Pronunciation Deviation Analysis Through Voice Cloning and Acoustic Comparison

Amir Ghasemkhani; Andrew Valdivia; Hailu Xu; Xin Qin; Yueming Zhang

REVIEW 4 major objections 4 minor 12 references

Pronunciation Deviation Analysis Through Voice Cloning and Acoustic Comparison

T0 review · 4 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Mispronounced words can be found by comparing a learner's own speech to a voice-cloned version of the same words spoken correctly in the learner's voice.

desk verdict Promising reference-signal idea that never verifies its own reference signal; the reported evaluation is in-sample and below trivial baselines for three of four speakers. read the letter →

arxiv 2507.10985 v1 pith:PNU4CLIG submitted 2025-07-15 cs.SD cs.AIeess.AS

Andrew Valdivia , Yueming Zhang , Hailu Xu , Amir Ghasemkhani , Xin Qin This is my paper · ORCID

classification cs.SDcs.AIeess.AS

keywords pronunciationanalysisvoicecloningspeechprocessinglanguagelearningacousticcomparisonmispronunciationdetectiondynamictimewarpingMFCCfeatures

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

The paper proposes that a language learner's mispronunciations can be located without phonetic rules or per-language training data by measuring the acoustic distance between the learner's original utterance and a voice-cloned version of the same words spoken correctly in the learner's own voice. The hypothesis is that the regions of maximal acoustic deviation between the original and cloned audio mark the mispronounced segments. The authors test this on annotated non-native English speech, generating clones of each utterance, aligning the real and cloned audio word by word, and computing frame-level distances between their MFCC (mel-frequency cepstral coefficient) feature envelopes. Mispronounced words show systematically larger distances than correctly pronounced words, and a per-word threshold labels speech as correct, incorrect, or ambiguous. If the approach holds, pronunciation feedback becomes personalized and language-agnostic, needing a text-to-speech clone of the learner rather than expert phonetic rules.

What carries the argument

The load-bearing mechanism is the pairing of a voice clone with a normalized multi-feature DTW distance. The clone is produced by a commercial text-to-speech service that reproduces the learner's voice while reading the target text with correct pronunciation, giving each learner a personalized 'correct version' of their own utterance. For each aligned word, 13-dimensional MFCC (mel-frequency cepstral coefficient) envelopes are extracted from both versions, the two time series are resampled to a common length, and a dynamic time warping distance (the minimal accumulated cost of aligning two time series) is computed per coefficient and averaged, with normalization by duration and dimensionality. Calibration fits kernel density estimates and 90th-percentile thresholds to the distances of correctly and incorrectly pronounced training words; at runtime each new word's distance $\bar{d}_j$ selects a class-appropriate threshold, yielding a CORRECT, INCORRECT, or AMBIGUOUS verdict. The whole argument depends on these deviations being interpretable as pronunciation error rather than voice idiosyncrasy or synthesis artifact.

What would settle it

Force-align or automatically transcribe the cloned audio for a word the learner is known to mispronounce, such as the 'fraud' case study: if the clone's transcription contains the same substituted phone as the learner's production (or phones that do not match the canonical form), the reference signal is corrupted and the deviation-based labels cannot be trusted. A cheaper quantitative check: if the distance distributions for correctly pronounced words and mispronounced words become statistically indistinguishable once the clone is generated by a different TTS voice or cloning recipe, the discriminative signal was an artifact of one synthesis pipeline.

Watch

Extended reading notes

Core claim

The paper's central claim is that a voice-cloned, correctly pronounced version of a learner's own utterance can serve as a personalized reference signal: the words whose real and cloned versions diverge most in acoustic feature space are the words the learner mispronounced. Concretely, the authors compute, for each word, a multi-feature dynamic time warping (DTW) distance between the 13-dimensional MFCC envelopes of the original and the synthesized audio, calibrate class-specific thresholds from annotated training distances, and then label unseen words as CORRECT, INCORRECT, or AMBIGUOUS depending on where the distance falls relative to the threshold selected by kernel density classification. Across four non-native speakers, mispronounced words exhibit larger mean distances than correctly pronounced ones, and the model's per-word labels track human-annotated errors; a worked case study on the word 'fraud', where the speaker's vowel is realized as /AW1/ instead of the target /AO1/, shows the kind of first-language interference error the method flags. The authors offer this as a proof of concept that mispronunciation detection can bypass predefined phonetic rule inventories and avoid extensive per-language training data.

Load-bearing premise

The entire pipeline rests on the unverified premise that the cloned audio actually pronounces the target words correctly while preserving the learner's voice; if the clone repeats the learner's mispronunciation or introduces its own artifacts, the measured acoustic deviation points at something other than pronunciation error.

Editorial extensions

If this is right

Pronunciation feedback becomes available for any language or dialect that a text-to-speech service can read and any voice it can clone, with no phonetic rule inventory required.
Because the calibration is learned from annotated distances rather than hand-built rules, the same pipeline transfers across learners with different first languages without retraining per language.
The three-way decision rule lets a tutoring system abstain from feedback on words whose acoustic evidence is inconclusive, rather than forcing a guess.
Per-speaker distance distributions differ enough that individualizing the model improves results, pointing toward fully personalized calibration as the next step.
The runtime cost is limited to cloning an utterance, aligning words, and computing frame distances, which the authors identify as the basis for eventual real-time interactive practice.

Reading between the lines

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

Since the paper never phonetically transcribes the cloned audio, a direct check of the foundation would be to force-align the clone: if the clone reproduces the learner's substituted phone rather than the canonical one (or adds its own artifacts), the deviation signal would measure something other than pronunciation error.
The per-coefficient DTW distances could double as an error typology: substitution errors should shift the spectral envelope shape across many coefficients, whereas deletions and additions should show up mainly as timing and alignment discontinuities, potentially separating error types without phonetic labels.
Because the reference is the learner's own voice, the same deviation geometry could expose prosodic problems such as misplaced stress or non-native rhythm, which phoneme-inventory methods cannot see; the paper lists this as future work.
If clone correctness is verified, the approach becomes a zero-shot pronunciation checker for under-resourced languages: any language a TTS system can speak and any learner whose voice can be cloned gets a personalized reference without annotated training data.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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

Colleague—

Short version: this is a clearly written paper with an appealing idea, but the load-bearing assumption that ElevenLabs cloning a learner's utterance from text produces a corrected, same-voice pronunciation is never validated, and the experiments as described are in-sample with no baselines. I would not trust the numbers, and the approach needs substantial rework before it warrants publication.

What is actually new: the specific pairing of a commercial voice-cloning API with DTW-based frame-level deviation to localize mispronunciations is an incremental but useful twist on prior work (Das & Gutierrez-Osuna; Nguyen et al.). The paper honestly cites those, and it does not pretend the broad “compare to a synthetic native-like reference” idea is new. The pipeline is well structured, and the qualitative “Fraud” example is a nice illustration of what the method could do.

The soft spots are serious. First, the core premise: the clone is assumed to contain the corrected pronunciation. There is no systematic verification—no phonetic transcription of the synthesized audio, no checks across speakers or phonemes. One manually inspected word (“Fraud”) is not evidence that the API reliably fixes mispronunciations. If the clone preserves the learner's error or introduces artifacts, the measured distances mean nothing.

Second, the evaluation is circular as reported. D_correct and D_incorrect are built using ground-truth labels, and the thresholds and KDEs are fit to those same distances. The Experiments section never states a held-out split, so Table 1 likely overstates performance. Even as reported, three of four speakers are below majority-class accuracy. That is a red flag.

Third, there are no baselines—no comparison to GOP scoring or any existing CAPT method. The claim of effectiveness is unsupported.

What is salvageable: the idea of a personalized, voice-preserving reference signal is worth exploring. But the authors need to (1) validate the clone's phonetic correctness, (2) run a proper train/test split with cross-validation, (3) add baselines, and (4) report per-speaker majority-class performance honestly.

Who it is for: someone working on CAPT or accent modification might find the concept worth a skim, but the manuscript currently should not be cited as evidence.

Recommendation: if I were the editor, I would desk-reject this version. The central assumption is unverified, and the experimental results are weak. I would encourage the authors to address the validation and evaluation issues and resubmit.

Referee Report

4 major / 4 minor

Summary. The paper proposes a mispronunciation detection method that compares a learner's original utterance with a voice-cloned version of the same text generated by ElevenLabs, under the assumption that the clone preserves the learner's voice but corrects pronunciation. For each word, the method extracts MFCC envelopes from both audio streams, resamples them to a common length, computes per-coefficient DTW distances, and averages them into a normalized distance. In a calibration phase, these distances are partitioned by ground-truth correctness labels, used to fit class-conditional KDEs (K_COR, K_INC) and to set class-specific 90th-percentile thresholds (tau_C, tau_I). At runtime, a word is classified as CORRECT, INCORRECT, or AMBIGUOUS depending on which KDE is larger and whether the distance falls below or above the selected threshold. Experiments on four L2-ARCTIC speakers report per-speaker accuracies between 0.461 and 0.646 and F1 scores up to 0.684, with a qualitative case study on the word "Fraud".

Significance. If the central premise were validated, the approach would be attractive because it offers a personalized reference signal without language-specific phonetic rules or large annotated training sets, and the use of a public corpus (L2-ARCTIC) is a strength. The pipeline is clearly described, and the qualitative example suggests the intended mechanism can operate in at least one instance. However, the significance is currently undermined by two load-bearing gaps: the claim that the cloned audio contains corrected pronunciation is supported only by a single manual inspection, and the reported evaluation does not establish any held-out train/test separation. Moreover, the quantitative results in Table 1 are weak: for NJS the accuracy (0.486) is well below the majority-class baseline (163/253 ≈ 0.644), and for ERMS and MBMPS the accuracy is also at or below the respective majority baselines. These issues prevent the paper from demonstrating the claimed effectiveness.

major comments (4)

[Sec. 3.5, 3.6, Algorithm 1, Sec. 4] The evaluation appears to be in-sample. The thresholds tau_C and tau_I in Eq. (4) are computed as the 90th percentiles of D_correct and D_incorrect, and the KDEs K_COR and K_INC in Sec. 3.6 are fit to the same labeled partitions; Algorithm 1 then applies these to each word without describing any train/test split. Section 4 reports only aggregate precision, recall, F1, and accuracy, with no statement that evaluation words or utterances were held out from threshold/KDE fitting. If the reported numbers include training data, they are optimistically biased and cannot support the claim of effective detection. The authors must specify a split (e.g., utterance-level or word-level) and re-report Table 1 on held-out data; ideally, confidence intervals should also be provided given the small per-speaker supports.
[Sec. 1 and Sec. 3.1] The load-bearing assumption that the ElevenLabs clone provides 'ideal pronunciation conditions' or 'corrected pronunciation' is never verified. Section 3.1 simply loads the clones, and the only evidence is the manual inspection of one word ('Fraud') in Section 4. The paper does not report any systematic phonetic transcription, forced alignment, or human listening evaluation of the cloned outputs across the four speakers, 245+ words per speaker, or different phoneme contexts. If the clone preserves the learner's original mispronunciation or introduces synthesis artifacts, then the distances D_correct and D_incorrect no longer reflect pronunciation quality, and the threshold/KDE classifier loses its reference signal. The authors should verify clone pronunciation correctness quantitatively (e.g., by comparing forced-alignment phone labels of the clone against canonical ARPABET transcriptions) and report the agreement rate per speaker.
[Table 1] The reported accuracies are at or below the majority-class baseline for three of four speakers, which directly contradicts the abstract's claim of 'effective' mispronunciation pinpointing. For NJS, accuracy is 0.486 versus a majority baseline of 163/253 ≈ 0.644; for MBMPS, 0.461 versus 154/245 ≈ 0.629; for ERMS, 0.537 versus 133/244 ≈ 0.545. Even if a held-out evaluation were added, the current numbers provide little evidence that the model distinguishes mispronounced words from correct ones beyond always predicting the majority class. The authors should report the majority-class baseline and discuss how their method improves upon it, or temper the central claim accordingly.
[Sec. 4] The explainability case study is anecdotal. The manual observation that the original 'Fraud' contains /AW1/ and the clone contains /AO1/ is presented as evidence that the clone corrects pronunciation, but this is a single word from a single speaker. The paper does not state how the phonetic labels in Figure 3 were obtained (manual transcription? forced alignment?), nor whether the same correction behavior occurs across the other 26,000+ utterances in L2-ARCTIC or even across the other three speakers in the evaluation. A quantitative comparison of original vs. clone phone error rates, computed with a standard forced aligner, is needed to support the premise.

minor comments (4)

[Sec. 3.6] The sentence 'Given the sets of training distances partitioned by prediction correctness D_correct and D_incorrect, we their probability density functions' is missing a verb; it should read 'we estimate their probability density functions.'
[Fig. 4] The caption of Figure 4 contains a typo: 'Bribary' should be 'Bribery'.
[References] References [2] and [4] appear to be the same paper (Korzekwa et al., 'Computer-assisted pronunciation training—speech synthesis is almost all you need') with different formatting; one duplicate should be removed.
[Sec. 3.4] The method is described as 'Multi-Feature Dynamic Time Warping,' but the implementation computes independent 1D DTW distances per MFCC coefficient after resampling; the terminology is somewhat misleading because it is not a true multi-dimensional DTW.

Circularity Check

1 steps flagged · score 6.0 of 10

In-sample evaluation: the thresholds and KDE classifiers are fit to the same label-partitioned distances that are then scored, so the reported detection metrics largely restate the fit; the clone-corrects-pronunciation premise is also unverified.

fitted input called prediction [Sections 3.5-3.7 (Eqs. 4-7, Algorithm 1) and Section 4 (Table 1)]
"Given the sets of training distances partitioned by prediction correctness: Dcorrect = { ¯dj | correct predictions}, Dincorrect = { ¯dj | incorrect predictions}, the class-specific thresholds τC and τI are computed as the 90th percentiles of their respective distance distributions (Sec. 3.5); 'Fit KDE: KCOR ← kde(Dcorrect), KINC ← kde(Dincorrect)' and 'Compute thresholds: τC ← Q0.9(Dcorrect), τI ← Q0.9(Dincorrect)' (Alg. 1); 'We tested our model on four L2-Arctic speakers (EBVS, ERMS, MBMPS, NJS), measuring classification performance via precision, recall, F1-score, and accuracy' (Sec. 4)."

The detector's parameters are computed from the very same label-partitioned distance sets that are then scored. Algorithm 1's training loop appends each word's distance to Dcorrect or Dincorrect 'per ground truth', fits the KDEs, and sets τC = Q0.9(Dcorrect), τI = Q0.9(Dincorrect). Section 4 reports classification accuracy on four L2-ARCTIC speakers without describing any split into held-out words or speakers. As written, the reported 'predictions' are a deterministic recoding of the training labels: by construction, τC is the 90th percentile of the correct training distances, so 90% of correct training examples satisfy the correct-class threshold, and the KDE densities are evaluated at the same points that generated them.

full rationale

The paper's feature-extraction and distance-computation chain (MFCC + DTW, Eqs. 2-3 and Sec. 3.4) is not circular: it is a self-contained acoustic comparison. No load-bearing self-citation occurs; the cited related work is external and mostly contextual. The circularity is concentrated in the evaluation protocol. In Sec. 3.5, the thresholds are defined as empirical percentiles of distance sets Dcorrect and Dincorrect, which are themselves formed by partitioning the training data according to ground-truth pronunciation labels. In Sec. 3.6 and Algorithm 1, the KDE classifiers are fit to the same Dcorrect and Dincorrect, and the thresholds are Q0.9 of those same sets. Section 4 then reports classification accuracy on four L2-ARCTIC speakers without stating that the evaluated words/speakers were held out from this fitting procedure. As written, the reported 'predictions' are a deterministic recoding of the training labels: τC = Q0.9(Dcorrect) guarantees that 90% of the correct training distances satisfy the correct-class threshold, and the KDE densities are evaluated at the same points that generated them. Table 1 therefore does not independently demonstrate generalization. A separate, non-circular but load-bearing threat is the unverified premise that the ElevenLabs clone contains corrected pronunciation while preserving the learner's voice; this is an empirical assumption about the reference signal, not a circular derivation, and would be better scored as a correctness/validity risk.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The paper introduces no invented entities. Its free parameters are the two class-specific thresholds, the quantile level, and the unspecified KDE bandwidth. The central claim rests on several domain assumptions, most importantly that the clone corrects pronunciation, that acoustic distance tracks pronunciation errors, and that the ground-truth annotations are reliable. These assumptions are plausible but unvalidated in the text.

free parameters (4)

tau_C (90th percentile of correct-class distances) = not reported
Fit to ground-truth-labeled training distances in Sec 3.5; used as cutoff for CORRECT predictions.
tau_I (90th percentile of incorrect-class distances) = not reported
Fit to ground-truth-labeled training distances in Sec 3.5; used as cutoff for INCORRECT predictions.
quantile level alpha (0.1) = 0.1
Chosen by hand to define the 90th percentiles in Sec 3.5; no sensitivity analysis is provided.
KDE bandwidth = not specified
Kernel density estimates K_COR and K_INC in Sec 3.6 require a bandwidth; the paper does not state the bandwidth or kernel choices.

assumptions (6)

domain assumption The ElevenLabs clone provides corrected pronunciation while preserving speaker identity.
Stated in Sec 1 and assumed in Sec 3.1; no phonetic verification of the synthesized audio is provided.
domain assumption Acoustic deviation between original and cloned MFCC envelopes is a valid indicator of mispronunciation.
This is the central hypothesis from the Abstract, used to define the detection pipeline; it is not independently benchmarked.
domain assumption L2-ARCTIC word-level correctness annotations are accurate ground truth.
These labels partition D_correct and D_incorrect in Sec 3.5; no audit of annotation quality is reported.
domain assumption MFCC plus DTW distance captures pronunciation-relevant differences rather than speaker or prosody differences.
The feature and distance choice in Sec 3.3-3.4 assumes this; no ablation or comparison to other distances is given.
domain assumption Word alignments between original and cloned audio are accurate.
Word boundaries in Sec 3.2 are taken from TextGrids and clone alignment; no alignment error analysis is reported.
standard math Standard definitions of DTW, KDE, and percentiles are valid.
Background methods used in Sec 3.4-3.6; no issues, but included for completeness.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Pronunciation Deviation Analysis Through Voice Cloning and Acoustic Comparison." pith.science (2026). https://pith.science/paper/PNU4CLIG

@misc{pith2026250710985,
  author       = {Pith},
  title        = {Pith review of: Pronunciation Deviation Analysis Through Voice Cloning and Acoustic Comparison},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PNU4CLIG}},
  note         = {Machine review of arXiv:2507.10985}
}

read the original abstract

This paper presents a novel approach for detecting mispronunciations by analyzing deviations between a user's original speech and their voice-cloned counterpart with corrected pronunciation. We hypothesize that regions with maximal acoustic deviation between the original and cloned utterances indicate potential mispronunciations. Our method leverages recent advances in voice cloning to generate a synthetic version of the user's voice with proper pronunciation, then performs frame-by-frame comparisons to identify problematic segments. Experimental results demonstrate the effectiveness of this approach in pinpointing specific pronunciation errors without requiring predefined phonetic rules or extensive training data for each target language.

Figures

Figures reproduced from arXiv: 2507.10985 by the authors.

**Figure 1.** Data-processing pipeline. To address these limitations, our approach leverages personalized, synthetically generated voices, finely tuned to replicate each learner’s unique vocal traits under ideal pronunciation conditions. We employ advanced voice cloning technologies, explicitly utilizing the ElevenLabs platform [5], renowned for its sophisticated deep-learning algorithms and realistic synthetic speech generat… view at source ↗

**Figure 2.** An example of DTW alignment between original and synthesized samples. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Phonetic transcription of “Robbery, Bribery, Fraud.” The utterance under examination, “Robbery, Bribery, Fraud,” is presented in Figure 3 with its corresponding phonetic transcription. We extract the relevant temporal segments from both the original recording and the synthesized audio; these segments are depicted in Figure 4. Subsequent alignment is performed via Dynamic Time Warping (DTW), as demonstrated in… view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Temporal segments from the original and synthesized signals. Mispro [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

**Figure 5.** Figure 5: STFT spectrograms of a real vs. cloned word slice (colour scale: [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

12 extracted references · 12 canonical work pages

[1]

Computer-assisted pronunciation training: A systematic review,

M. Amrate and P.-H. Tsai, “Computer-assisted pronunciation training: A systematic review,” ReCALL, 2024, published online Sep. 2024

work page 2024
[2]

Computer-assisted pronun- ciation training—speech synthesis is almost all you need,

D. Korzekwa, J. L. Trueba, T. Drugman, and B. Kostek, “Computer-assisted pronun- ciation training—speech synthesis is almost all you need,”Speech Communication, vol. 142, pp. 22–33, 2022

work page 2022
[3]

Orthography-based pronunciation scoring for better capt feedback,

C. Richter, R. Pálsson, L. O’Brien, K. Friðriksdóttir, B. Bédi, E. H. Magnúsdót- tir, and J. Guðnason, “Orthography-based pronunciation scoring for better capt feedback,” in Proc. Interspeech 2023, 2023, pp. 1004–1008

work page 2023
[4]

Computer- assisted pronunciation training—speech synthesis is almost all you need,

D. Korzekwa, J. Lorenzo-Trueba, T. Drugman, and B. Kostek, “Computer- assisted pronunciation training—speech synthesis is almost all you need,” Speech Communication , vol. 142, p. 22–33, Jul. 2022. [Online]. Available: http://dx.doi.org/10.1016/j.specom.2022.06.003

work page doi:10.1016/j.specom.2022.06.003 2022
[5]

Elevenlabs text-to-speech api,

ElevenLabs, “Elevenlabs text-to-speech api,” https://elevenlabs.io/, 2025, accessed 11 June 2025

work page 2025
[6]

Improving mispronunciation detection using speech reconstruction,

A. Das and R. Gutierrez-Osuna, “Improving mispronunciation detection using speech reconstruction,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 32, pp. 4420–4433, 2024

work page 2024
[7]

Improving pronunciation and accent conversion through knowledge distillation and synthetic ground-truth from native tts,

T. N. Nguyen, S. Akti, N. Q. Pham, and A. Waibel, “Improving pronunciation and accent conversion through knowledge distillation and synthetic ground-truth from native tts,” inICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2025, pp. 1–5

work page 2025
[8]

L2-arctic: A non-native english speech corpus,

G. Zhao, S. Sonsaat, A. Silpachai, I. Lucić, E. Chukharev-Hudilainen, J. Levis, and R. Gutierrez-Osuna, “L2-arctic: A non-native english speech corpus,” inProc. Interspeech 2018, 2018, pp. 2783–2787

work page 2018

Show all 12 references

[9]

Capturing accents: An approach to personalize pronunciation training for learners with different l1 backgrounds,

V. Khaustova, E. Pyshkin, V. Khaustov, J. Blake, and N. Bogach, “Capturing accents: An approach to personalize pronunciation training for learners with different l1 backgrounds,” inSpeech and Computer (SPECOM 2023) , ser. Lecture Notes in Computer Science, vol. 14339, 2023, pp. 59–70

2023
[10]

Accent conversion without parallel data using pretrained tts models,

Y. Zhou, Z. Wu, M. Zhang, X. Tian, and H. Li, “Accent conversion without parallel data using pretrained tts models,”arXiv preprint arXiv:2212.10204 , 2022

2022 arXiv
[11]

Bäckström, O

T. Bäckström, O. Räsänen, A. Zewoudie, P. P. Zarazaga, L. Koivusalo, S. Das, E. G. Mellado, M. B. Mansali, D. Ramos, S. Kadiri, P. Alku, and M. H. Vali, Introduction to Speech Processing , 2nd ed., 2022. [Online]. Available: https://speechprocessingbook.aalto.fi

2022
[12]

B. W. Silverman,Density estimation for statistics and data analysis . Routledge, 2018

2018

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

Computer-assisted pronunciation training: A systematic review,

M. Amrate and P.-H. Tsai, “Computer-assisted pronunciation training: A systematic review,” ReCALL, 2024, published online Sep. 2024

work page 2024

[2] [2]

Computer-assisted pronun- ciation training—speech synthesis is almost all you need,

D. Korzekwa, J. L. Trueba, T. Drugman, and B. Kostek, “Computer-assisted pronun- ciation training—speech synthesis is almost all you need,”Speech Communication, vol. 142, pp. 22–33, 2022

work page 2022

[3] [3]

Orthography-based pronunciation scoring for better capt feedback,

C. Richter, R. Pálsson, L. O’Brien, K. Friðriksdóttir, B. Bédi, E. H. Magnúsdót- tir, and J. Guðnason, “Orthography-based pronunciation scoring for better capt feedback,” in Proc. Interspeech 2023, 2023, pp. 1004–1008

work page 2023

[4] [4]

Computer- assisted pronunciation training—speech synthesis is almost all you need,

D. Korzekwa, J. Lorenzo-Trueba, T. Drugman, and B. Kostek, “Computer- assisted pronunciation training—speech synthesis is almost all you need,” Speech Communication , vol. 142, p. 22–33, Jul. 2022. [Online]. Available: http://dx.doi.org/10.1016/j.specom.2022.06.003

work page doi:10.1016/j.specom.2022.06.003 2022

[5] [5]

Elevenlabs text-to-speech api,

ElevenLabs, “Elevenlabs text-to-speech api,” https://elevenlabs.io/, 2025, accessed 11 June 2025

work page 2025

[6] [6]

Improving mispronunciation detection using speech reconstruction,

A. Das and R. Gutierrez-Osuna, “Improving mispronunciation detection using speech reconstruction,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 32, pp. 4420–4433, 2024

work page 2024

[7] [7]

Improving pronunciation and accent conversion through knowledge distillation and synthetic ground-truth from native tts,

T. N. Nguyen, S. Akti, N. Q. Pham, and A. Waibel, “Improving pronunciation and accent conversion through knowledge distillation and synthetic ground-truth from native tts,” inICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2025, pp. 1–5

work page 2025

[8] [8]

L2-arctic: A non-native english speech corpus,

G. Zhao, S. Sonsaat, A. Silpachai, I. Lucić, E. Chukharev-Hudilainen, J. Levis, and R. Gutierrez-Osuna, “L2-arctic: A non-native english speech corpus,” inProc. Interspeech 2018, 2018, pp. 2783–2787

work page 2018