Challenging the Boundaries of Speech Recognition: The MALACH Corpus

Brian Kingsbury; George Saon; Kartik Audhkhasi; Michael Picheny; Xiaodong Cui; Z\'oltan T\"uske

REVIEW 3 major objections 5 minor 45 references

Challenging the Boundaries of Speech Recognition: The MALACH Corpus

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read A 375-hour corpus of Holocaust testimonies still defeats state-of-the-art speech recognition, with the best pure-play system reaching 21.7% word error rate.

desk verdict A useful, honest benchmark release for hard ASR, but the unfinished reference transcriptions make the headline 21.7% WER a number to quote with a caveat. read the letter →

arxiv 1908.03455 v1 pith:F2IE36GK submitted 2019-08-09 cs.CL cs.SDeess.AS

Michael Picheny , Zóltan Tüske , Brian Kingsbury , Kartik Audhkhasi , Xiaodong Cui , George Saon This is my paper · ORCID

classification cs.CLcs.SDeess.AS

keywords MALACHcorpusspeechrecognitionaccenteddisfluentemotionaloralhistoryarchivesLSTMacousticmodelworderrorrate

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 re-releases the MALACH corpus, a 375-hour collection of unconstrained, emotional, heavily accented and disfluent Holocaust testimonies, and proposes it as a benchmark for speech recognition systems that must cope with speech far outside clean read or conversational data. The authors argue that while modern deep learning has nearly solved tasks like Switchboard, MALACH still exposes large gaps in robustness to accents, disfluencies, age-related coarticulation, and emotional speech. To lower the barrier to entry, the paper defines training and test partitions, provides a lexicon and scoring files, and reports baseline results from a progression of systems, from classical HMMs through LSTMs with sequence training and a neural language model. The central empirical claim is that a system trained purely on 176 hours of MALACH transcripts reaches 21.7% WER on the minitest, compared to 32.1% for the best published result from the original MALACH program, which used 600 hours of data and interpolated language models. The paper's aim is to make this difficult, socially important data accessible so the community can build on these baselines.

What carries the argument

The central object is the MALACH corpus itself, now supplemented with a speech-recognition edition (LDC2019S11) that defines training and test partitions, a lexicon, a GLM file, and scoring setup. The load-bearing process is the construction of a pure-play baseline: 176 hours of manually transcribed speech for training, a 1.5-hour minitest (identical to the original MALACH test minus two unreleased conversations), a 4-gram language model with modified Kneser-Ney smoothing built solely from the training transcripts, and a sequence of acoustic models ranging from classical HMMs to a bi-directional LSTM with sMBR training and an LSTM language model. The key identity that carries the argument is the comparison of WER across this progression, which isolates the contribution of each modeling layer and establishes that even the strongest current technology leaves a large gap on this data.

What would settle it

A direct check would be to take the released minitest segments, have several independent expert transcribers produce new references without seeing the existing transcripts, and measure inter-annotator agreement. If the agreement is far below the reported 21.7% WER gap between systems, the comparison and the claimed improvements could be artifacts of reference noise rather than genuine progress. A more targeted test would be to rescore the same lattices with a resegmented gold-standard reference, since the paper says the current segmentation contains errors.

Watch

Extended reading notes

Core claim

The core discovery is that state-of-the-art speech recognition, when trained only on the 176 hours of manually transcribed MALACH speech and its 1.3M-token transcripts, achieves 21.7% word error rate on the 1.5-hour minitest, a substantial improvement over the 32.1% reported during the original MALACH project (which used 600 hours of transcribed and untranscribed data plus interpolated language models). The result is presented as a pure-play baseline: no external or augmented acoustic data is used, and the language model is built solely from the MALACH transcripts. The paper establishes a clear progression of systems—context-dependent HMM (40.8%), VTLN/FSA/MLLR (33.4%), fMMI+BMMI+MLLR (29.8%), DNN+XE (29.2%), CNN+XE (28.7%), DNN+HF (27.2%), CNN+HF (26.8%), LSTM (25.9%), LSTM+splicing (25.4%), +sMBR (23.9%), and finally +LSTM-LM (21.7%)—demonstrating that each modern technique contributes, yet the task remains far from solved, with the best system still making more than one error in five words on a corpus filled with accents, disfluencies, and emotional speech. The authors also highlight the societal motivation: making the roughly 115,000 hours of oral histories searchable requires accurate speech recognition, which remains the bottleneck.

Load-bearing premise

The manual transcriptions and segmentations used as training targets and test references are accurate enough that measured word error rates reflect speech recognition quality rather than annotation errors; the paper itself notes that the test segmentation still has accuracy issues and needs further verification passes (Section 6).

Editorial extensions

If this is right

If the MALACH baseline stands, any future speech recognition system that claims robustness to accents, disfluencies, or emotional speech can be evaluated against a fixed, publicly available setup, making progress measurable. The paper provides the exact partition, lexicon, GLM, and scoring files for this.
The 21.7% WER result implies that the remaining errors are concentrated in phenomena that current models handle poorly: disfluencies, partial words, heavy non-native accents, and emotional delivery. Systems that explicitly model these phenomena, rather than treating them as noise, are the next step.
The finding that an LSTM language model rescoring (from 23.9% to 21.7%) yields a significant gain, even though sentences are fragmented, suggests that better long-span modeling—potentially with resegmented test data—could close more of the gap.
The paper's comparison with 13.5% WER on Broadcast News and 9.19% on LibriSpeech (100-hour setup) frames MALACH as a stress test: a system that performs well on MALACH will likely be more robust in real-world conditions where speakers are elderly, accented, or emotionally distressed.
Because the full test set (3.1 hours) is still undergoing verification, the minitest serves as a provisional benchmark; once the gold-standard reference script is completed, results can be reported on the larger set, which contains 65 tagged emotional events not present in the minitest.

Reading between the lines

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

A natural extension is to test whether adaptation from a large general-purpose acoustic model, or interpolation with external language-model text, can push the pure-play 21.7% substantially lower; the paper hints this is plausible but does not attempt it.
The 522 tagged emotional events in the training data could be used to train an emotion-conditional or emotion-aware acoustic model, a direction the paper does not explore but which is directly supported by its released metadata.
The corpus's mix of accents, disfluencies, and emotional speech makes it a better proxy for clinical, legal, or archival transcription tasks than standard benchmarks; reporting results on MALACH alongside LibriSpeech would give a more honest picture of deployment readiness.
A testable extension would be to measure human transcription error rates on the same minitest segments, providing a human ceiling against which machine WER can be compared, following the kind of human-versus-machine analysis done for conversational telephone speech.

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

You should know this paper is a resource-release, not a methods contribution. The genuinely new thing is the LDC release of train/test partitions, a lexicon, a GLM scoring file, and baseline WERs for the MALACH corpus. That is useful, and the authors have done the community a real service by making the setup easy to adopt. The WER numbers are genuine measurements on a held-out minitest, and the best pure-play result (21.7% with an LSTM acoustic model plus LSTM LM, trained only on 176 hours of MALACH transcripts) is a solid improvement over the old 32.1% from the original program. However, note that the old number used 600 hours of data and an interpolated LM, so the comparison is not apples-to-apples; the improvement is real but partly reflects the extra data the original system had.

The paper does several things well. It describes the pipeline in enough detail to reproduce, runs a sensible ladder of systems from CD-HMM to LSTM, and is refreshingly candid about limitations. Section 6 admits that manual segmentation accuracy is imperfect and that a gold-standard reference script is still needed. That honesty is a point in its favor.

The soft spot is exactly that admitted limitation. The headline 21.7% is measured against reference transcriptions and segmentations that the authors themselves say are unfinished. The test set is only 1.5 hours, and the full test set is still under verification. For a benchmark, this is a load-bearing issue: the community will start quoting 21.7% as the number to beat, but systematic errors in the reference could bias it. This does not undermine the qualitative conclusion that MALACH is a hard task for modern ASR, but it means the precise WER is not yet a stable target. The representativeness of the minitest also rests on informal experiments, which is thin.

Who is this for? Researchers working on accented, disfluent, or emotional speech who want a public benchmark with lower barriers to entry. They should use the released setup but treat the reference as provisional and verify or fix it as part of their own work. This paper deserves a serious referee: it is a citable resource with reproducible baselines, and its limitations are stated up front rather than hidden. I would accept it for peer review with the expectation that the final version includes a more prominent warning about reference quality.

Referee Report

3 major / 5 minor

Summary. The paper reintroduces the MALACH corpus, a 375-hour subset of Holocaust survivor testimonies, as a challenging benchmark for accented, disfluent, and emotional speech recognition. It defines training and test partitions, releases a lexicon, a GLM file, and scoring resources through the LDC, and presents baseline results from a progression of systems ranging from context-dependent HMMs to LSTM-based hybrids with sequence training and an LSTM language model. The best 'pure-play' system, trained only on 176 hours of MALACH transcripts, reports 21.7% WER on a 1.5-hour minitest, compared with 32.1% for the best result from the original MALACH program. The paper claims that these results demonstrate both the progress of modern ASR and the continued difficulty of this type of spontaneous, heavily accented speech.

Significance. If the reported benchmark is reliable, this paper has clear value for the ASR community: it lowers the barrier to entry on a unique corpus combining disfluencies, heavy accents, code-switching, and emotional speech, and it establishes reproducible training/test setups and a reference lexicon. The explicit release of metadata and scoring files is a concrete, citable contribution. The paper is also honest about its own limitations, stating in Section 6 that the reference segmentations are not yet gold standard and that verification passes remain. The main significance rests on the headline comparison between current deep-learning baselines and historical MALACH results, which is a useful calibration point if the reference quality concern is addressed. The work is largely empirical, with no theoretical machinery; its strengths are the released resource, the documented experimental pipeline, and the direct comparison across technology generations.

major comments (3)

[Section 6] The paper explicitly states that 'a number of issues still remained with respect to the accuracy of the manual segmentation' and that 'multiple additional verification passes are needed to really obtain a gold standard reference script even for the minitest.' Since every WER in Table 1 and the headline comparison to the original 32.1% result are computed against this same reference, the reported absolute numbers may be biased and the minitest cannot yet serve as a stable community target. The authors should either complete the verification passes before final release or clearly label all reported WERs as preliminary and provide a sensitivity analysis showing how results change under plausible segmentation or transcription corrections.
[Section 3] The representativeness of the 1.5-hour minitest rests on informal experiments described only as showing no appreciable change in overall WER when the test set was abbreviated from the original MALACH test data. No details are given regarding the number of conversations, the duration, the WERs on the original versus abbreviated sets, or the variability across subsets. Because the minitest is the sole evaluation set for all baseline results, the paper should document these experiments quantitatively and ideally provide bootstrap confidence intervals to support the claim that the shortened test set is a faithful proxy.
[Table 1 and Section 5] The paper reports incremental WER gains between systems (e.g., 25.9% to 25.4% with splicing, 25.4% to 23.9% with sMBR, and 23.9% to 21.7% with the LSTM LM) and describes them as significant, but no confidence intervals or significance tests are provided. On a test set of only 1.5 hours and roughly 26K tokens, several differences are small and could be within scoring noise. The authors should report bootstrap or matched-pair confidence intervals for the key comparisons, or at least state the statistical uncertainty of the reported WERs.

minor comments (5)

[Section 4.3] There is a typo in the LSTM configuration paragraph: 'droput factor' should be 'dropout factor.'
[Section 4.1] There are minor typos: 'occured' should be 'occurred' and 'processsing' should be 'processing.'
[Table 1] The table header reads 'MALACH 50-Hour minitest,' which is confusing because Section 3 defines the minitest as 1.5 hours; please clarify whether '50-Hour' refers to the training data condition, the Broadcast News setup, or something else.
[Section 6] There is a typo in the final paragraph: 'peformance' should be 'performance.'
[Section 3] The sentence 'This leaves 102 additional conversations that can be used for a broader test set to evaluate aspects of disfluencies, emotional speech, etc., or rolled into the training data' is slightly ambiguous because it follows the count 674 + 8 = 682 from 784 total interviews; the arithmetic is correct, but the phrasing could be clarified to indicate that the 102 conversations are in addition to the 682 used.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper reports held-out WER measurements; no prediction reduces to its training inputs or to self-citations.

full rationale

The paper's central claims are empirical benchmarks on a held-out minitest. Acoustic and language models are trained on 176 hours of labeled interviews and evaluated on a 1.5-hour minitest (Section 3), with scoring via SCTK; the test set is not used for fitting. The WER numbers in Table 1 are measurements, not derivations, and no equation or fitted parameter is renamed as a prediction. The comparison to earlier MALACH results (43.8% to 32.1%) cites the authors' own prior publications [20,21], but those citations function as historical records of previous system performance, not as justifications that force the current result; the current result would stand even if those numbers were different. The only flagged weakness, stated in Section 6, is that 'a number of issues still remained with respect to the accuracy of the manual segmentation' and a gold-standard reference needs additional verification passes. That is a data-quality/validity caveat about the reference transcripts, not a circularity: the measured WER is still computed against an external reference rather than being defined as an input. There is no self-definitional loop, no fitted input called prediction, and no imported uniqueness theorem. Accordingly, the circularity score is 0.

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

This is an empirical benchmark paper. There are no physical constants or fitted theoretical parameters. The reported WERs are measurements on a defined test set. The main load-bearing assumptions concern data quality and test-set representativeness, which the paper openly flags.

assumptions (2)

domain assumption Manual transcriptions and segmentations in the LDC corpus are sufficiently accurate to serve as reference labels for WER evaluation.
All baseline WERs depend on this; the paper itself questions it in Section 6, noting segmentation accuracy issues and the need for further verification passes.
domain assumption The 1.5-hour minitest is representative of the full 3.1-hour test set.
Section 3 states this is supported by informal experiments, but no quantitative evidence is shown.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Challenging the Boundaries of Speech Recognition: The MALACH Corpus." pith.science (2026). https://pith.science/paper/F2IE36GK

@misc{pith2026190803455,
  author       = {Pith},
  title        = {Pith review of: Challenging the Boundaries of Speech Recognition: The MALACH Corpus},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F2IE36GK}},
  note         = {Machine review of arXiv:1908.03455}
}

read the original abstract

There has been huge progress in speech recognition over the last several years. Tasks once thought extremely difficult, such as SWITCHBOARD, now approach levels of human performance. The MALACH corpus (LDC catalog LDC2012S05), a 375-Hour subset of a large archive of Holocaust testimonies collected by the Survivors of the Shoah Visual History Foundation, presents significant challenges to the speech community. The collection consists of unconstrained, natural speech filled with disfluencies, heavy accents, age-related coarticulations, un-cued speaker and language switching, and emotional speech - all still open problems for speech recognition systems. Transcription is challenging even for skilled human annotators. This paper proposes that the community place focus on the MALACH corpus to develop speech recognition systems that are more robust with respect to accents, disfluencies and emotional speech. To reduce the barrier for entry, a lexicon and training and testing setups have been created and baseline results using current deep learning technologies are presented. The metadata has just been released by LDC (LDC2019S11). It is hoped that this resource will enable the community to build on top of these baselines so that the extremely important information in these and related oral histories becomes accessible to a wider audience.

Figures

Figures reproduced from arXiv: 1908.03455 by the authors.

**Figure 1.** Large Spoken Archive Search System Architecture proposed in MALACH [14] these various phenomena. For example, there are disfluencies in SWITCHBOARD, accented speech in the Mozilla corpus [10], and very informal speech in the AMI corpus [11]. However, very few corpora demonstrate all these phenomena - SWITCHBOARD is relatively accent free, the Mozilla Corpus is read speech, etc. The 1996/1997 LDC releases of the Bro… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 45 canonical work pages

[1]

Introduction There has been huge progress in speech recognition over the last several years. Tasks previously considered merely hard, such as open vocabulary voice search and voice messaging, are now in wide deployment across popular consumer devices such as smartphones [3] and smart speakers [4]. Tasks once thought extremely difﬁcult, such as SWITCHBOARD...

work page Pith review arXiv 1996
[2]

Prior Speech Recognition Results The MALACH project ran from 2001-2006. As that time, the dominant speech recognition technology was speaker-adaptive processing [17], sometimes combined with more sophisticated techniques such as MMI/MPE training [18] or (by the end of the project) fMPE processing [19]. Most of these technologies have now been subsumed by ...

work page 2001
[3]

There were many ways in which the 784 interviews could have been divided into training and test data

Proposed Training and Testing Partitions The MALACH corpus previously released through the LDC consists of selections from 784 interviews ranging from approx- imately 15 minutes to 30 minutes in length. There were many ways in which the 784 interviews could have been divided into training and test data. In the interest of continuity with prior work [20, 1...

work page
[4]

Since then, due to the success of deep learning, there has been a major revolution in speech recognition

Baseline Results As described in Section 2, the original MALACH results were based on an older generation of speech recognition technology. Since then, due to the success of deep learning, there has been a major revolution in speech recognition. Systems today are almost unrecognizable from those of 10-15 years ago, and tech- nology changes on almost a dai...

work page
[5]

To obtain better complementar- ity, a 64-dimensional ﬁlter bank was created and used to extract logmel, delta, and delta-delta parameters

Additional Recognition Results In [5] one of the simpler but more successful techniques to produce acoustic model improvements was the application of feature fusion - speciﬁcally, combining logmel features with feature-space adapted features. To obtain better complementar- ity, a 64-dimensional ﬁlter bank was created and used to extract logmel, delta, and...

work page
[6]

pure play

Discussion As can be seen from the above results, although recent tech- nology advances have made signiﬁcant inroads in performance, MALACH remains a challenging speech recognition task. To put things in perspective, the error rate on the popular Lib- rispeech read speech corpus when trained using the 100-hour clean subset of the training data using a DNN...

work page
[7]

A range of systems were built spanning traditional HMMs all the way to hybrid LSTM- based acoustic and language models

Summary The MALACH corpus is re-introduced as an important cor- pus because of its societal interest and to challenge the speech recognition community in areas such as modeling of accents, disﬂuencies, and emotional speech. A range of systems were built spanning traditional HMMs all the way to hybrid LSTM- based acoustic and language models. The best syst...

work page
[8]

This work would not be possible without the support of Sam Gustman from the Visual History Foundation

Acknowledgements We are grateful to our former colleague, Bhuvana Ramabhad- ran, who did most of the original speech recognition research on MALACH during the course of the MALACH project, includ- ing all the data preparation. This work would not be possible without the support of Sam Gustman from the Visual History Foundation. Thanks also goes to Doug Oa...

work page

Show all 45 references

[9]

Librispeech: an ASR corpus based on public domain audio books,

Vassil Panayotov, Guoguo Chen, Daniel Povey, and Sanjeev Khu- danpur, “Librispeech: an ASR corpus based on public domain audio books,” in IEEE International Conference on Acoustics, Speech and Signal Processing. IEEE, 2015, pp. 5206–5210

2015
[10]

USC-SFI MALACH Interviews and Transcripts English,

“USC-SFI MALACH Interviews and Transcripts English,” https://catalog.ldc.upenn.edu/LDC2012S05, 2012

2012
[11]

USC-SFI MALACH Interviews and Transcripts English - Speech Recognition Edition,

“USC-SFI MALACH Interviews and Transcripts English - Speech Recognition Edition,” https://catalog.ldc. upenn.edu/LDC2019S11, 2019

2019
[12]

State-of-the-art speech recognition with sequence-to-sequence models,

Chung-Cheng Chiu, Tara N Sainath, Yonghui Wu, Rohit Prab- havalkar, Patrick Nguyen, Zhifeng Chen, Anjuli Kannan, Ron J Weiss, Kanishka Rao, Ekaterina Gonina, et al., “State-of-the-art speech recognition with sequence-to-sequence models,” in IEEE International Conference on Aco...

2018
[13]

Contextual ASR adaptation for conversational agents,

Anirudh Raju, Behnam Hedayatnia, Linda Liu, Ankur Gandhe, Chandra Khatri, Angeliki Metallinou, Anu Venkatesh, and Ariya Rastrow, “Contextual ASR adaptation for conversational agents,” in Interspeech, 2018

2018
[14]

All results presented in this paper are on this minitest, as the full test set is still undergoing veriﬁca- tion

except for the two conversations (four speakers) that were not released publicly. All results presented in this paper are on this minitest, as the full test set is still undergoing veriﬁca- tion. The training vocabulary covered 98.1% of the tokens in the minitest
[15]

English conversational telephone speech recognition by humans and ma- chines,

George Saon, Gakuto Kurata, Tom Sercu, Kartik Audhkhasi, Samuel Thomas, Dimitrios Dimitriadis, Xiaodong Cui, Bhuvana Ramabhadran, Michael Picheny, Lynn-Li Lim, et al., “English conversational telephone speech recognition by humans and ma- chines,” in Interspeech, 2017

2017
[16]

Comparing human and ma- chine errors in conversational speech transcription,

Andreas Stolcke and Jasha Droppo, “Comparing human and ma- chine errors in conversational speech transcription,” inEighteenth Annual Conference of the International Speech Communication Association, 2017

2017
[17]

Speech recognition: What’s left?,

Michael Picheny, “Speech recognition: What’s left?,” in Text, Speech, and Dialogue: 20th International Conference, TSD 2017, Prague, Czech Republic, August 27-31, 2017, Proceedings . 2017, Springer

2017
[18]

The design for the Wall Street Journal-based CSR corpus,

Douglas B Paul and Janet M Baker, “The design for the Wall Street Journal-based CSR corpus,” in Proceedings of the work- shop on Speech and Natural Language . Association for Compu- tational Linguistics, 1992, pp. 357–362

1992
[19]

Mozilla common voice data,

“Mozilla common voice data,” https://voice.mozilla. org/en/data
[20]

Announcing the AMI meeting corpus,

Jean Carletta, “Announcing the AMI meeting corpus,” The ELRA Newsletter, vol. 11, no. 1, pp. 3–5, 2006

2006
[21]

The devel- opment of the 1996 HTK broadcast news transcription system,

PC Woodland, MJF Gales, D Pye, and SJ Young, “The devel- opment of the 1996 HTK broadcast news transcription system,” in DARPA speech recognition workshop. Morgan Kaufmann Pub, 1997, pp. 73–78

1996
[22]

The MGB challenge: Evalu- ating multi-genre broadcast media recognition,

Peter Bell, Mark JF Gales, Thomas Hain, Jonathan Kilgour, Pierre Lanchantin, Xunying Liu, Andrew McParland, Steve Renals, Os- car Saz, Mirjam Wester, et al., “The MGB challenge: Evalu- ating multi-genre broadcast media recognition,” in 2015 IEEE Workshop on Automatic Speech Re...

2015
[23]

Automatic recognition of spontaneous speech for access to multilingual oral history archives,

William Byrne, David Doermann, Martin Franz, Samuel Gust- man, Jan Hajic, Douglas Oard, Michael Picheny, Josef Psutka, Bhuvana Ramabhadran, Dagobert Soergel, et al., “Automatic recognition of spontaneous speech for access to multilingual oral history archives,” IEEE Transactio...

2004
[24]

USC Shoah foundation,

“USC Shoah foundation,” http://sfi.usc.edu/
[25]

USC-SFI MALACH Interviews and Transcripts Czech,

“USC-SFI MALACH Interviews and Transcripts Czech,” https://catalog.ldc.upenn.edu/LDC2014S04, 2014

2014
[26]

Maximum likelihood linear transformations for HMM-based speech recognition,

M.J.F Gales, “Maximum likelihood linear transformations for HMM-based speech recognition,” Computer Speech and Lan- guage, vol. 12, pp. 75–98, 1998

1998
[27]

Minimum phone error and I- smoothing for improved discriminative training,

D. Povey and P. C. Woodland, “Minimum phone error and I- smoothing for improved discriminative training,” in Proc. of ICASSP, 2002, pp. 105–108

2002
[28]

fMPE: Discriminatively trained features for speech recognition,

D. Povey, B. Kingsbury, L. Mangu, G. Saon, H. Soltau, and G. Zweig, “fMPE: Discriminatively trained features for speech recognition,” in Proc. of ICASSP, 2005, pp. 961–964

2005
[29]

To- wards automatic transcription of large spoken archives-English ASR for the MALACH project,

Bhuvana Ramabhadran, Jing Huang, and Michael Picheny, “To- wards automatic transcription of large spoken archives-English ASR for the MALACH project,” inAcoustics, Speech, and Signal Processing, 2003. Proceedings.(ICASSP’03). 2003 IEEE Interna- tional Conference on. IEEE, 2003...

2003
[30]

Exploiting large quantities of sponta- neous speech for unsupervised training of acoustic models,

Bhuvana Ramabhadran, “Exploiting large quantities of sponta- neous speech for unsupervised training of acoustic models,” in Ninth European Conference on Speech Communication and Tech- nology, 2005

2005
[31]

CMU Pronunciation Dictionary version 0.7b,

“CMU Pronunciation Dictionary version 0.7b,” https: //github.com/Alexir/CMUdict/blob/master/ cmudict-0.7b
[32]

Conditional and joint models for grapheme-to- phoneme conversion,

Stanley F Chen, “Conditional and joint models for grapheme-to- phoneme conversion,” in Eighth European Conference on Speech Communication and Technology, 2003

2003
[33]

An empirical study of smoothing techniques for language modeling,

Stanley F. Chen and Joshua Goodman, “An empirical study of smoothing techniques for language modeling,” Computer Speech & Language, vol. 13, no. 4, pp. 359–393, 1999

1999
[34]

Evaluation of proposed modiﬁca- tions to mpe for large scale discriminative training,

D. Povey and.B. Kingsbury, “Evaluation of proposed modiﬁca- tions to mpe for large scale discriminative training,” in Proc. of ICASSP, 2007, vol. IV , pp. 321–324

2007
[35]

The IBM Attila speech recognition toolkit,

H. Soltau, G. Saon, and B. Kingsbury, “The IBM Attila speech recognition toolkit,” in Proc. of IEEE Workshop on Spoken Lan- guage Technology (SLT), 2010, pp. 97–102

2010
[36]

Pytorch 0.4.0,

“Pytorch 0.4.0,” https://pytorch.org/
[37]

Deep convolutional neural networks for large-scale speech tasks,

Tara N Sainath, Brian Kingsbury, George Saon, Hagen Soltau, Abdel-rahman Mohamed, George Dahl, and Bhuvana Ramabhad- ran, “Deep convolutional neural networks for large-scale speech tasks,” Neural Networks, vol. 64, pp. 39–48, 2015

2015
[38]

Scalable minimum Bayes risk training of deep neural network acoustic models us- ing distributed Hessian-free optimization,

B. Kingsbury, T. Sainath, and H. Soltau, “Scalable minimum Bayes risk training of deep neural network acoustic models us- ing distributed Hessian-free optimization,” in Proc. Interspeech, 2012

2012
[39]

NIST scoring software SCTK 2.4.10,

“NIST scoring software SCTK 2.4.10,” http://www. openslr.org/4/
[40]

Lattice-based optimization of sequence classiﬁ- cation criteria for neural-network acoustic modeling,

B. Kingsbury, “Lattice-based optimization of sequence classiﬁ- cation criteria for neural-network acoustic modeling,” in Proc. of ICASSP, 2009, pp. 3761–3764

2009
[41]

Reg- ularizing and optimizing LSTM language models,

Stephen Merity, Nitish Shirish Keskar, and Richard Socher, “Reg- ularizing and optimizing LSTM language models,” in Interna- tional Conference on Learning Representations (ICLR) , 2018

2018
[42]

Dropout: A simple way to prevent neural networks from overﬁtting,

Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov, “Dropout: A simple way to prevent neural networks from overﬁtting,” Journal of Machine Learning Research, vol. 15, pp. 1929–1958, 2014

1929
[43]

Regularization of neural networks using dropconnect,

Li Wan, Matthew Zeiler, Sixin Zhang, Yann Le Cun, and Rob Fergus, “Regularization of neural networks using dropconnect,” in International Conference on Machine Learning (ICML) , 2013, vol. 28, pp. 1058–1066

2013
[44]

Kaldi results on Librispeech corpus,

“Kaldi results on Librispeech corpus,” https://github. com/kaldi-asr/kaldi/blob/master/egs/ librispeech/s5/RESULTS
[45]

Kaldi results on AMI corpus,

“Kaldi results on AMI corpus,” https://github.com/ kaldi-asr/kaldi/blob/master/egs/ami/s5b/ RESULTS

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Introduction There has been huge progress in speech recognition over the last several years. Tasks previously considered merely hard, such as open vocabulary voice search and voice messaging, are now in wide deployment across popular consumer devices such as smartphones [3] and smart speakers [4]. Tasks once thought extremely difﬁcult, such as SWITCHBOARD...

work page Pith review arXiv 1996

[2] [2]

Prior Speech Recognition Results The MALACH project ran from 2001-2006. As that time, the dominant speech recognition technology was speaker-adaptive processing [17], sometimes combined with more sophisticated techniques such as MMI/MPE training [18] or (by the end of the project) fMPE processing [19]. Most of these technologies have now been subsumed by ...

work page 2001

[3] [3]

There were many ways in which the 784 interviews could have been divided into training and test data

Proposed Training and Testing Partitions The MALACH corpus previously released through the LDC consists of selections from 784 interviews ranging from approx- imately 15 minutes to 30 minutes in length. There were many ways in which the 784 interviews could have been divided into training and test data. In the interest of continuity with prior work [20, 1...

work page

[4] [4]

Since then, due to the success of deep learning, there has been a major revolution in speech recognition

Baseline Results As described in Section 2, the original MALACH results were based on an older generation of speech recognition technology. Since then, due to the success of deep learning, there has been a major revolution in speech recognition. Systems today are almost unrecognizable from those of 10-15 years ago, and tech- nology changes on almost a dai...

work page

[5] [5]

To obtain better complementar- ity, a 64-dimensional ﬁlter bank was created and used to extract logmel, delta, and delta-delta parameters

Additional Recognition Results In [5] one of the simpler but more successful techniques to produce acoustic model improvements was the application of feature fusion - speciﬁcally, combining logmel features with feature-space adapted features. To obtain better complementar- ity, a 64-dimensional ﬁlter bank was created and used to extract logmel, delta, and...

work page

[6] [6]

pure play

Discussion As can be seen from the above results, although recent tech- nology advances have made signiﬁcant inroads in performance, MALACH remains a challenging speech recognition task. To put things in perspective, the error rate on the popular Lib- rispeech read speech corpus when trained using the 100-hour clean subset of the training data using a DNN...

work page

[7] [7]

A range of systems were built spanning traditional HMMs all the way to hybrid LSTM- based acoustic and language models

Summary The MALACH corpus is re-introduced as an important cor- pus because of its societal interest and to challenge the speech recognition community in areas such as modeling of accents, disﬂuencies, and emotional speech. A range of systems were built spanning traditional HMMs all the way to hybrid LSTM- based acoustic and language models. The best syst...

work page

[8] [8]

This work would not be possible without the support of Sam Gustman from the Visual History Foundation

Acknowledgements We are grateful to our former colleague, Bhuvana Ramabhad- ran, who did most of the original speech recognition research on MALACH during the course of the MALACH project, includ- ing all the data preparation. This work would not be possible without the support of Sam Gustman from the Visual History Foundation. Thanks also goes to Doug Oa...

work page