English Pronunciation Evaluation without Complex Joint Training: LoRA Fine-tuned Speech Multimodal LLM

Hosung Nam; Taekyung Ahn

REVIEW 3 major objections 6 minor 26 references

English Pronunciation Evaluation without Complex Joint Training: LoRA Fine-tuned Speech Multimodal LLM

T0 review · 3 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that a Multimodal Large Language Model adapted only through LoRA can, in a single training pass, both score pronunciation and detect mispronunciations, reaching PCC above 0.7 and phoneme error rates below 15 percent on Spe

desk verdict Useful fine-tuning comparison for pronunciation scoring, but the MDD claim is really just ASR; the paper needs a proper MDD evaluation before the joint claim holds. read the letter →

arxiv 2509.02915 v1 pith:BG7L2YJB submitted 2025-09-03 cs.CL

Taekyung Ahn , Hosung Nam This is my paper · ORCID

classification cs.CL

keywords automaticpronunciationassessmentmispronunciationdetectionanddiagnosismultimodallargelanguagemodellow-rankadaptationcomputer-assistedtrainingSpeechocean762phonemeerrorrateEnglishL2learners

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that a single pretrained multimodal large language model, adapted with low-rank (LoRA) fine-tuning, can do both jobs of computer-assisted pronunciation training at once: scoring how native-like an utterance is, and transcribing what was actually said at word and phoneme level. The authors show this on the Speechocean762 corpus of Mandarin-accented English, using prompts and task-control tokens instead of task-specific model heads, and report score correlations with human raters above 0.7 in several dimensions and word/phoneme error rates below 0.15. They also report that fine-tuning only the LoRA adapter matches or beats additionally unfreezing the audio encoder and projector, which matters because it makes the method cheap and accessible. If true, the result removes the need for the separate datasets, architectures, and training pipelines previously used for joint APA and MDD systems.

What carries the argument

The carrying mechanism is Low-Rank Adaptation (LoRA) applied to Phi-4-multimodal-instruct, a pretrained speech-language MLLM that already contains an audio encoder, an audio projector, and a mixture-of-LoRAs structure. Only the low-rank adapter weights (or, in one variant, also the audio encoder and projector) are updated during supervised fine-tuning, while the large language model stays frozen. Two control tokens, <|APA|> and <|MDD|>, are prepended to prompts to switch the same model between scoring and transcription during inference, and both outputs are produced in the pretrained model's token vocabulary (including ARPABET phoneme symbols), so no new output heads or phoneme tokens are ne

What would settle it

Run the same LoRA fine-tuning recipe on an out-of-domain L2 corpus, such as Korean- or Spanish-accented English, and check whether the accuracy-score Pearson correlation stays above 0.7; if it drops toward zero, the claimed transferability to general L2 learners is contradicted. Alternatively, replacing the pretrained audio encoder with a randomly initialized encoder and repeating the recipe would test whether the pretrained speech features are load-bearing.

Watch

Extended reading notes

Core claim

The central claim is that pronunciation assessment does not need a bespoke joint architecture: the pretrained Phi-4-multimodal-instruct model, carrying an audio encoder, projector, and LoRA adapter attached to a language model, can be taught both APA and MDD from a single dataset using only supervised fine-tuning of the adapter (or optionally the audio layers). With the Speechocean762 training set, the model predicts four pronunciation scores (accuracy, fluency, prosodic, total) in JSON and outputs orthographic and phoneme transcriptions in the same pass, with control tokens <|APA|> and <|MDD|> selecting the task. The best LoRA run produces Pearson correlations up to 0.733 with human scores

Load-bearing premise

The method assumes that the audio-processing parts of the pretrained model already understand accented non-native speech well enough that updating only a small adapter is sufficient to reach usable pronunciation scores; the paper does not validate this premise independently on accented speech.

Editorial extensions

If this is right

A single model can serve both APA and MDD in one inference pass, so a CAPT system can output a pronunciation score and a phoneme-level transcript from the same audio and prompt.
LoRA-only tuning reaches comparable or better MDD metrics than unfreezing the audio encoder and projector, so users with one GPU can adapt a large MLLM without full fine-tuning.
Because ARPABET symbols are ordinary alphabetic tokens, phoneme-level transcription can be handled without training or adding new phoneme tokens.
The negative correlation between predicted accuracy scores and phoneme error rate supports the interpretation that the scored output is tied to actual mispronunciation detection, not just dataset statistics.
The same prompt-and-control-token recipe should transfer to other learner populations and languages using the model's multilingual capabilities, though the paper only tests Mandarin-L1 English speakers.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A testable extension of the paper's setup is to apply the same LoRA-trained adapter to other accented-English corpora (for example, Korean or Spanish L1) with no additional training; if the pretrained audio encoder is as transferable as assumed, accuracy-score correlation should remain in the same range.
The control-token mechanism could generalize to other CAPT subskills, such as stress, intonation, or rhythm scoring, by defining new tokens and rubrics instead of adding new output heads; the paper does not implement this.
The reported negative correlation between predicted accuracy scores and phoneme error rate could be operationalized as an explainability signal: a deployment could generate learner feedback by diffing the model's phoneme transcript against the target, something the paper leaves implicit.
The LoRA-only versus unfreeze comparison covers only four epochs, so the relative advantage of LoRA-only on MDD metrics is specific to that training budget; longer unfreezing is an open experimental question.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: this is an honest, straightforward fine-tuning study showing that LoRA-only adaptation of Phi-4-multimodal gives reasonable pronunciation-scoring correlations on Speechocean762, and that unfreezing the audio encoder doesn't help. The part that doesn't hold up is the claim that the same setup also does MDD. The evaluation never feeds the model the target sentence, so the WER/PER/F1 numbers are ASR/phone-recognition metrics, not mispronunciation detection.

What's genuinely useful: the control-token setup for APA vs MDD in one model, the direct comparison of LoRA-only vs unfreezing audio layers (the null result is informative), and use of the standard train/test split. The authors also deserve credit for the limitation section; they admit no full fine-tuning and no suprasegmental modeling.

Soft spots, in order. First, the MDD evaluation in Section 4.3 is the big one. Speechocean762 has mispronunciation annotations, but the model is prompted to 'transcribe the audio utterance' without the intended sentence. That is ASR, not MDD. The precision/recall/F1 are never defined against the dataset's labels; the formulas define TP/FP conceptually but no alignment or scoring protocol is given. So Table 3's F1 of 0.724 is plausibly just a phoneme recognition number. The abstract's 'simultaneous APA and MDD' claim rests on this, which means the central claim is overstated. Second, the abstract says 'PCC > 0.7' but accuracy and total PCC for LoRA are below 0.7 (0.54-0.68). The strong correlation holds only for fluency/prosodic at the best epoch, not all metrics. Minor but sloppy. Third, Table 4 compares both models at 4 epochs even though LoRA peaks at epoch 3; that slightly biases the SOTA comparison in favor of Unfreeze for accuracy. They should compare best epochs or report all. Fourth, no code, no uncertainty bars, and apparently single runs. Common for this type of paper, but noted. The correlation analysis in Figure 5 is descriptive and weak, but clearly labeled as such, so fine.

Bottom line: the APA half is evidence, the MDD half isn't. The paper deserves a serious referee because the core question is worth asking and the fix is straightforward: redo MDD evaluation with target sentences and the dataset's mispronunciation labels. Without that, this is a solid demo of pronunciation scoring, not a joint APA+MDD system. It is the kind of paper that would benefit from peer review rather than desk rejection; I would send it out and require the MDD evaluation to be reworked or the claim softened.

Referee Report

3 major / 6 minor

Summary. The paper proposes a unified automatic pronunciation assessment (APA) and mispronunciation detection and diagnosis (MDD) system by LoRA fine-tuning the multimodal LLM Phi-4-multimodal-instruct on the Speechocean762 dataset. The model is trained with task-specific control tokens and prompts to output four sentence-level pronunciation scores (accuracy, fluency, prosodic, total) as well as word-level and phoneme-level transcriptions. The authors compare two fine-tuning strategies: LoRA-only and unfreezing the audio encoder/projector. They report Pearson correlation coefficients up to 0.733 for fluency, WER down to 0.139, PER down to 0.114, and F1 up to 0.724, concluding that LoRA-only tuning is comparable to audio-layer unfreezing. They also report a moderate negative correlation between PER and accuracy scores.

Significance. If the MDD claim were properly validated, the contribution would be a simple, resource-efficient recipe for building a unified CAPT system from an MLLM, and the LoRA-only versus unfreezing comparison would be a useful practical datapoint. The APA results are plausible and competitive on several dimensions, and the paper is honest about its computational limitations. However, the MDD evaluation as presented does not measure mispronunciation detection or diagnosis: the model is prompted to transcribe, not to compare against a target sentence, and no alignment or scoring protocol against the Speechocean762 mispronunciation annotations is described. The abstract also overstates the correlation results. The central claim of simultaneous APA and MDD is therefore not yet supported; the paper's strongest defensible contribution is an APA-scoring and general ASR experiment on a public benchmark.

major comments (3)

[§4.3 and Appendix 7.2] The MDD evaluation does not validate the claimed MDD capability. The MDD prompt is 'Transcribe the audio utterance, providing both a word-level transcript and phoneme-level breakdown.' No target sentence or canonical phoneme sequence is provided, and no alignment or scoring protocol is described. The F1 formulas in (3)-(5) define TP/FP in terms of 'correctly identified mispronunciations,' but the paper never explains how a mispronunciation is identified from an unconstrained transcription against the dataset's reference phoneme annotations. Consequently, the reported WER, PER, and F1 in Table 3 are measures of ASR/phoneme-recognition performance, not of mispronunciation detection or diagnosis. The MDD half of the central claim is therefore unsupported.
[Abstract and §4.4] The abstract's statement 'PCC > 0.7' is not supported by Table 3 for the LoRA model as a whole: Table 3 gives LoRA total PCC 0.668, accuracy 0.645, and only fluency and prosodic PCC above 0.7 at some epochs. Similarly, §4.4's claim that 'The LoRA approach consistently yielded superior results compared to the Unfreeze strategy across all metrics' is contradicted by Table 3: at epoch 4 Unfreeze accuracy is 0.743 versus LoRA 0.645, and Unfreeze WER is 0.142 versus LoRA 0.148. The data support 'comparable' or 'mixed,' not 'consistently superior.'
[§4.2-§4.4 and Table 4] The paper does not describe a validation-based epoch selection procedure. Table 3 reports all epochs, but §4.4 and Table 4 compare 'four epochs' (LoRA and Unfreeze) even though the best LoRA APA/MDD numbers occur at epochs 2-3. If hyperparameters or epochs are selected using the test set, the reported comparisons are biased; if not, the selection rule should be stated. This matters for the central LoRA-versus-unfreeze comparison and for the comparison with prior work in Table 4.

minor comments (6)

[Table 3 caption] The caption states 'All p-values for APA tasks are under 0.05, except for underlined values,' but the table does not display p-values or underlining. Either include them or remove the claim.
[§4.2] The training description gives batch size 8 and gradient accumulation step 8, but not the effective batch size, number of training steps, warmup, or learning-rate schedule. Please add these details for reproducibility.
[§2.4] The text says Phi-4 'incorporates a Mixture-of-LoRAs structure,' but the proposed method then fine-tunes LoRA adapters. Clarify whether the authors add new LoRA adapters on top of the model's existing Mixture-of-LoRAs or replace/reuse them.
[§4.1] The exclusion of the 'completeness' metric because all test values are 10 is plausible, but the sentence should also state whether the same was true for the training set, since excluding a metric from training on this basis is unusual.
[Table 2 and Appendix 7.1] The label 'prosodic' in the appendix and 'prosody' in Table 2 are used inconsistently. Please unify the terminology.
[Throughout] The paper uses 'simultaneously' for APA and MDD, but the two tasks are triggered by different control tokens and prompts. Clarify whether 'simultaneous' means a single shared model/training run rather than parallel generation in one forward pass.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the paper fine-tunes on the Speechocean762 train split and evaluates on the held-out test split; all reported numbers are measured, not fitted, and there are no load-bearing self-citations.

full rationale

The paper's derivation chain is a standard supervised fine-tuning pipeline. APA scores are trained on the Speechocean762 training split and evaluated by Pearson correlation against human scores on the test split; MDD metrics (WER, PER, F1) are computed by comparing model outputs to reference transcriptions from the same held-out test split. No parameter is fitted to the test set, and no 'prediction' is obtained by construction from its own input. The reliance on Phi-4-multimodal-instruct's pretrained audio encoder is an external design choice, not a self-citation, and the cited technical report [1] does not share authors with this paper. The correlation analysis between PER and accuracy scores is descriptive and not used to derive the central claim. The strongest substantive concern in the paper is that the MDD evaluation may conflate ASR/phoneme recognition with mispronunciation detection, since the MDD prompt asks for transcription and the F1 definitions are not operationalized against the dataset's mispronunciation annotations. That is a construct-validity or evaluation-design issue, not a circularity: the reported numbers are still measured on held-out data and do not reduce to the training targets by definition. Under the specific circularity criteria (self-definition, fitted-input-called-prediction, load-bearing self-citation, imported uniqueness, ansatz-via-citation, renaming), no circular step is present.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The paper relies on a pretrained MLLM's representations and on a standard annotated corpus. It introduces no new entities or fitted constants; the learned LoRA weights are the result of training, not assumed inputs. Standard hyperparameters (learning rate, batch size) are not counted as free parameters because they are not fitted to the test set and do not enter the central claim as tunable constants.

assumptions (3)

domain assumption The pretrained Phi-4-multimodal-instruct audio encoder and projector already encode task-relevant acoustic features for non-native English, so only LoRA adaptation (or unfreezing audio layers) is needed.
Taken from the Phi-4 technical report [1] and invoked in Section 4.2: 'the audio encoder and projector have been sufficiently trained... eliminating the need for additional training.' This is not independently validated for L2 accented speech in this paper.
domain assumption Speechocean762 human scores are reliable and consistent ground truth for pronunciation assessment.
The dataset's human annotations are treated as gold labels; any noise or bias in these scores directly limits the PCC ceiling. Not validated in this paper.
domain assumption The CMUDict phone set and the dataset's reference phone transcriptions are correct references for MDD evaluation.
PER and F1 are computed against these references; if the references are flawed, the MDD metrics are misleading.

how reviews work

0 comments

Cite this review

Pith. "Pith review of English Pronunciation Evaluation without Complex Joint Training: LoRA Fine-tuned Speech Multimodal LLM." pith.science (2026). https://pith.science/paper/BG7L2YJB

@misc{pith2026250902915,
  author       = {Pith},
  title        = {Pith review of: English Pronunciation Evaluation without Complex Joint Training: LoRA Fine-tuned Speech Multimodal LLM},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BG7L2YJB}},
  note         = {Machine review of arXiv:2509.02915}
}

read the original abstract

This study demonstrates that a Multimodal Large Language Model (MLLM) adapted via Low-Rank Adaptation (LoRA) can perform both Automatic Pronunciation Assessment (APA) and Mispronunciation Detection and Diagnosis (MDD) simultaneously. Leveraging Microsoft's Phi-4-multimodal-instruct, our fine-tuning method eliminates the need for complex architectural changes or separate training procedures conventionally required for these distinct tasks. Fine-tuned on the Speechocean762 dataset, the pronunciation evaluation scores predicted by the model exhibited a strong Pearson Correlation Coefficient (PCC > 0.7) with human-assigned scores, while achieving low Word Error Rate (WER) and Phoneme Error Rate (PER) (both < 0.15). Notably, fine-tuning only the LoRA layers was sufficient to achieve performance levels comparable to those achieved by fine-tuning all audio layers. This research highlights that an integrated pronunciation assessment system can be established by adapting large multimodal models without full fine-tuning, utilizing a significantly simpler training methodology compared to previous joint models designed for simultaneous APA and MDD. This efficient LoRA-based approach paves the way for more accessible, integrated, and effective Computer-Assisted Pronunciation Training (CAPT) technologies for English L2 learners.

Figures

Figures reproduced from arXiv: 2509.02915 by the authors.

**Figure 1.** The overview of our proposed method. [20], revealing a significant correlation between these two functionalities. However, these approaches still require separate datasets and independent model architectures for each task, necessitating distinct training procedures and computational resources despite their functional interdependence. Recently, LLMs (Large Language Models) have exhibited significant advancements in… view at source ↗

**Figure 2.** The Comparison of our model and past studies. [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** The methods for using the prompts and control tokens. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: The method of unfreezing layers. While conventional practice typically limits fine-tuning to LoRA adapter layers, the official documentation [17] suggests potential benefits from unfreezing all audio layers. For this reason, our experiment directly compares these appr…

**Figure 5.** Figure 5: Pearson Correlation Coefficients between Accuracy Score and Phoneme Error Rate. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 20 canonical work pages

[1]

Abouelenin, A

A. Abouelenin, A. Ashfaq, A. Atkinson, H. Awadalla, N. Bach, J. Bao, et al

work page
[2]

A. Babu, C. Wang, A. Tjandra, K. Lakhotia, Q. Xu, N. Goyal, K. Singh, P. von Platen, Y. Saraf, J. Pino, A. Baevski, A. Conneau, and M. Auli. 2022. XLS-R: Self-supervised cross-lingual speech representation learning at scale. In Proc. Interspeech 2022. 2278–2282. doi:10.21437/Interspeech.2022-143

doi:10.21437/interspeech.2022-143 2022
[3]

Baevski, Y

A. Baevski, Y. Zhou, A. Mohamed, and M. Auli. 2020. Wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations. In Advances in Neural Information Processing Systems, Vol. 33. 12449–12460

work page 2020
[4]

Bernstein, M

J. Bernstein, M. Cohen, H. Murveit, D. Rtischev, and M. Weintraub. 1990. Au- tomatic evaluation and training in English pronunciation. In Proceedings of the ICSLP-90: 1990 International Conference on Spoken Language Processing . 1185– 1188

work page 1990
[5]

Conneau, A

A. Conneau, A. Baevski, R. Collobert, A. Mohamed, and M. Auli. 2021. Unsu- pervised cross-lingual representation learning for speech recognition. In Proc. Interspeech 2021. 2426–2430. doi:10.21437/Interspeech.2021-329

doi:10.21437/interspeech.2021-329 2021
[6]

Eskenazi

M. Eskenazi. 2009. An overview of spoken language technology for education. Speech Communication 51, 10 (2009), 832–844

work page 2009
[7]

Y. Feng, G. Fu, Q. Chen, and K. Chen. 2020. SED-MDD: Towards sentence dependent end-to-end mispronunciation detection and diagnosis. In ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 3492–3496

work page 2020
[8]

Fouz-González

J. Fouz-González. 2015. Trends and directions in computer-assisted pronunciation training. In Investigating English pronunciation: Trends and directions . 314–342

work page 2015

Show all 26 references

[9]

Franco, V

H. Franco, V. Abrash, K. Precoda, H. Bratt, R. Rao, and J. Butzberger. 2000. The SRI EduSpeak system: recognition and pronunciation scoring for language learning. In Proceedings of Intelligent Speech Technology in Language Learning, InSTiLL- 2000

2000
[10]

Franco, L

H. Franco, L. Neumeyer, V. Digalakis, and O. Ronen. 2000. Combination of machine scores for automatic grading of pronunciation quality. Speech Commu- nication 30 (2000), 121–130

2000
[11]

K. Fu, L. Peng, N. Yang, and S. Zhou. 2024. Pronunciation Assessment with Multi-modal Large Language Models. arXiv preprint arXiv:2407.09209 (2024)

2024 arXiv
[12]

Getman, R

Y. Getman, R. Al-Ghezi, K. Voskoboinik, T. Grósz, M. Kurimo, G. Salvi, T. Svend- sen, and S. Strömbergsson. 2022. Wav2vec2-Based Speech Rating System for Children with Speech Sound Disorder. In Interspeech 2022. ISCA

2022
[13]

Y. Gong, Z. Chen, I. H. Chu, P. Chang, and J. Glass. 2022. Transformer-based multi- aspect multi-granularity non-native english speaker pronunciation assessment. In ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 7262–7266

2022
[14]

J. Heo. 2024. Practical AI application development using LLMs . onlybook

2024
[15]

Y. Hong. 2021. Automated evaluation of pronunciation for L2 English learners . Ph. D. Dissertation. Graduate School, Korea University

2021
[16]

Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, et al. 2024. Gpt-4o system card. arXiv preprint arXiv:2410.21276 (2024)

2024 arXiv
[17]

Microsoft. 2025. Phi-4-multimodal-instruct. Hugging Face. https://huggingface. co/microsoft/Phi-4-multimodal-instruct

2025
[18]

L. Peng, K. Fu, B. Lin, D. Ke, and J. Zhang. 2021. A Study on Fine-Tuning wav2vec2. 0 Model for the Task of Mispronunciation Detection and Diagnosis. InInterspeech. ISCA

2021
[19]

X. Qian, H. Meng, and F. Soong. 2010. The use of DBN-HMMs for mispro- nunciation detection and diagnosis in L2 English to support computer-aided pronunciation training. In Proceedings of INTERSPEECH . 775–778

2010
[20]

H. Ryu, S. Kim, and M. Chung. 2023. A Joint Model for Pronunciation Assessment and Mispronunciation Detection and Diagnosis with Multi-task Learning. In Proc. Interspeech 2023. 959–963. doi:10.21437/Interspeech.2023-337

2023 doi
[21]

K. Wang, L. He, K. Liu, Y. Deng, W. Wei, and S. Zhao. 2025. Exploring the Potential of Large Multimodal Models as Effective Alternatives for Pronunciation Assessment. arXiv preprint arXiv:2503.11229 (2025)

2025 arXiv
[22]

Weide et al

R. Weide et al. 1998. The Carnegie Mellon pronouncing dictionary (Release 0.6). Retrieved from www.cs.cmu.edu

1998
[23]

S. M. Witt and S. J. Young. 2000. Phone-level pronunciation scoring and assess- ment for interactive language learning. Speech Communication 30, 2-3 (2000), 95–108

2000
[24]

M. Yang, K. Hirschi, S. D. Looney, O. Kang, and J. H. L. Hansen. 2022. Improving Mispronunciation Detection with Wav2vec2-Based Momentum Pseudo-Labeling for Accentedness and Intelligibility Assessment. In Interspeech 2022. ISCA

2022
[25]

Zhang, Z

J. Zhang, Z. Zhang, Y. Wang, Z. Yan, Q. Song, Y. Huang, et al . 2021. spee- chocean762: An open-source non-native english speech corpus for pronunciation assessment. arXiv preprint arXiv:2104.01378 (2021)

2021 arXiv
[2025]

arXiv preprint arXiv:2503.01743 (2025)

Phi-4-Mini Technical Report: Compact yet Powerful Multimodal Language Models via Mixture-of-LoRAs. arXiv preprint arXiv:2503.01743 (2025)

2025 arXiv

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

Abouelenin, A

A. Abouelenin, A. Ashfaq, A. Atkinson, H. Awadalla, N. Bach, J. Bao, et al

work page

[2] [2]

A. Babu, C. Wang, A. Tjandra, K. Lakhotia, Q. Xu, N. Goyal, K. Singh, P. von Platen, Y. Saraf, J. Pino, A. Baevski, A. Conneau, and M. Auli. 2022. XLS-R: Self-supervised cross-lingual speech representation learning at scale. In Proc. Interspeech 2022. 2278–2282. doi:10.21437/Interspeech.2022-143

doi:10.21437/interspeech.2022-143 2022

[3] [3]

Baevski, Y

A. Baevski, Y. Zhou, A. Mohamed, and M. Auli. 2020. Wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations. In Advances in Neural Information Processing Systems, Vol. 33. 12449–12460

work page 2020

[4] [4]

Bernstein, M

J. Bernstein, M. Cohen, H. Murveit, D. Rtischev, and M. Weintraub. 1990. Au- tomatic evaluation and training in English pronunciation. In Proceedings of the ICSLP-90: 1990 International Conference on Spoken Language Processing . 1185– 1188

work page 1990

[5] [5]

Conneau, A

A. Conneau, A. Baevski, R. Collobert, A. Mohamed, and M. Auli. 2021. Unsu- pervised cross-lingual representation learning for speech recognition. In Proc. Interspeech 2021. 2426–2430. doi:10.21437/Interspeech.2021-329

doi:10.21437/interspeech.2021-329 2021

[6] [6]

Eskenazi

M. Eskenazi. 2009. An overview of spoken language technology for education. Speech Communication 51, 10 (2009), 832–844

work page 2009

[7] [7]

Y. Feng, G. Fu, Q. Chen, and K. Chen. 2020. SED-MDD: Towards sentence dependent end-to-end mispronunciation detection and diagnosis. In ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 3492–3496

work page 2020

[8] [8]

Fouz-González

J. Fouz-González. 2015. Trends and directions in computer-assisted pronunciation training. In Investigating English pronunciation: Trends and directions . 314–342

work page 2015