Search for: All records

Automatic pronunciation assessment (APA) plays an important role in providing feedback for self-directed language learners in computer-assisted pronunciation training (CAPT). Several mispronunciation detection and diagnosis (MDD) systems have achieved promising performance based on end-to-end phoneme recognition. However, assessing the intelligibility of second language (L2) remains a challenging problem. One issue is the lack of large-scale labeled speech data from non-native speakers. Additionally, relying only on one aspect (e.g., accuracy) at a phonetic level may not provide a sufficient assessment of pronunciation quality and L2 intelligibility. It is possible to leverage segmental/phonetic-level features such as goodness of pronunciation (GOP), however, feature granularity may cause a discrepancy in prosodic-level (suprasegmental) pronunciation assessment. In this study, Wav2vec 2.0-based MDD and Goodness Of Pronunciation feature-based Transformer are employed to characterize L2 intelligibility. Here, an L2 speech dataset, with human-annotated prosodic (suprasegmental) labels, is used for multi-granular and multi-aspect pronunciation assessment and identification of factors important for intelligibility in L2 English speech. The study provides a transformative comparative assessment of automated pronunciation scores versus the relationship between suprasegmental features and listener perceptions, which taken collectively can help support the development of instantaneous assessment tools and solutions for L2 learners. 

While a range of measures based on speech production, language, and perception are possible (Manun et al., 2020) for the prediction and estimation of speech intelligibility, what constitutes second language (L2) intelligibility remains under-defined. Prosodic and temporal features (i.e., stress, speech rate, rhythm, and pause placement) have been shown to impact listener perception (Kang et al., 2020). Still, their relationship with highly intelligible speech is yet unclear. This study aimed to characterize L2 speech intelligibility. Acoustic analyses, including PRAAT and Python scripts, were conducted on 405 speech samples (30 s) from 102 L2 English speakers with a wide variety of backgrounds, proficiency levels, and intelligibility levels. The results indicate that highly intelligible speakers of English employ between 2 and 4 syllables per second and that higher or lower speeds are less intelligible. Silent pauses between 0.3 and 0.8 s were associated with the highest levels of intelligibility. Rhythm, measured by Δ syllable length of all content syllables, was marginally associated with intelligibility. Finally, lexical stress accuracy did not interfere substantially with intelligibility until less than 70% of the polysyllabic words were incorrect. These findings inform the fields of first and second language research as well as language education and pathology.

 

Current leading mispronunciation detection and diagnosis (MDD) systems achieve promising performance via end-to-end phoneme recognition. One challenge of such end-to-end solutions is the scarcity of human-annotated phonemes on natural L2 speech. In this work, we leverage unlabeled L2 speech via a pseudo-labeling (PL) procedure and extend the fine-tuning approach based on pre-trained self-supervised learning (SSL) models. Specifically, we use Wav2vec 2.0 as our SSL model, and fine-tune it using original labeled L2 speech samples plus the created pseudo-labeled L2 speech samples. Our pseudo labels are dynamic and are produced by an ensemble of the online model on-the-fly, which ensures that our model is robust to pseudo label noise. We show that fine-tuning with pseudo labels achieves a 5.35% phoneme error rate reduction and 2.48% MDD F1 score improvement over a labeled-samples-only finetuning baseline. The proposed PL method is also shown to outperform conventional offline PL methods. Compared to the state-of-the-art MDD systems, our MDD solution produces a more accurate and consistent phonetic error diagnosis. In addition, we conduct an open test on a separate UTD-4Accents dataset, where our system recognition outputs show a strong correlation with human perception, based on accentedness and intelligibility.