit-basedmultilingual models outperformed the corresponding unilingual models with therelative performance improvements of 5.47% to 19.87% and 5.74% to 16.77%,respectively. The rounded phone unit-based multilingual models outperformed theequivalent basic phone unit-based models with relative performance improvementsof 0.95% to 4.98%. Overall, we discovered that multilingual DNN modelingmethods are profoundly effective to develop Chaha speech recognizers. Both thebasic and rounded phone acoustic units are convenient to build Chaha ASRsystem. However, the rounded phone unit-based models are superior inperformance and faster in recognition speed over the corresponding basic phoneunit-based models. Hence, the rounded phone units are the most suitableacoustic units to develop Chaha ASR systems.