Evaluation of English Speech Recognition for Japanese Learners Using DNN-Based Acoustic Models

Language modeling in speech recognition for grammatical error detection based on neural machine translation

“…To evaluate the proposed method for language modeling in ASR, we fixed the acoustic model referring to our previous work [4]. This DNN-based AM was trained using the ERJ corpus with the conditions shown in Table 3.…”

“…The quality of ASR depends on its two main components: acoustic models (AMs) and language models (LMs). So far, with the widespread application of deep learning, a large number of research studies on deep neural network (DNN)based acoustic models have increasingly improved the speech recognition rate for non-native speech [3,4]. However, in these ordinary DNN-based systems, less attention has been paid to establishing LMs in ASR for improving the recognition of L2 speech with grammatical errors.…”

Complexities Faced by Non-native Learners in Accuracy and Fluency in English Language speaking at BS Level

“…Students who are unable to communicate fluently in the target language will struggle to do so in that language. There is a possibility that students' and their professors' viewpoints on education may not always coincide 100% of the time (Fu et al, 2018). Because of this, it is impossible for them to put their educational ideas into practice in the classroom because those ideas are in direct competition with one another.…”