Pith. sign in

REVIEW 3 major objections 5 minor 76 references

Speaker Retrieval in the Wild: Challenges, Effectiveness and Robustness

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 Pre-trained speaker models retrieve named speakers from a 1948-1979 BBC archive with 86-87% top-one accuracy.

desk verdict A useful empirical speaker-retrieval study on BBC Rewind, but the central P@1 numbers rest on a missing statement about whether query files stay in the search index; a referee should demand that clarification before trusting the results. read the letter →

arxiv 2504.18950 v2 pith:YSRAOZ73 submitted 2025-04-26 cs.SD cs.IRcs.LGeess.AS

Erfan Loweimi , Mengjie Qian , Kate Knill , Mark Gales This is my paper · ORCID
classification cs.SDcs.IRcs.LGeess.AS
keywords speakerretrievaldiarisationembeddingsmediaarchiveszero-shotmetadata-derivedlabelsnoiserobustnessreverberation
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 argues that speaker retrieval in large, aged, weakly annotated audio archives can be done without any training or fine-tuning on the archive itself, by chaining pre-trained speaker diarisation with pre-trained speaker embeddings. It takes on two 'in the wild' problems: journalist-written synopses name people who may be visible but silent, and the recordings span quiet studios to street noise, overlapping speech, and 31 years of changing technology. After filtering the query set down to the 451 files where the named person is actually audible, the ECAPA-TDNN and TitaNet-Large systems rank the correct file first for 86.3% and 86.9% of queries on the BBC Rewind Northern Ireland subset. Systematic distortion tests show the pipeline absorbs additive noise, 8-bit audio, and 8 kHz bandwidth, while 4 kHz bandwidth and strong reverberation degrade it sharply, with model-specific failure points.

What carries the argument

The engine of the system is a four-stage chain: diarisation splits an archive file into speaker-labelled segments; a pre-trained embedding model converts each segment into a fixed-length vector; duration-proportional weights aggregate those vectors into one speaker embedding per detected speaker, with linear weighting chosen over uniform, softmax, and rank weighting because it gave the best retrieval with no hyperparameter; and retrieval ranks a file by the maximum cosine similarity between the query embedding and any speaker embedding in that file. A second mechanism, the evaluation taxonomy that separates synopsis-derived names into audio-visual presence, audio-only presence, and silent presence, is what lets the authors measure the system on labels that actually correspond to an audible voice.

What would settle it

Have two independent annotators, blind to the retrieval results, label every one of the 523 query files as audio-visual presence, audio-only presence, or silent presence; if their labels disagree with the paper's assignments on more than a few percent of files, the reported 86-87% top-one accuracy on the refined query set is not reproducible.

Watch

Extended reading notes

Core claim

On the refined query set of 451 files (19.3 hours) whose synopsis names correspond to an audible voice, the best pre-trained systems retrieve the correct file at rank one for 86.3% (ECAPA-TDNN) and 86.9% (TitaNet-Large) of queries, with top-three precision around 84%. The paper concludes that speaker retrieval in the wild is achievable with zero in-domain training: pre-trained diarisation segments each file, a pre-trained embedding model represents each segment, duration-weighted averaging forms one embedding per detected speaker, and a file is scored by the maximum cosine similarity between the query and that file's speaker embeddings. It further shows that segment-level retrieval matches speaker-level accuracy, that interpolating the scores of the two best embeddings gives a small consistent gain, and that the silent-presence label category explains most of the residual failures on the unfiltered query set.

Load-bearing premise

The headline accuracy numbers assume the manual classification of which query files actually contain the named speaker's voice is correct, and that classification was done by the authors without a published annotation protocol or a second annotator.

Editorial extensions

If this is right

  • Deploying the pipeline on a new archive requires no manual annotation or fine-tuning; only archive audio and weak name metadata are needed to produce ranked speaker searches.
  • On a query set whose labels genuinely correspond to an audible voice, top-one precision of 86-87% is the realistic operating point, while on unfiltered metadata the number falls to about 75% because silent-presence queries act as noise.
  • Segment-level scoring is nearly as accurate as speaker-level scoring, so the same framework can return the timestamps where the target speaker appears with little accuracy cost.
  • The distortion tests map the operating envelope: babble and white noise are the hardest noise types, music the easiest; 8 kHz and 8-bit audio are tolerable, 4 kHz audio is catastrophic; ECAPA-TDNN holds up to roughly one second of reverberation time while TitaNet-Large collapses above 0.25 seconds.
  • Simply interpolating ECAPA-TDNN and TitaNet-Large scores improves P@1, P@3, P@5, and P@10 consistently, indicating the two models carry complementary information.

Reading between the lines

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

  • A practical system should use query-side audio verification before trusting the ranked list, because 15% of raw synopsis-derived queries in this archive are silent-presence and produce near-zero retrieval accuracy; the paper's own cumulative analysis implies this filter is what makes the headline numbers hold.
  • The near-perfect precision on the top 60% of queries suggests residual errors are dominated by label and prominence effects rather than acoustics, so a query set engineered to be uniformly difficult would be a sharper test of embedding quality.
  • The TitaNet-Large reverberation collapse, despite its reverberation-augmented training, invites a testable fix: prepend dereverberation or adapt the embeddings with reverberant augmentation and measure whether its 3.8% top-one accuracy at two seconds of reverberation recovers toward the ECAPA system's 33.3%.
  • The metadata recipe of named-entity extraction plus manual presence auditing should transfer to other broadcast archives, but each corpus would need its own audit because the mix of audio-visual, audio-only, and silent-presence entries depends on how journalists wrote the synopses.
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.

Referee Report

3 major / 5 minor

Summary. The paper addresses speaker retrieval on large, aged media archives, using the BBC Rewind corpus (12,594 Northern Ireland files, 409 hours, 1948–1979) as a testbed. The proposed system performs diarisation, extracts segment- and speaker-level embeddings with pre-trained models (x-vector, ECAPA-TDNN, TitaNet), and ranks archive files by cosine similarity. To handle unreliable synopsis metadata, the authors manually categorize 523 query files into Audio-Visual Presence (AVP), Silent Presence (SP), and Audio-only Presence (AoP), and use the AVP∪AoP subset (Q†, 451 files) for headline evaluation. In the clean setup, ECAPA-TDNN-SB and TitaNet-L-NeMo achieve P@1 of 86.3% and 86.9% on Q† (Table IV); the paper also reports robustness experiments against additive noise, bit-depth reduction, sampling-rate reduction, and reverberation. The central claims are that pre-trained models provide effective and robust speaker retrieval with no training or fine-tuning on the target archive, and that the framework generalizes beyond BBC Rewind.

Significance. If the evaluation protocol is sound, the paper makes a useful practical contribution: it demonstrates that off-the-shelf speaker embeddings and diarisation, combined with a metadata-driven query-selection procedure, can retrieve speakers from a difficult archival corpus without any target-domain training. The work also provides a systematic robustness comparison across several distortion types, which is valuable for practitioners. The paper is clear about the distinction between AVP/SP/AoP labels and proposes a sensible way to reduce label noise. The results are reproducible in principle because the pre-trained model checkpoints and the BBC Rewind corpus are public, though query-file annotations are not released. The main significance depends on whether the retrieval evaluation truly excludes the query files themselves and whether the manual label categorization is reliable; both points require clarification.

major comments (3)
  1. [Section V-A / Section II-D] The manuscript never states whether each query's own video file is excluded from the archive index during retrieval. The query set is described as 523 files 'randomly selected from the Rewind corpus' (Section V-A), and Eq. (6) in Section II-D scores 'each archive file' with no hold-out note. If a query's own file remains in the index, its speaker embedding is derived from the same recording as the query, making it a near-identical, trivially relevant candidate. The reported P@1 values (e.g., 86.3% for ECAPA-TDNN-SB on Q† in Table IV) would then measure, at least in part, re-identification of the query recording rather than retrieval of previously unseen files containing the target speaker. This is load-bearing for the central effectiveness and robustness claims in Tables III--V and Figs. 9--10. The authors must state explicitly whether query files were removed from the archive before scoring; if they were removed in the implementation, that statement is missing and must be added.
  2. [Section V-C (Table III)] The construction of Q† depends on a manual categorization of the 523 query videos into AVP, SP, and AoP categories, but no annotation protocol, annotator training, or inter-annotator agreement is reported. The paper says only that the authors 'thoroughly examined both the visual and auditory contents' (Section V-C). Because Q† is the test set for all headline results, the reliability of this categorization is critical: if the identification of silent-presence files is imperfect, the reported P@1 values overstate system quality. The authors should provide a detailed annotation guideline, report inter-annotator statistics on at least a subset, and state how borderline cases were resolved. The cumulative analysis in Fig. 6, which sorts files by system performance before manual labeling, also risks biasing the subsequent exclusion of SP files; this should be acknowledged or the annotation should be performed independently of the system scores.
  3. [Section V-B and Fig. 7] Several hyperparameters and method choices are selected on the same query set that is later used for evaluation, without a separate validation split. Table II compares uniform, linear, softmax (with τ tuned over 1, 5, 10, 15), and ranking weighting schemes on the 523-file query set and then adopts linear weighting for the remaining experiments; this is a model-selection step performed on the test data. Similarly, Fig. 7 sweeps the fusion weight λ in Eq. (10) on Q† and reports that λ=0.5 gives the best P@K. Such in-sample selection can inflate performance and complicates the interpretation of the reported numbers. The paper should either use a hold-out validation set for these choices or explicitly disclose that the reported results include development-on-test selection. This is not fatal to the core methodology, but it must be disclosed for a rigorous evaluation.
minor comments (5)
  1. [Section VI-B] The abbreviation AvgRPR is used in Tables VII--VIII and X but is never defined; it should be spelled out (e.g., average relative performance reduction) when first used.
  2. [Fig. 10 caption] The caption says 'TitNet-L-NeMo'; this should be 'TitaNet-L-NeMo'.
  3. [Section II-C / Table II] The temperature values evaluated for softmax weighting are τ = 1, 5, 10, 15; the paper does not explain why this grid is chosen, and a wider range may be needed to support the claim that linear weighting is superior to any softmax setting.
  4. [Section V-A] The query selection is described as 'randomly selected' but no random seed or selection procedure is given; this limits reproducibility of the exact query set. Releasing the query file IDs would be helpful.
  5. [Section III-B] The count of 5,800 distinct names is produced by spaCy NER, but no error analysis of the NER output is provided; a small manual validation of name extraction accuracy would strengthen the metadata statistics.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: retrieval is an empirical zero-shot benchmark using externally pre-trained embeddings, not a derivation equivalent to its inputs.

full rationale

The paper's central effectiveness claim (P@1 86.3/86.9 on Q†, Table IV) is an empirical evaluation, not a derived prediction. The speaker embeddings (x-vector, ECAPA-TDNN, TitaNet) and diarisation (PyAnnote) are pre-trained on external corpora (VoxCeleb 1/2, LibriSpeech, Switchboard, Fisher, SRE, RIR noise) and are applied to BBC Rewind without training or fine-tuning on the target archive; the query speakers are 38 Northern Ireland politicians from 1948-1979 recordings, so the result does not reduce to a fitted parameter or to a self-citation. The duration weighting equations (2)-(5) are fixed aggregation rules, and Table II is an empirical comparison, not a fit whose output is then relabelled as a prediction. The Q† refinement (AVP∪AoP) is a label-quality filter based on manual audiovisual inspection, and the full 523-query result (P@1 74.6) is also reported, so the improved subset result is not a hidden substitution. Self-citations [8], [9], [48], [49] are contextual (related work/project background) and not load-bearing. The main caveats are evaluation-protocol issues rather than circularity: the manuscript does not state whether a query's own file is excluded from the ranked index, and λ in Eq. (10) is selected on Q† before reporting the fusion curve. These could inflate or overfit the reported numbers, but there is no equation-level reduction showing that the effectiveness result is equivalent to its inputs; they belong in a correctness/robustness review, not a circularity finding.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

No new physical or architectural entities are introduced. The numerical free parameters are limited to test-set-selected weighting and fusion choices; the main assumptions concern domain transfer and metadata trustworthiness. This is a low-parameter empirical system paper, but the query-set selection and manual labels are not released, so reproducibility is limited.

free parameters (2)
  • Softmax temperature tau for segment weighting = tested at 1, 5, 10, 15; linear weighting later chosen instead
    Introduced in Eq. (4) and evaluated in Table II; the best tau is selected on the same Q† test set, then abandoned in favour of linear weighting.
  • Interpolation weight lambda for score fusion = 0.5
    Eq. (10); Fig. 7 sweeps lambda on the Q† test set and reports lambda = 0.5 as most consistent, so the fusion gain is tuned to the evaluation data.
assumptions (4)
  • domain assumption Synopsis names extracted via spaCy NER correspond to individuals relevant to the video, with errors rare for prominent politicians.
    Used throughout; formalized in Section IV-C as basis for using synopses as evaluation labels. Manual review of 523 queries reduces, but does not eliminate, reliance on this.
  • domain assumption Pre-trained models (PyAnnote, x-vector, ECAPA, TitaNet) trained on modern corpora transfer to aged noisy BBC audio without adaptation.
    Section IV-B argues domain mismatch is handled by diverse training data; the entire zero-shot framework depends on this.
  • domain assumption Cosine similarity of speaker embeddings is a valid relevance score for ranking archive files by speaker presence.
    Eqs. (6) and (7) define the retrieval score; no learned scoring or calibration is used.
  • domain assumption Speaker diarisation segments are accurate enough that averaged segment embeddings faithfully represent each speaker.
    The embedding extraction in Eq. (1) averages segments assigned by PyAnnote; diarisation errors propagate directly into the retrieval scores.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Speaker Retrieval in the Wild: Challenges, Effectiveness and Robustness." pith.science (2026). https://pith.science/paper/YSRAOZ73

@misc{pith2026250418950,
  author       = {Pith},
  title        = {Pith review of: Speaker Retrieval in the Wild: Challenges, Effectiveness and Robustness},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YSRAOZ73}},
  note         = {Machine review of arXiv:2504.18950}
}
read the original abstract

There is a growing abundance of publicly available or company-owned audio/video archives, highlighting the increasing importance of efficient access to desired content and information retrieval from these archives. This paper investigates the challenges, solutions, effectiveness, and robustness of speaker retrieval systems developed "in the wild" which involves addressing two primary challenges: extraction of task-relevant labels from limited metadata for system development and evaluation, as well as the unconstrained acoustic conditions encountered in the archive, ranging from quiet studios to adverse noisy environments. While we focus on the publicly-available BBC Rewind archive (spanning 1948 to 1979), our framework addresses the broader issue of speaker retrieval on extensive and possibly aged archives with no control over the content and acoustic conditions. Typically, these archives offer a brief and general file description, mostly inadequate for specific applications like speaker retrieval, and manual annotation of such large-scale archives is unfeasible. We explore various aspects of system development (e.g., speaker diarisation, embedding extraction, query selection) and analyse the challenges, possible solutions, and their functionality. To evaluate the performance, we conduct systematic experiments in both clean setup and against various distortions simulating real-world applications. Our findings demonstrate the effectiveness and robustness of the developed speaker retrieval systems, establishing the versatility and scalability of the proposed framework for a wide range of applications beyond the BBC Rewind corpus.

Figures

Figures reproduced from arXiv: 2504.18950 by the authors.

Figure 1
Figure 1. Workflow of the speaker retrieval system consisting of pre-processing, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Extraction of segment (e) and speaker (s) level embeddings using pre-trained diarisation and embedding extraction models from input audio. N and M, respectively, denote the number of segments and speakers, both automatically determined by the speaker diarisation block. B. Speaker Diarisation The next block is speaker diarisation which aims at answer￾ing “who spoke when?” through temporal decomposition of the speech … view at source ↗
Figure 4
Figure 4. Histograms of the #speakers per file (according to diarisation) along [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figures from the paper (7 more)
Figure 3
Figure 3. Figure 3: Histogram of the information extracted using diarisation. (a) number of [PITH_FULL_IMAGE:figures/full_fig_p006_3.png]
Figure 5
Figure 5. Figure 5: Occurrences of names vs. name index, computed by calculating [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Cumulative performance analysis. Precision@K (P@ [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Effect of ECAPA-TDNN and TitaNet-L speaker embeddings combi [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 9
Figure 9. Figure 9: Retrieval performance for Babble, White Gaussian (WG), Street and [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 8
Figure 8. Figure 8: Scatter plots in 2D using t-SNE dimensionality reduction method. (a) [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]
Figure 10
Figure 10. Figure 10: Speaker retrieval performance at different SNRs for different [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

76 extracted references · 65 canonical work pages

  1. [1]

    Multimedia information retrieval: Theory and techniques,

    L. Stone, “Multimedia information retrieval: Theory and techniques,” Library Review, vol. 63, no. 4/5, pp. 373–374, 2014

    work page 2014

  2. [2]

    R ¨uger, Multimedia information retrieval , ser

    S. R ¨uger, Multimedia information retrieval , ser. Synthesis Lectures on Information Concepts, Retrieval and Services. Morgan & Claypool Publishers, 2010

    work page 2010

  3. [3]

    Scalable identity-oriented speech retrieval,

    C. Chen, D. Jiang, J. Peng, R. Lian, Y . Li, C. Zhang, L. Chen, and L. Fan, “Scalable identity-oriented speech retrieval,” IEEE Transactions on Knowledge and Data Engineering , vol. 35, no. 3, pp. 3261–3265, 2023

    work page 2023

  4. [4]

    Spoken content retrieval—beyond cascading speech recognition with text retrieval,

    L.-s. Lee, J. Glass, H.-y. Lee, and C.-a. Chan, “Spoken content retrieval—beyond cascading speech recognition with text retrieval,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 23, no. 9, pp. 1389–1420, 2015

    2015

  5. [5]

    Spoken content retrieval: A survey of techniques and technologies,

    M. Larson and G. J. F. Jones, “Spoken content retrieval: A survey of techniques and technologies,” Found. Trends Inf. Retr., vol. 5, no. 4—5, p. 235–422, apr 2012

    work page 2012

  6. [6]

    Tur and R

    G. Tur and R. De Mori, Speech Retrieval. John Wiley & Sons, Ltd, 2011, pp. 417–446

    work page 2011

  7. [7]

    A framework for speaker retrieval and identification through unsupervised learning,

    V . de Abreu Campos and D. C. G. Pedronette, “A framework for speaker retrieval and identification through unsupervised learning,” Computer Speech & Language , vol. 58, pp. 153–174, 2019

    work page 2019

  8. [8]

    On the usefulness of speaker embeddings for speaker retrieval in the wild: A comparative study of x-vector and ecapa-tdnn models,

    E. Loweimi, M. Qian, K. Knill, and M. Gales, “On the usefulness of speaker embeddings for speaker retrieval in the wild: A comparative study of x-vector and ecapa-tdnn models,” in INTERSPEECH, 2024, pp. 3774–3778

    work page 2024

Show all 76 references
  1. [9]

    Zero- shot audio topic reranking using large language models,

    M. Qian, R. Ma, A. Liusie, E. Loweimi, K. Knill, and M. Gales, “Zero- shot audio topic reranking using large language models,” inIEEE Spoken Language Technology Workshop (SLT), 2024

    2024

  2. [10]

    Ai-based seman- tic multimedia indexing and retrieval for social media on smartphones,

    S. Wagenpfeil, F. Engel, P. M. Kevitt, and M. Hemmje, “Ai-based seman- tic multimedia indexing and retrieval for social media on smartphones,” Information, vol. 12, no. 1, p. 43, 2021

    2021

  3. [11]

    Adversarial attacks against audio surveillance systems,

    S. Ntalampiras, “Adversarial attacks against audio surveillance systems,” in EUSIPCO. IEEE, 2022, pp. 284–288

    2022

  4. [12]

    Audio surveillance: A systematic review,

    M. Crocco, M. Cristani, A. Trucco, and V . Murino, “Audio surveillance: A systematic review,” ACM Comput. Surv., vol. 48, no. 4, feb 2016

    2016

  5. [13]

    Audio forensics behind the iron curtain: from raw sounds to expert testimony,

    A. Kvicalova, “Audio forensics behind the iron curtain: from raw sounds to expert testimony,” Sound Studies, vol. 9, no. 2, pp. 187–208, 2023

    2023

  6. [14]

    Digital forensics approach for handling audio and video files,

    S. M. Pedapudi and N. Vadlamani, “Digital forensics approach for handling audio and video files,” Measurement: Sensors , vol. 29, p. 100860, 2023

    2023

  7. [15]

    Evaluation of a forensic auto- matic speaker recognition system with emotional speech recordings,

    R. Essery, P. Harrison, and V . Hughes, “Evaluation of a forensic auto- matic speaker recognition system with emotional speech recordings,” in INTERSPEECH, 2023, pp. 2568–2572

    2023

  8. [16]

    An overview of recent work in media forensics: Methods and threats,

    K. Bhagtani, A. K. S. Yadav, E. R. Bartusiak, Z. Xiang, R. Shao, S. Baireddy, and E. J. Delp, “An overview of recent work in media forensics: Methods and threats,” ArXiv, vol. abs/2204.12067, 2022

    2022 arXiv

  9. [17]

    Artificial intelligence and music: Open questions of copyright law and engineering praxis,

    B. L. Sturm, M. Iglesias, O. Ben-Tal, M. Miron, and E. G ´omez, “Artificial intelligence and music: Open questions of copyright law and engineering praxis,” Arts, vol. 8, no. 3, p. 115, 2019

    2019

  10. [18]

    Ethical dimensions of music information retrieval technology,

    A. Holzapfel, B. L. Sturm, and M. Coeckelbergh, “Ethical dimensions of music information retrieval technology,” Transactions of the Interna- tional Society for Music Information Retrieval , vol. 1, no. 1, pp. 44–55, 2018

    2018

  11. [19]

    A review of speaker diarization: Recent advances with deep learning,

    T. J. Park, N. Kanda, D. Dimitriadis, K. J. Han, S. Watanabe, and S. Narayanan, “A review of speaker diarization: Recent advances with deep learning,” Computer Speech & Language, vol. 72, p. 101317, 2022

    2022

  12. [20]

    Speaker recognition based on deep learning: An overview,

    Z. Bai and X.-L. Zhang, “Speaker recognition based on deep learning: An overview,” Neural Networks, vol. 140, pp. 65–99, 2021

    2021

  13. [21]

    Automatic speaker verification systems and spoof detection techniques: review and analysis,

    A. Mittal and M. Dua, “Automatic speaker verification systems and spoof detection techniques: review and analysis,” International Journal of Speech Technology, vol. 25, pp. 105–134, 2022

    2022

  14. [22]

    Representation learning: A review and new perspectives,

    Y . Bengio, A. Courville, and P. Vincent, “Representation learning: A review and new perspectives,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 35, no. 8, pp. 1798–1828, 2013

    2013

  15. [23]

    Front- end factor analysis for speaker verification,

    N. Dehak, P. J. Kenny, R. Dehak, P. Dumouchel, and P. Ouellet, “Front- end factor analysis for speaker verification,” IEEE Transactions on Audio, Speech, and Language Processing , vol. 19, no. 4, pp. 788–798, 2011

    2011

  16. [24]

    Deep neural networks for small footprint text-dependent speaker verification,

    E. Variani, X. Lei, E. McDermott, I. L. Moreno, and J. Gonzalez- Dominguez, “Deep neural networks for small footprint text-dependent speaker verification,” in ICASSP, 2014, pp. 4052–4056

    2014

  17. [25]

    Generalized end-to-end loss for speaker verification,

    L. Wan, Q. Wang, A. Papir, and I. L. Moreno, “Generalized end-to-end loss for speaker verification,” in ICASSP, 2018, pp. 4879–4883

    2018

  18. [26]

    X- vectors: Robust dnn embeddings for speaker recognition,

    D. Snyder, D. Garcia-Romero, G. Sell, D. Povey, and S. Khudanpur, “X- vectors: Robust dnn embeddings for speaker recognition,” in ICASSP, 2018, pp. 5329–5333

    2018

  19. [27]

    Speaker embedding extraction with phonetic information,

    Y . Liu, L. He, J. Liu, and M. T. Johnson, “Speaker embedding extraction with phonetic information,” in INTERSPEECH, 2018, pp. 2247–2251. 13

    2018

  20. [28]

    Phonetic-aware speaker embedding for far-field speaker verification,

    Z. Jin, Y . Tu, and M.-W. Mak, “Phonetic-aware speaker embedding for far-field speaker verification,” ArXiv, vol. abs/2311.15627, 2023

    2023 arXiv

  21. [29]

    ECAPA-TDNN: Emphasized Channel Attention, Propagation and Aggregation in TDNN Based Speaker Verification,

    B. Desplanques, J. Thienpondt, and K. Demuynck, “ECAPA-TDNN: Emphasized Channel Attention, Propagation and Aggregation in TDNN Based Speaker Verification,” in INTERSPEECH, 2020, pp. 3830–3834

    2020

  22. [30]

    TitaNet: Neural model for speaker representation with 1d depth-wise separable convolutions and global context,

    N. R. Koluguri, T. Park, and B. Ginsburg, “TitaNet: Neural model for speaker representation with 1d depth-wise separable convolutions and global context,” in ICASSP, 2021, pp. 8102–8106

    2021

  23. [31]

    (2024) BBC Rewind

    British Broadcasting Corporation. (2024) BBC Rewind. [Online]. Available: https://bbcrewind.co.uk/

    2024

  24. [32]

    The speakers in the wild (SITW) speaker recognition database,

    M. McLaren, L. Ferrer, D. Castan, and A. Lawson, “The speakers in the wild (SITW) speaker recognition database,” in INTERSPEECH, 2016, pp. 818–822

    2016

  25. [33]

    V oxCeleb: A Large-Scale Speaker Identification Dataset,

    A. Nagrani, J. S. Chung, and A. Zisserman, “V oxCeleb: A Large-Scale Speaker Identification Dataset,” in INTERSPEECH, 2017, pp. 2616– 2620

    2017

  26. [34]

    V oxCeleb: Large- scale speaker verification in the wild,

    A. Nagrani, J. S. Chung, W. Xie, and A. Zisserman, “V oxCeleb: Large- scale speaker verification in the wild,” Computer Speech & Language , vol. 60, p. 101027, 2020

    2020

  27. [35]

    Finding difficult speakers in automatic speaker recognition,

    L. L. Stoll, “Finding difficult speakers in automatic speaker recognition,” University of California, Berkeley, Electrical Engineering and Computer Sciences, Technical Report UCB/EECS-2011-152, 2011

    2011

  28. [36]

    pyannote.audio: neural building blocks for speaker diarization,

    H. Bredin, R. Yin, J. M. Coria, G. Gelly, P. Korshunov, M. Lavechin, D. Fustes, H. Titeux, W. Bouaziz, and M.-P. Gill, “pyannote.audio: neural building blocks for speaker diarization,” in ICASSP, 2020

    2020

  29. [37]

    End-to-end speaker segmentation for overlap-aware resegmentation,

    H. Bredin and A. Laurent, “End-to-end speaker segmentation for overlap-aware resegmentation,” in INTERSPEECH, 2021

    2021

  30. [38]

    Bidirectional recurrent neural networks,

    M. Schuster and K. Paliwal, “Bidirectional recurrent neural networks,” IEEE Transactions on Signal Processing, vol. 45, no. 11, pp. 2673–2681, 1997

    1997

  31. [39]

    Bidirectional LSTM networks for improved phoneme classification and recognition,

    A. Graves, S. Fern ´andez, and J. Schmidhuber, “Bidirectional LSTM networks for improved phoneme classification and recognition,” in ICANN, 2005

    2005

  32. [40]

    Phoneme recognition using time-delay neural networks,

    A. Waibel, T. Hanazawa, G. Hinton, K. Shikano, and K. Lang, “Phoneme recognition using time-delay neural networks,” IEEE Transactions on Acoustics, Speech, and Signal Processing , vol. 37, no. 3, pp. 328–339, 1989

    1989

  33. [41]

    Attentive statistics pooling for deep speaker embedding,

    K. Okabe, T. Koshinaka, and K. Shinoda, “Attentive statistics pooling for deep speaker embedding,” in INTERSPEECH, 2018, pp. 2252–2256

    2018

  34. [42]

    Arcface: Additive angular margin loss for deep face recognition,

    J. Deng, J. Guo, N. Xue, and S. Zafeiriou, “Arcface: Additive angular margin loss for deep face recognition,” in CVPR, 2019, pp. 4685–4694

    2019

  35. [43]

    Margin mat- ters: Towards more discriminative deep neural network embeddings for speaker recognition,

    X. Xiang, S. Wang, H. Huang, Y . Qian, and K. Yu, “Margin mat- ters: Towards more discriminative deep neural network embeddings for speaker recognition,” 2019 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) , pp. 1652–1656, 2019

    2019

  36. [44]

    Res2net: A new multi-scale backbone architecture,

    S. Gao, M. Cheng, K. Zhao, X. Zhang, M. Yang, and P. Torr, “Res2net: A new multi-scale backbone architecture,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 43, no. 02, pp. 652–662, 2021

    2021

  37. [45]

    Squeeze-and-excitation networks,

    J. Hu, L. Shen, and G. Sun, “Squeeze-and-excitation networks,” in 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2018, pp. 7132–7141

    2018

  38. [46]

    ContextNet: Improving Convolutional Neural Networks for Automatic Speech Recognition with Global Context,

    W. Han, Z. Zhang, Y . Zhang, J. Yu, C.-C. Chiu, J. Qin, A. Gulati, R. Pang, and Y . Wu, “ContextNet: Improving Convolutional Neural Networks for Automatic Speech Recognition with Global Context,” in INTERSPEECH, 2020, pp. 3610–3614

    2020

  39. [47]

    Quartznet: Deep automatic speech recognition with 1d time-channel separable convolutions,

    S. Kriman et al., “Quartznet: Deep automatic speech recognition with 1d time-channel separable convolutions,” in ICASSP, 2020, pp. 6124–6128

    2020

  40. [48]

    Multimodal Video Search by Examples (MVSE),

    H. Wang, M. Mulvenna, R. Bond et al., “Multimodal Video Search by Examples (MVSE),” EPSRC Grant Reference EP/V002740/2

  41. [49]

    Multi-modal video search by examples—a video quality impact analysis,

    G. Wu, A. Haider, X. Tian, E. Loweimi, C. H. Chan, M. Qian, A. Muhammad, I. Spence, R. Cooper, W. W. Y . Ng, J. Kittler, M. Gales, and H. Wang, “Multi-modal video search by examples—a video quality impact analysis,” IET Computer Vision , pp. 1–17, 2024

    2024

  42. [50]

    Robust signal-to-noise ratio estimation based on waveform amplitude distribution analysis,

    C. Kim and R. M. Stern, “Robust signal-to-noise ratio estimation based on waveform amplitude distribution analysis,” in INTERSPEECH, 2008, pp. 2598–2601

    2008

  43. [51]

    spaCy: Industrial-strength natural language processing in Python,

    E. AI, “spaCy: Industrial-strength natural language processing in Python,” 2015–. [Online]. Available: https://spacy.io

    2015

  44. [52]

    Beautiful soup: HTML and XML parsing for python,

    L. Richardson and M. L. A. Wicking, “Beautiful soup: HTML and XML parsing for python,” 2004–. [Online]. Available: https: //www.crummy.com/software/BeautifulSoup/

    2004

  45. [53]

    P. S. Foundation, Python unicodedata module , Python Software Foundation, Wilmington, DE, 2001–. [Online]. Available: https: //docs.python.org/3/library/unicodedata.html

    2001

  46. [54]

    SpeechBrain’s x-vector implementation for speaker embedding extraction,

    M. Ravanelli et al. , “SpeechBrain’s x-vector implementation for speaker embedding extraction,” Hugging Face, 2021, released on 05-03-21. [Online]. Available: https://huggingface.co/speechbrain/ spkrec-xvect-voxceleb

    2021

  47. [55]

    SpeechBrain’s ECAPA-TDNN implementation for speaker embedding extraction,

    ——, “SpeechBrain’s ECAPA-TDNN implementation for speaker embedding extraction,” Hugging Face, 2021, released on 05-03-21. [Online]. Available: https://huggingface.co/speechbrain/ spkrec-ecapa-voxceleb

    2021

  48. [56]

    ECAPA-TDNN model for speaker verification and diarization tasks,

    O. Kuchaiev et al. , “ECAPA-TDNN model for speaker verification and diarization tasks,” NeMo toolkit, 2023, version 1.16.0, Modified April 4, 2023, Size 85.79 MB. [Online]. Available: https://github.com/ NVIDIA/NeMo

    2023

  49. [57]

    TitaNet model for speaker verification and diarization tasks,

    ——, “TitaNet model for speaker verification and diarization tasks,” NeMo toolkit, 2023, version 1.10.0, Modified April 4, 2023, Size 96.91 MB. [Online]. Available: https://github.com/NVIDIA/NeMo

    2023

  50. [58]

    V oxCeleb2: Deep speaker recognition,

    J. S. Chung, A. Nagrani, and A. Zisserman, “V oxCeleb2: Deep speaker recognition,” in Proc. Interspeech 2018 , 2018, pp. 1086–1090

    2018

  51. [59]

    Librispeech: An asr corpus based on public domain audio books,

    V . Panayotov, G. Chen, D. Povey, and S. Khudanpur, “Librispeech: An asr corpus based on public domain audio books,” in ICASSP, 2015, pp. 5206–5210

    2015

  52. [60]

    Switchboard: telephone speech corpus for research and development,

    J. Godfrey, E. Holliman, and J. McDaniel, “Switchboard: telephone speech corpus for research and development,” in ICASSP, vol. 1, 1992, pp. 517–520 vol.1

    1992

  53. [61]

    Fisher english training speech parts 1 and 2 transcripts,

    C. Cieri, D. Graff, O. Kimball, D. Miller, and K. Walker, “Fisher english training speech parts 1 and 2 transcripts,” Linguistic Data Consortium (LDC), Philadelphia, Tech. Rep. LDC2004T19, LDC2005T19, 2004, 2005a

    2004

  54. [62]

    The NIST speaker recognition eval- uations: 1996-2001,

    A. F. Martin and M. A. Przybocki, “The NIST speaker recognition eval- uations: 1996-2001,” in Proc. The Speaker and Language Recognition Workshop (Odyssey 2001), 2001, pp. 39–43

    1996

  55. [63]

    A study on data augmentation of reverberant speech for robust speech recognition,

    T. Ko, V . Peddinti, D. Povey, M. L. Seltzer, and S. Khudanpur, “A study on data augmentation of reverberant speech for robust speech recognition,” in ICASSP, 2017, pp. 5220–5224

    2017

  56. [64]

    Proceedings of the 8th text retrieval conference,

    E. M. V oorhees, “Proceedings of the 8th text retrieval conference,” in TREC-8 Question Answering Track Report , 1999, pp. 77–82

    1999

  57. [65]

    Cumulated gain-based evaluation of ir techniques,

    K. J ¨arvelin and J. Kek ¨al¨ainen, “Cumulated gain-based evaluation of ir techniques,” ACM Transactions on Information Systems , vol. 20, no. 4, pp. 422–446, October 2002, cited on pages 51, 81

    2002

  58. [66]

    A theoretical analysis of NDCG ranking measures,

    Y . Wang, L. Wang, Y . Li, D. He, W. Chen, and T.-Y . Liu, “A theoretical analysis of NDCG ranking measures,” in Proceedings of the 26th Annual Conference on Learning Theory (COLT 2013) , 2013

    2013

  59. [67]

    C. D. Manning, P. Raghavan, and H. Sch ¨utze, Chapter 8: Evaluation in Information Retrieval. Cambridge University Press, 2009

    2009

  60. [68]

    Carterette and E

    B. Carterette and E. M. V oorhees, Overview of Information Retrieval Evaluation. Berlin, Heidelberg: Springer Berlin Heidelberg, 2011, pp. 69–85

    2011

  61. [69]

    Visualizing data using t-sne

    L. Van der Maaten and G. Hinton, “Visualizing data using t-sne.” Journal of Machine Learning Research (JMLR) , vol. 9, no. 11, 2008

    2008

  62. [70]

    Scikit-learn: Machine learning in Python,

    F. Pedregosa, G. Varoquaux, A. Gramfort, V . Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V . Dubourg, J. Vander- plas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duch- esnay, “Scikit-learn: Machine learning in Python,” Journal of Machine ...

    2011

  63. [71]

    Silhouettes: a graphical aid to the interpretation and validation of cluster analysis,

    P. J. Rousseeuw, “Silhouettes: a graphical aid to the interpretation and validation of cluster analysis,” Computational and Applied Mathematics, vol. 20, pp. 53–65, 1987

    1987

  64. [72]

    A binaural room impulse response database for the evaluation of dereverberation algorithms,

    M. Jeub, M. Sch ¨afer, and P. Vary, “A binaural room impulse response database for the evaluation of dereverberation algorithms,” in Proceed- ings of International Conference on Digital Signal Processing (DSP) . IEEE, 2009

    2009

  65. [73]

    Do we need dereverberation for hand-held telephony?

    M. Jeub, M. Sch ¨afer, H. Kr¨uger, C. M. Nelke, C. Beaugeant, and P. Vary, “Do we need dereverberation for hand-held telephony?” in International Congress on Acoustics (ICA) . Australian Acoustical Society, 2010

    2010

  66. [74]

    RIR-Generator,

    Nils Werner, “RIR-Generator,” https://pypi.org/project/rir-generator/, 2022, version 0.2.0

    2022

  67. [75]

    Room impulse response generator,

    E. A. Habets, “Room impulse response generator,” Technische Univer- siteit Eindhoven, Tech. Rep , vol. 2, no. 2.4, p. 1, 2006

    2006

  68. [76]

    Image method for efficiently simulating small-room acoustics,

    J. B. Allen and D. A. Berkley, “Image method for efficiently simulating small-room acoustics,” Journal of the Acoustical Society of America , vol. 65, no. 4, p. 943, April 1979

    1979

Pith tools

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