Evaluating Pre-trained Speech Encoders for Spontaneous Speech Detection and Out of Domain Synthetic Speech Generalisation in Indic Languages

Nihar Desai; Pavan Kumar J; Sujith Pulikodan; Varun Rai

REVIEW 3 major objections 5 minor 30 references

Evaluating Pre-trained Speech Encoders for Spontaneous Speech Detection and Out of Domain Synthetic Speech Generalisation in Indic Languages

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read In a 22-language Indic study, out-of-domain synthetic speech recall is predicted by training TTS proximity to the unseen TTS embeddings, not by distance from natural speech.

desk verdict Useful Indic benchmark undercut by a circular 'proximity predicts generalisation' claim. read the letter →

arxiv 2608.12536 v1 pith:NAZJL6DF submitted 2026-08-12 eess.AS

Varun Rai , Pavan Kumar J , Sujith Pulikodan , Nihar Desai This is my paper · ORCID

classification eess.AS

keywords spontaneousspeechdetectionsyntheticIndiclanguagestext-to-speechgeneralisationembeddingcentroidanalysislanguageisolationprobingfrozenencodersWhisperembeddings

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 evaluates five frozen transformer speech encoders on two tasks across 22 Indic languages: distinguishing spontaneous from scripted speech, and distinguishing natural from synthetic speech. Its central result is that a detector's ability to generalize to unseen text-to-speech (TTS) systems is governed by how close the training TTS systems' embedding centroids lie to the unseen synthetic speech, not by how close they are to natural speech. Expanding the training pool from one to four Indic TTS systems raises out-of-domain synthetic recall from 7% to 51%. The paper also shows that an encoder's language-discriminability can trade off against its spontaneity-detection accuracy, with Wav2vec2 and AST showing a negative correlation while Whisper and Vaani do not.

What carries the argument

The analysis runs on frozen encoder embeddings rather than fine-tuned models. Whisper-small supplies the 768-dimensional embeddings for the synthetic-speech experiments, while the spontaneity task also compares AST, Vaani-FastConformer, Wav2vec2-large, and BEATs, each feeding a compact three-block DNN classifier. The explanatory tools are centroid statistics in the Whisper embedding space, where each TTS system is represented by its mean vector and pairwise cosine and Euclidean distances to other systems, and a language isolation probe, a multinomial logistic regression trained to predict an utterance's language from its pooled embedding. The centroid distances carry the generalisation argument by linking training-pool composition to out-of-domain recall, while the language isolation score carries the trade-off finding when correlated with per-language spontaneity accuracy.

What would settle it

Compute the claimed proximity predictor without access to the unseen test utterances: estimate the centroids of freevc24 and xttsv2 from a development sample or model artifacts, choose the training configuration by centroid proximity, and measure OOD recall; the rule fails if a configuration chosen this way does not beat a natural-distance baseline. A simpler check is whether Table 2's distances, recomputed with the OOD systems removed, still rank the training configurations in the order of Table 1's recall.

Watch

Extended reading notes

Core claim

The paper's load-bearing claim is that out-of-domain generalisation is predicted by a training system's proximity to unseen TTS embeddings, not its distance from natural speech. In the Whisper-small embedding space, the OmniVoice centroid sits closest to the two unseen evaluation systems freevc24 and xttsv2, with Euclidean distances of 1.91 and 1.27, and training configurations that include OmniVoice achieve the highest out-of-domain synthetic recall. F5 has the closest centroid to natural speech, at 1.12, yet produces poor out-of-domain recall. Across all tested configurations, out-of-domain synthetic recall rises from 7% with a single training TTS system to 51% with all four, and the paper attributes the remaining ceiling to a distributional gap that composition alone cannot close within the current system pool.

Load-bearing premise

The proximity measure that predicts out-of-domain recall is computed from the unseen TTS test utterances themselves in the same Whisper embedding space the classifier uses, so applying the rule to choose training data presupposes access to the target distribution before it has been observed.

Editorial extensions

If this is right

Training data selection for real-world deepfake detectors should prioritize proximity to the expected synthetic voice embeddings over closeness to natural speech or raw diversity.
Including at least one training TTS system whose embedding centroid lies near the unseen evaluation region is the main lever for out-of-domain generalization, and adding more systems raises OOD synthetic recall from 7% to 51%.
A 51% recall ceiling persists across all tested configurations, indicating that pool composition alone cannot fully close the distributional gap to unseen TTS systems within the current system pool.
Encoders differ in whether their language-discriminability competes with spontaneity detection: Whisper and Vaani maintain high accuracy across languages, while Wav2vec2 and AST show a significant negative correlation between language isolation and spontaneity accuracy.
Frozen Whisper embeddings plus a lightweight classifier provide a strong cross-lingual baseline on Indic synthetic speech, extending the earlier finding that foundation-model representations beat task-specific end-to-end detectors on out-of-distribution synthetic inputs.

Reading between the lines

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

Beyond the paper: the proximity rule could be made actionable by estimating future TTS centroids from model checkpoints or a small development sample before training, and testing whether such an estimate still ranks training configurations by OOD recall.
Beyond the paper: the 51% ceiling suggests a capacity or representation limit of the frozen encoder; a testable follow-up is whether fine-tuning Whisper or adding one nearby TTS system beyond four breaks the ceiling.
Beyond the paper: the language isolation trade-off implies that spontaneity detection can be made more language-agnostic by selecting encoders whose spontaneity decisions do not depend on language identity, which could be verified by probing fine-tuned variants of Whisper and Vaani.
Beyond the paper: the same centroid-proximity analysis could be applied to voice conversion systems, which occupy a different region of embedding space than TTS, to see whether the proximity rule transfers to a broader class of deepfake sources.

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,

The benchmark work here is genuinely useful, but the headline claim about proximity predicting out-of-domain generalisation does not hold up. The centroid distances in Table 2 are computed from the same unseen TTS test utterances whose recall they are used to explain in Table 1. That is in-sample explanation, not prediction: the 'predictor' has direct access to the target distribution. The stress-test note is correct. The paper should be reframed as a post-hoc geometric analysis or validated out-of-sample by holding out entire TTS systems from both training and centroid computation.

What is actually new and good: this is the first systematic evaluation of five frozen speech encoders for spontaneous versus read classification across 22 scheduled Indic languages, using the IndicVoices scenario tags. The finding that Whisper and Vaani are consistently strong while Wav2vec2 shows a language-discriminability/spontaneity trade-off is a useful empirical result worth probing further. The training-diversity experiment is also a clean demonstration: expanding the training pool from one to four TTS systems raises OOD recall from 7% to 51%. That result stands regardless of the geometric explanation.

The soft spots are real and load-bearing. As noted, the central 'prediction' is circular because the geometry is fit on the test set. It is also a correlation over N=2 OOD systems, so even without the circularity it would be thin. There are no error bars, seeds, or repeats; class sizes for the spontaneity task are not reported; and the '51% ceiling' is an artifact of the limited system pool, not a general bound. The paper also overstates the 'not distance from natural speech' conclusion, since that comparison is based on the same test-derived centroids.

Who gets value from this? Researchers working on Indic speech benchmarks and on deepfake detection will find the spontaneity results and the diversity sweep useful, but should treat the centroid-proximity claim as an untested hypothesis. The paper deserves a serious referee because the benchmark is relevant and the flaw is instructive, but it needs major revision before publication. I would accept it for review.

Best.

Referee Report

3 major / 5 minor

Summary. The paper evaluates five frozen transformer speech encoders (AST, Vaani-FastConformer, Wav2vec2, Whisper, BEATs) on two binary tasks: spontaneous-versus-read classification across 22 Indic languages plus IEMOCAP, and natural-versus-synthetic speech detection trained on four in-domain Indic TTS systems (F5, Omni, Indic VITS, M4) and tested on two held-out OOD systems (freevc24, xttsv2). The authors report that OOD synthetic recall rises from 7% to 51% as the training pool expands to four TTS systems, and that centroid proximity between a training system's Whisper embeddings and the unseen OOD embeddings predicts this generalization, while proximity to natural speech does not. They also present a language-isolation probing analysis suggesting an encoder-dependent trade-off between language discriminability and spontaneity accuracy for Wav2vec2 and AST, but not for Whisper and Vaani.

Significance. If the central claim were established, the finding that training data should be selected for closeness to expected synthetic voices rather than to natural speech would be practically valuable for deepfake detector deployment. The paper also supplies a useful descriptive benchmark of five encoders on a 22-language Indic corpus, addressing a real gap in the literature. Strengths include the large language coverage, the use of scenario tags from IndicVoices as clean spontaneity labels, a speaker-independent IEMOCAP split, and the embedding-geometry framing. However, the headline 'prediction' claim is not supported as stated because the predictor is computed from the OOD test data itself, and the relationship rests on only two OOD systems with no uncertainty estimates.

major comments (3)

[§3.5, Table 2] The headline 'prediction' in Key Finding 2 is circular as presented. Table 2 reports cosine and Euclidean distances between system centroids in the Whisper-small embedding space, and the OOD rows (freevc24, xttsv2) are computed from the same OOD test set whose recall in Table 1 is being explained. The manuscript never states that the OOD centroids are derived from data that would be available at training-selection time; in fact §3.2 says these systems were used solely for OOD evaluation and embedding geometry analysis, meaning the geometry analysis and evaluation share the same utterances. A predictor that requires the unseen target distribution is not an out-of-sample predictor. The authors should either recompute the proximity measure from embeddings available before deployment (e.g., from a held-out development set or from natural speech plus known TTS systems), or explicitly reframe the result as a post-hoc interpretation of classifier behavior and remove the 'direct implications for training data selection' claim.
[§3.5] The proximity-recall relationship is evaluated on only two OOD systems (freevc24 and xttsv2). With N=2 target conditions, no regression, correlation, or statistical comparison is possible, and the claim that OOD generalization 'is predicted by' proximity is therefore unsupported even setting aside the circularity. The paper should be honest that this is an illustrative observation, or substantially increase the number of OOD evaluation systems.
[Table 1, §3.4] No uncertainty estimates accompany any of the quantitative results. Table 1 appears to report a single run per training configuration; there are no random seeds, repeats, confidence intervals, or error bars. The test-set size is not stated precisely for the OOD partition (200 utterances per language per model for up to nine languages implies a range, but the exact language count and total n are not given). Without this information, the 7% to 51% improvement cannot be assessed for significance, and the 'performance ceiling of 51%' in §3.5 is only the maximum over the tested configurations, not an established upper bound.

minor comments (5)

[Table 1, §3.2] The column labeled 'IndicSynth' is not defined in the caption. Please state explicitly that it reports pooled recall over freevc24 and xttsv2 and over which set of languages is included.
[Figure 1, §2.4] Reporting only a heatmap without numerical values makes it difficult to compare per-language accuracies across encoders; a table or a link to a CSV with the exact numbers would improve reproducibility.
[§2.4.1] The symbol R is used for both Pearson and Spearman correlations in the text and in Figure 2; please disambiguate the coefficient and the test in the figure legend and in each paragraph.
[§2.3, §3.3] Section 3.3 states the classifier architecture for Whisper-small only, while Section 2.3 covers five encoders; clarify which encoder is used for the natural-versus-synthetic experiments and why only Whisper is retained there.
[§4] The final sentence of the conclusion appears unfinished: '...explore joint spontaneity-authenticity modelling for robust real-world Indic' is followed immediately by Table 2; please complete the sentence or remove the dangling phrase.

Circularity Check

1 steps flagged · score 6.0 of 10

The key 'proximity predicts generalisation' claim is computed from the same unseen TTS test embeddings whose recall it is used to explain.

fitted input called prediction [Section 3.2 and Section 3.5 Key Finding 2, Table 2 vs Table 1]
"These systems were excluded from classifier training and used solely for out-of-distribution (OOD) evaluation and embedding geometry analysis. ... This analysis shows that out-of-domain generalisation is predicted by a training system's proximity to unseen TTS embeddings, not its distance from natural speech, a finding with direct implications for training data selection in real-world deepfake detectors."

The centroid distances that are said to predict generalisation (Omni to xttsv2 = 1.2688, Omni to freevc24 = 1.9178 in Table 2) are computed from the freevc24 and xttsv2 utterances themselves, in the same frozen Whisper embedding space used by the classifier. Those same utterances constitute the OOD evaluation partition whose recall (7% to 51% in Table 1) the proximity is invoked to explain. There is no temporal or statistical separation: the 'predictor' already contains the target distribution it is supposed to forecast. Consequently the claim reduces to a post-hoc description of classifier behaviour in that embedding space, not an independent rule for selecting training data before the unseen TTS system appears.

full rationale

The spontaneous-versus-scripted classification experiments, the language isolation probing, and the observed 7%-to-51% improvement from expanding the training pool are empirically self-contained: they use held-out validation partitions and report measured outcomes rather than fitted predictions. The circularity is concentrated in Key Finding 2 and the abstract/conclusion statements derived from it. There the 'predictor' is constructed from the same out-of-distribution test embeddings whose recall it claims to explain, so the headline generalisation rule is partially definitional rather than an out-of-sample forecast. The trend also rests on only two OOD systems, which is a statistical limitation but not itself circularity. The cited Vaani encoder is prior work by overlapping authors, but it is used as a benchmark system evaluated here, not as an unverified premise, so it does not add circularity. Overall, the central training-diversity result is real, but the paper's strongest interpretive claim reduces by construction to the geometry of the very test set it claims to predict, warranting a score of 6 rather than a higher score.

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

The central claims rest on the IndicVoices scenario labels and IEMOCAP filename labels being trustworthy, and on Whisper embedding space being a meaningful geometry for TTS similarity. No novel entities are introduced. The classifier's trainable weights are standard supervised fitting and are not counted as free parameters.

assumptions (4)

domain assumption IndicVoices scenario tags (Read, Extempore, Conversation) are a clean binary ground truth for spontaneous versus read speech.
Section 2.2 uses the scenario column as labels, assuming Extempore and Conversation correspond to spontaneous speech and Read to scripted speech.
domain assumption The IEMOCAP filename convention (_script and _impro) is a valid ground truth for scripted versus improvised speech.
Section 2.2 states that the filename convention is the sole source of ground truth labels for all IEMOCAP utterances.
domain assumption Whisper-small embedding space is an appropriate geometry for measuring TTS-system similarity via centroid Euclidean distance.
Section 3.4 and Table 2 compute pairwise cosine and Euclidean distances in the 768-dimensional Whisper embedding space and use them to explain classifier recall.
domain assumption A multinomial logistic regression probe trained on 70% of a 20% validation partition yields a valid language isolation score.
Section 2.4.1 describes the probe and uses per-class recall as a measure of language-discriminability; this assumes the probe settings are appropriate for the embedding spaces.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Evaluating Pre-trained Speech Encoders for Spontaneous Speech Detection and Out of Domain Synthetic Speech Generalisation in Indic Languages." pith.science (2026). https://pith.science/paper/NAZJL6DF

@misc{pith2026260812536,
  author       = {Pith},
  title        = {Pith review of: Evaluating Pre-trained Speech Encoders for Spontaneous Speech Detection and Out of Domain Synthetic Speech Generalisation in Indic Languages},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NAZJL6DF}},
  note         = {Machine review of arXiv:2608.12536}
}

read the original abstract

Transformer-based models have shown strong accuracy in distinguishing spontaneous from scripted speech and natural from synthetic speech, but these results are established on a narrow set of well-resourced language benchmarks and have not been extended across Indic languages, nor has embedding geometry been used to explain encoder behaviour or deepfake generalisation failure. We address these gaps by evaluating five frozen transformer encoders, AST, Vaani-FastConformer, Wav2vec2, Whisper and BEATs, across 22 Indic languages, and by conducting a multi-system TTS generalisation experiment across four TTS models. Beyond accuracy, we present language isolation probing and centroid proximity analysis. Probing reveals an encoder-dependent trade-off between language-discriminability and spontaneity detection. Centroid analysis shows that out-of-domain generalisation is predicted by a training system's proximity to unseen TTS embeddings, not its distance from natural speech, a finding with direct implications for training data selection in real-world deepfake detectors.

Figures

Figures reproduced from arXiv: 2608.12536 by the authors.

**Figure 3.** tsne plot to unseen TTS embeddings, not its distance from natural speech, a finding with direct implications for training data selection in real-world deepfake detectors. 4 Conclusion We present a unified study of spontaneous speech detection and synthetic speech generalisation across 22 Indic languages. Encoder choice matters: Wav2vec2 (Baevski et al., 2020) shows a language-discriminability/spontaneity trade-off … view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 27 canonical work pages

[1]

Can You Tell Apart Spontaneous and Read Speech if You Just Look at Prosody? , booktitle =

Batliner, Anton and Kompe, Ralf and Kie. Can You Tell Apart Spontaneous and Read Speech if You Just Look at Prosody? , booktitle =. 1995 , doi =

work page 1995
[2]

and Lee, Sungbok and Narayanan, Shrikanth S

Busso, Carlos and Bulut, Murtaza and Lee, Chi-Chun and Kazemzadeh, Abe and Mower, Emily and Kim, Samuel and Chang, Jeannette N. and Lee, Sungbok and Narayanan, Shrikanth S. , title =. Language Resources and Evaluation , volume =. 2008 , doi =

work page 2008
[3]

Modeling Prosody for Language Identification on Read and Spontaneous Speech , booktitle =

Rouas, Jean-Luc and Farinas, Jerome and Pellegrino, Fran. Modeling Prosody for Language Identification on Read and Spontaneous Speech , booktitle =. 2003 , doi =

work page 2003
[4]

Computer Speech & Language , volume =

Nakamura, Masanobu and Iwano, Koji and Furui, Sadaoki , title =. Computer Speech & Language , volume =. 2008 , doi =

work page 2008
[5]

Proceedings of Speech Prosody 2020 , pages =

Christodoulides, George , title =. Proceedings of Speech Prosody 2020 , pages =. 2020 , doi =

work page 2020
[6]

Proceedings of Interspeech 2016 , pages =

Ryant, Neville and Liberman, Mark , title =. Proceedings of Interspeech 2016 , pages =. 2016 , doi =

work page 2016
[7]

arXiv preprint , year =

Elisha, Shahar and others , title =. arXiv preprint , year =

work page
[8]

, title =

Javed, Tahir and Nawale, Janki and George, Eldho and Joshi, Sakshi and Bhogale, Kaushal and Kumar, Pratyush and Khapra, Mitesh M. , title =. Findings of the Association for Computational Linguistics: ACL 2024 , pages =. 2024 , doi =

work page 2024

Show all 30 references

[9]

Proceedings of ICASSP 2022 , pages =

Jung, Jee-weon and Heo, Hee-Soo and Tak, Hemlata and Shim, Hye-jin and Chung, Joon Son and Lee, Bong-Jin and Yu, Ha-Jin and Evans, Nicholas , title =. Proceedings of ICASSP 2022 , pages =. 2022 , doi =

2022
[10]

Proceedings of ICASSP 2021 , pages =

Tak, Hemlata and Patino, Jose and Todisco, Massimiliano and Nautsch, Andreas and Evans, Nicholas and Larcher, Anthony , title =. Proceedings of ICASSP 2021 , pages =. 2021 , doi =

2021
[11]

arXiv preprint , year =

Li, Yuankun and others , title =. arXiv preprint , year =
[12]

Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages =

Sharma, Divya V and Ekbote, Vijval and Gupta, Anubha , title =. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages =. 2025 , url =

2025
[13]

Proceedings of Interspeech 2018 , publisher =

Mangalam, Karttikeya and Guha, Tanaya , title =. Proceedings of Interspeech 2018 , publisher =. 2018 , doi =

2018
[14]

2022 , url =

Robust Speech Recognition via Large-Scale Weak Supervision , author =. 2022 , url =

2022
[15]

Advances in Neural Information Processing Systems (NeurIPS) , volume =

wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations , author =. Advances in Neural Information Processing Systems (NeurIPS) , volume =. 2020 , url =

2020
[16]

Proceedings of Interspeech 2021 , pages =

AST: Audio Spectrogram Transformer , author =. Proceedings of Interspeech 2021 , pages =. 2021 , doi =

2021
[17]

Proceedings of the 40th International Conference on Machine Learning (ICML) , pages =

BEATs: Audio Pre-Training with Acoustic Tokenizers , author =. Proceedings of the 40th International Conference on Machine Learning (ICML) , pages =. 2023 , publisher =

2023
[18]

arXiv preprint arXiv:2603.28714 , year =

VAANI: Capturing the Language Landscape for an Inclusive Digital India , author =. arXiv preprint arXiv:2603.28714 , year =

arXiv
[19]

Proceedings of Interspeech 2017 , pages =

Todisco, Massimiliano and Delgado, Héctor and Evans, Nicholas , title =. Proceedings of Interspeech 2017 , pages =. 2017 , doi =

2017
[20]

Speech Communication , volume =

Wu, Zhizheng and Kinnunen, Tomi and Evans, Nicholas and Yamagishi, Junichi and Hanilçi, Cemal and Sahidullah, Md and Alegre, Federico , title =. Speech Communication , volume =. 2015 , doi =

2015
[21]

Proceedings of NeurIPS 2022 , year =

Chen, Sanyuan and Wang, Chengyi and Chen, Zhuo and Wu, Yu and Jia, Shujie and Hua, Wei and Wang, Dong and Zhou, Ming and Li, Jinyu , title =. Proceedings of NeurIPS 2022 , year =

2022
[22]

2024 , howpublished =

2024
[23]

, title =

S V, Praveen and Anand, Srija and Siddhartha, Soma and Khapra, Mitesh M. , title =. 2025 , howpublished =

2025
[24]

Proceedings of ICASSP 2023 , publisher =

Li, Jingyi and Tu, Weiping and Xiao, Li , title =. Proceedings of ICASSP 2023 , publisher =. 2023 , url =

2023
[25]

Proceedings of Interspeech 2024 , pages =

Casanova, Edresson and Davis, Kelly and G. Proceedings of Interspeech 2024 , pages =. 2024 , doi =

2024
[26]

arXiv preprint arXiv:2604.00688 , year =

Zhu, Han and Ye, Lingxuan and Kang, Wei and Yao, Zengwei and Guo, Liyong and Kuang, Fangjun and Han, Zhifeng and Zhuang, Weiji and Lin, Long and Povey, Daniel , title =. arXiv preprint arXiv:2604.00688 , year =

arXiv
[27]

Proceedings of Interspeech 2019 , pages =

Wang, Xin and Kinnunen, Tomi and Sahidullah, Md and Delgado, Héctor and Evans, Nicholas and Yamagishi, Junichi and Todisco, Massimiliano , title =. Proceedings of Interspeech 2019 , pages =. 2019 , doi =

2019
[28]

Proceedings of Interspeech 2021 , pages =

Yamagishi, Junichi and Wang, Xin and Todisco, Massimiliano and Patino, Jose and Nautsch, Andreas and Evans, Nicholas and Kinnunen, Tomi and Sahidullah, Md and Delgado, Héctor , title =. Proceedings of Interspeech 2021 , pages =. 2021 , doi =

2021
[29]

Proceedings of Interspeech 2023 , pages =

Yamagishi, Junichi and Todisco, Massimiliano and Wang, Xin and Nautsch, Andreas and Evans, Nicholas and Patino, Jose and Kinnunen, Tomi and Sahidullah, Md and Delgado, Héctor , title =. Proceedings of Interspeech 2023 , pages =. 2023 , doi =

2023
[30]

Seamless: Multilingual Expressive and Streaming Speech Translation , journal =

Barrault, Lo. Seamless: Multilingual Expressive and Streaming Speech Translation , journal =. 2023 , url =

2023

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.

[1] [1]

Can You Tell Apart Spontaneous and Read Speech if You Just Look at Prosody? , booktitle =

Batliner, Anton and Kompe, Ralf and Kie. Can You Tell Apart Spontaneous and Read Speech if You Just Look at Prosody? , booktitle =. 1995 , doi =

work page 1995

[2] [2]

and Lee, Sungbok and Narayanan, Shrikanth S

Busso, Carlos and Bulut, Murtaza and Lee, Chi-Chun and Kazemzadeh, Abe and Mower, Emily and Kim, Samuel and Chang, Jeannette N. and Lee, Sungbok and Narayanan, Shrikanth S. , title =. Language Resources and Evaluation , volume =. 2008 , doi =

work page 2008

[3] [3]

Modeling Prosody for Language Identification on Read and Spontaneous Speech , booktitle =

Rouas, Jean-Luc and Farinas, Jerome and Pellegrino, Fran. Modeling Prosody for Language Identification on Read and Spontaneous Speech , booktitle =. 2003 , doi =

work page 2003

[4] [4]

Computer Speech & Language , volume =

Nakamura, Masanobu and Iwano, Koji and Furui, Sadaoki , title =. Computer Speech & Language , volume =. 2008 , doi =

work page 2008

[5] [5]

Proceedings of Speech Prosody 2020 , pages =

Christodoulides, George , title =. Proceedings of Speech Prosody 2020 , pages =. 2020 , doi =

work page 2020

[6] [6]

Proceedings of Interspeech 2016 , pages =

Ryant, Neville and Liberman, Mark , title =. Proceedings of Interspeech 2016 , pages =. 2016 , doi =

work page 2016

[7] [7]

arXiv preprint , year =

Elisha, Shahar and others , title =. arXiv preprint , year =

work page

[8] [8]

, title =

Javed, Tahir and Nawale, Janki and George, Eldho and Joshi, Sakshi and Bhogale, Kaushal and Kumar, Pratyush and Khapra, Mitesh M. , title =. Findings of the Association for Computational Linguistics: ACL 2024 , pages =. 2024 , doi =

work page 2024