Weak Supervision Techniques towards Enhanced ASR Models in Industry-level CRM Systems

Jiamou Liu; Jia Wang; Sijie Wang; Yung-I Liang; Yuxi Zhang; Zhongsheng Wang

REVIEW 3 major objections 5 minor 22 references

Weak Supervision Techniques towards Enhanced ASR Models in Industry-level CRM Systems

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper claims that fine-tuning ASR models on synthetic voice-text pairs generated by an LLM and a TTS model, then filtered by a pre-trained ASR model, cuts industrial CRM transcription errors by up to 63 percent (51.5 percent on…

desk verdict A sensible applied ASR data-augmentation recipe with a test-set-selection problem that makes the headline 63%/51.5% gains untrustworthy. read the letter →

arxiv 2507.16843 v1 pith:JLA5VXYH submitted 2025-07-20 cs.SD cs.AIcs.CLeess.AS

Zhongsheng Wang , Sijie Wang , Jia Wang , Yung-I Liang , Yuxi Zhang , Jiamou Liu This is my paper

classification cs.SDcs.AIcs.CLeess.AS

keywords weaksupervisionASRinCRMdataaugmentationsyntheticspeechmodelfine-tuningintegratederrorratelow-resourceindustrialapplication

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

Industry speech recognition usually needs large labeled corpora that companies cannot afford. The authors claim that a weak-supervision loop can replace most of that data: a large language model imitates a small set of real transcripts and a keyword list to generate thousands of synthetic text labels, a text-to-speech model turns them into audio, and a strong pre-trained ASR model filters out pairs whose recognition error is too high. The surviving synthetic pairs are used to fine-tune three versions of a pre-trained ASR model with a parameter-efficient adapter. Reported results on a luxury-retail CRM test set show up to 63% improvement over the native model, with an average improvement of 51.5%, from only 100 real samples per brand. The paper also proposes the integrated error rate (IER), a metric that weights word, character, and industry-keyword errors for mixed Chinese-foreign speech.

What carries the argument

The mechanism that carries the argument is a generate-filter-fine-tune loop. Starting from 100 real voice-text pairs and a curated keyword list, a large language model produces new text labels in the style of the real labels while covering keyword categories; a text-to-speech model reads those labels as audio, with dialect and accent noise added. A strong pre-trained ASR model then transcribes each synthetic clip, and pairs whose character error rate against the generated label exceeds a fixed threshold are discarded, leaving a filtered synthetic dataset of roughly 20,000 samples. A parameter-efficient low-rank adapter is used to fine-tune three versions of a pre-trained ASR model on this data with modest GPU requirements. A second named component is the integrated error rate (IER), which decomposes each transcription into words, characters, and industry keywords, weights their error rates by their frequencies, averages over two word-segmentation tools, and de-duplicates overlapping errors, yielding a single score for mixed-language industrial speech.

What would settle it

Repeat the pipeline with a separate validation set used only to select the epoch count and CER threshold, then evaluate on a disjoint test set that was never touched during development; if the up-to-63% and average-51.5% improvements disappear, the central claim fails.

Watch

Extended reading notes

Core claim

The central claim is that high-quality domain-specific ASR can be produced by synthetic data alone, as long as the synthetic corpus is generated from real examples and filtered by a pre-existing ASR model. In the paper's CRM setting, sales staff dictate customer portraits that mix Chinese with untranslated luxury brand names, and the authors show that models fine-tuned on 20,000 synthetic samples generated from 100 real samples per brand substantially outperform the unmodified pre-trained models on the same test set. They attribute the gain to the keyword-driven label generation, which concentrates training on the words general models miss, and to the CER filter, which removes synthetic audio the pre-trained model cannot reliably transcribe. The best fine-tuned model also achieves near-error-free output on some test utterances, and the authors report that the method has been adopted in their industrial deployment. IER is introduced as a more comprehensive evaluation for such hybrid speech, averaging weighted word, character, and keyword errors across multiple Chinese word-segmentation tools with de-duplication of overlapping errors.

Load-bearing premise

The reported gains are measured on a test set that was also used to choose the training epochs and the filtering threshold, so the numbers may reflect tuning to that particular test set rather than a general property of the method.

Editorial extensions

If this is right

Organizations with as few as 100 labeled utterances per domain can produce a domain-tuned ASR model with roughly half the transcription error of the generic base model.
Because the keyword list is part of the generation prompt, the pipeline automatically injects brand names, product lines, and colloquial terms into the training data, targeting exactly the vocabulary that generic ASR fails on.
The CER-based filtering step gives a reusable quality gate for any synthetic speech corpus, independent of the CRM application.
The IER metric provides a single-number benchmark for code-mixed speech where WER and CER disagree, which could support more consistent evaluation of multilingual industrial ASR.

Reading between the lines

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

A testable extension would filter the synthetic corpus with a different pre-trained ASR model than the one being fine-tuned; if the reported gains shrink, part of the improvement comes from the filter shaping data toward its own recognition biases.
The same recipe likely transfers to other jargon-heavy verticals such as finance, healthcare, or technical support, where the main per-domain cost is building the keyword list and manually verifying a small seed corpus.
If the LLM imitates the seed transcripts too closely, the synthetic corpus may under-represent spontaneous speech, so real-world gains could be smaller than test-set numbers; replacing part of the synthetic audio with lightly augmented real audio would test this.
IER's keyword weighting could be adopted more broadly as a benchmark for code-mixed speech, provided the keyword list is fixed before evaluation so the metric is not tuned to a specific model's mistakes.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Quick take: the pipeline is a reasonable industry recipe, but the headline numbers are not clean estimates. The paper fine-tunes Whisper models on LLM-generated text plus ChatTTS audio, filters synthetic samples with a CER threshold, and uses LoRA; on a 1000-utterance luxury CRM test set it reports CER improvements up to 63%. The composition is not novel component-wise—TTS augmentation and LLM-generated text are both in the cited literature—but the specific application and the careful comparison against real-only fine-tuning are useful. Table 3 shows synthetic fine-tuning beats both native models and real-data fine-tuning in most rows; that directional evidence is the paper's real value. The IER metric is an attempt to handle Chinese-brand-name mixed speech; it is not well-defined enough to reproduce, and manual verification of all indicators makes it an effortful but qualitative check.

The soft spots are real. Section 8.2 picks the epoch count (5) by fine-tuning to 10/15/20/30 and choosing the one with the best CER on the same Test Set used to report final results in Table 3. That makes the final CER numbers optimistically biased. The stress-test note also worries the CER filtering threshold was selected on the test set; reading Section 8.2, the threshold plot is about remaining synthetic data, so that specific concern is weaker. A separate issue: the filtering model is whisper-large-v2, the same model that gets the best final results, so the synthetic data is selectively enriched with examples the target model already handles. That overstates gains for that model and leaves the method's generality untested. No confidence intervals or significance tests, no held-out validation, and no released code/data; the 200-sample IER subset is unspecified. These are not fatal to the idea, but they are fatal to the precise claim of 63%/51.5%.

Who is this for? Engineers building domain-specific ASR systems with little labeled data, and researchers studying synthetic data pipelines. It is not a conceptual breakthrough. My read is that it deserves a serious referee, not a desk rejection, but the revision needs a proper validation split, repeated runs with variance, and a filter model that is not the final model. I would not cite it as evidence for a general method, but I would mention it as an applied data point when discussing low-resource ASR augmentation.

Referee Report

3 major / 5 minor

Summary. The paper proposes a weak-supervision pipeline for fine-tuning ASR models in domain-specific CRM systems. Starting from a small set of real labeled audio examples and a curated keyword list, the authors use an LLM (DeepSeek V2) to generate synthetic transcripts and a TTS model (ChatTTS) to synthesize speech, then filter the synthetic corpus with a pre-trained ASR model's CER score. The filtered synthetic data (10,000 samples per brand) are used to LoRA-fine-tune Whisper models (medium, large-v2, large-v3). Evaluation on a 1,000-utterance luxury-retail test set reports CER/WER improvements, with the best fine-tuned models achieving up to 63% CER reduction and an average improvement of 51.5% over native models; the paper also introduces an Integrated Error Rate (IER) metric for mixed Chinese/foreign-keyword speech and reports a real industrial deployment.

Significance. If the reported results are clean, the paper makes a useful practical contribution: it demonstrates that LLM-plus-TTS synthetic data generation combined with ASR-based filtering can improve domain-specific ASR with only 100 real labeled samples per domain. The proposed IER metric addresses a genuine evaluation need for multilingual, keyword-heavy industrial speech, and the deployment at Atom Intelligence gives the work ecological validity. The paper is less novel methodologically, as TTS-based data augmentation and self-training are known techniques, but the specific combination for CRM luxury-retail ASR is a reasonable engineering contribution. However, the significance is conditional on the evaluation protocol being sound; the current manuscript has several load-bearing methodological gaps.

major comments (3)

[§8.2 and §6.4, Table 3] The same Test Set of 1,000 utterances is used both for selecting hyperparameters and for reporting final performance. Section 8.2 states that the number of fine-tuning epochs (5) was chosen by further fine-tuning and computing CER metrics, and Figure 6b is used to choose 0.15 as the CER filtering boundary; Table 3 then reports CER on that same test set. Because the threshold determines which synthetic data enter the training set and the epoch count controls fitting to synthetic data, the reported 63% highest and 51.5% average improvements are optimistically biased and are not a clean estimate of the method's expected generalization. Please separate a validation set for epoch and threshold selection and report test-set results only after those choices are fixed.
[§6.4, Table 4] The IER metric is computed on an unspecified subset: Section 6.4 states 'we reduced the amount of data in the test set to 200' without describing how the 200 utterances were selected. Table 3 uses the full 1,000-utterance test set, while Table 4 reports IER only for this smaller subset, making the two tables not directly comparable. In addition, Table 4 provides no native-model IER baseline, so the claimed IER advantage cannot be quantified. Please define the subset selection procedure, report IER on the full test set or justify the reduction, and give confidence intervals or significance tests.
[§5.1 and §9] The data filtering step uses whisper-large-v2 as the quality checker, and the best-performing fine-tuned models in Table 3 are also whisper-large-v2. This creates a risk that the synthetic training data are enriched with examples that the target model already transcribes well, inflating the apparent benefit of synthetic fine-tuning. The authors explicitly acknowledge in Section 9 that 'using multiple models during data construction and fine-tuning can introduce some bias and error accumulation.' Please assess this bias concretely, for example by using a held-out filter model from a different ASR family and reporting how the filtered dataset and downstream CER change when the filter and fine-tuning model are decoupled.

minor comments (5)

[§1 and Table 3] The headline '63% highest performance improvement, average improvement 51.5%' does not state which models or rows are used for the average; please specify that these are CER improvements for the best fine-tuned settings per model and add a table reference.
[Table 3] Several entries are inconsistently formatted (e.g., WER_cn values reported as '0.58' and '0.70' with two decimals while others have five) and the dataset name 'GUCCCI&LV' contains a typo; please standardize formatting and correct the label.
[§8.1 and Figure 5] The claim that whisper-large-v2 'has achieved almost zero error translation performance' after fine-tuning is not quantified; please provide example-level error counts or a table snippet with CER/WER for the shown cases.
[§5.2] The IER formula would benefit from an explicit equation number and a precise statement of how de-duplication across WER and CER is applied algorithmically, since the current description ('we count it only once') is ambiguous when a word and one of its characters are both wrong.
[§6.1] The prompt-generation procedure samples 's sentences' and 'i keywords' (s=5, i=8), but the paper does not describe how repeated sampling is avoided across the 10,000 generated examples or how diversity was verified; please add these details.

Circularity Check

1 steps flagged · score 6.0 of 10

Reported 63%/51.5% improvements are partly test-set-fitted: Section 8.2 selects the fine-tuning epoch count using the same Test Set that produces the final CER values in Table 3.

fitted input called prediction [Section 6.1 (Test Set independence claim) vs Section 8.2 (epoch selection) and Table 3 / Abstract (reported improvements)]
"In the experimental part of this paper, the Test Set is mainly used for testing the fine-tuned ASR models and does not participate in the construction of any fine-tuning datasets. ... compared with the native model, the highest performance improvement is 63%, and the average improvement is 51.5%. ... Sub-figure 6a of Figure 6 shows the results of further fine-tuning the three best-performing models while keeping other parameters constant and calculating their CER metrics. ... Therefore, we chose 5 epochs as one of the parameters for fine-tuning the ASR model."

The headline improvement figures are computed from Table 3, whose CER values are measured on the Test Set (Section 6.4 states that result testing and metric calculation are performed on the test set). Section 8.2 then uses the same Test Set to select the fine-tuning epoch count (5) after comparing CER metrics at 10, 15, 20, and 30 epochs. The Test Set therefore influences which model configuration is reported, so the final numbers are test-set-selected estimates rather than independent out-of-sample predictions. This is the fitted-input-called-prediction pattern: the evaluation data were used to choose the configuration whose CER is then reported as the method's improvement.

full rationale

The one central circular step is the reuse of the Test Set for hyperparameter selection and final reporting. Sections 6.1 and 6.4 describe the Test Set as the evaluation set for metric calculation, and Section 8.2 chooses 5 fine-tuning epochs by comparing CER metrics after further fine-tuning for 10, 15, 20, and 30 epochs. Since Table 3 and the abstract's 63%/51.5% improvement claims are based on CER values from that same Test Set, the headline quantitative claims are optimistically biased by construction; the Test Set did participate in model configuration even though it did not participate in training-data construction. A secondary concern is that Section 5.1 filters synthetic data with whisper-large-v2, and whisper-large-v2 is also the best-reported model; this may favor that model, but it is not a definitional circularity because the filter is a fixed pre-trained model and the fine-tuning objective is not defined in terms of the filter's outputs. No load-bearing self-citation was found: reference [7], which shares an author with this paper, is a standard LoRA survey used to justify a standard technique, and the paper's central method does not depend on an unverified uniqueness claim. Section 9's own admission that using multiple models 'can introduce some bias and error accumulation' is consistent with the filter-model concern but does not by itself make the derivation circular. Overall, the method has independent content, but the headline 'predictions' are partially fitted to the evaluation set, so the circularity score is 6 rather than 0-2.

Assumptions & free parameters 4 free parameters · 3 assumptions · 1 invented entities

The central result depends on several fitted design choices: the filtering threshold, number of epochs, and sampling counts. It also assumes that TTS audio adequately represents real conditions, that the keyword list is comprehensive, and that using Whisper as the quality filter is unbiased. The only invented entity is the IER metric, which lacks independent validation.

free parameters (4)

CER filtering threshold tau = 0.15
Chosen in Section 8.2 'by weighing the amount and quality of data' after trying different CER boundaries; not derived or cross-validated.
Fine-tuning epochs = 5
Selected in Section 8.2 by comparing CER after 10, 15, 20, and 30 epochs on the test set; no separate validation set.
Sample count s and keyword count i = 5 and 8
Set in Section 6.3 without sensitivity analysis; affects diversity and size of the synthetic dataset.
IER test set size = 200
Section 6.4 states 'We reduced the amount of data in the test set to 200' for IER calculations, reducing statistical power.

assumptions (3)

domain assumption Synthetic audio from ChatTTS is representative enough of real-world accents and noise to improve ASR on real speech.
Section 5.1 claims the TTS simulates dialects and accents, but Section 9 admits the TTS may produce partially unusable audio and that 'false-labeled speech generated by ChatTTS deviates from the pronunciation of other languages' (Section 7).
domain assumption The keyword list K covers most industry-relevant professional terms.
Section 6.1 describes K as seven categories harvested from social media and manual cleaning, but no coverage evaluation is provided. If K is incomplete, the synthetic labels miss domain vocabulary.
domain assumption Whisper's CER on synthetic data is a valid proxy for data quality in the filtering step.
Section 5.1 uses whisper-large-v2 to compare inferred tags with generated tags and excludes samples above a CER threshold. This assumes the base model's errors are a reliable measure of label quality, which is not independently established.

invented entities (1)

Integrated Error Rate (IER) metric
purpose: Evaluate ASR performance on mixed Chinese-English speech with brand-specific keywords, combining word, character, and keyword errors.
Introduced in Section 5.2 with a hand-defined formula and manually verified results. No external benchmark, public implementation, or independent validation is provided, so its reliability as a measurement tool is not established.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Weak Supervision Techniques towards Enhanced ASR Models in Industry-level CRM Systems." pith.science (2026). https://pith.science/paper/JLA5VXYH

@misc{pith2026250716843,
  author       = {Pith},
  title        = {Pith review of: Weak Supervision Techniques towards Enhanced ASR Models in Industry-level CRM Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JLA5VXYH}},
  note         = {Machine review of arXiv:2507.16843}
}

read the original abstract

In the design of customer relationship management (CRM) systems, accurately identifying customer types and offering personalized services are key to enhancing customer satisfaction and loyalty. However, this process faces the challenge of discerning customer voices and intentions, and general pre-trained automatic speech recognition (ASR) models make it difficult to effectively address industry-specific speech recognition tasks. To address this issue, we innovatively proposed a solution for fine-tuning industry-specific ASR models, which significantly improved the performance of the fine-tuned ASR models in industry applications. Experimental results show that our method substantially improves the crucial auxiliary role of the ASR model in industry CRM systems, and this approach has also been adopted in actual industrial applications.

Figures

Figures reproduced from arXiv: 2507.16843 by the authors.

**Figure 2.** The detailed main framework of the ASR fine-tuning method, which com [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

**Figure 3.** Original true label content of a fictitious label and the version inferred [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: The true labels and English translations. These examples are fictitious, [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

**Figure 5.** Figure 5: The original labels and inference results by different models in the test [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]

**Figure 6.** Figure 6: Results of additional experiments. In addition, a CER threshold is determined to screen the dataset. In subfigure 6b, we show the remaining data in the two synthetic datasets under different CER boundaries. By weighing the amount and quality of data, we choose 0.15 a…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

22 extracted references · 17 canonical work pages

[1]

IOP Publishing (2019)

Adlin, F.N., Ferdiana, R., Fauziati, S.: Current trend and literature on electronic crmadoptionreview.In:JournalofPhysics:ConferenceSeries.vol.1201,p.012058. IOP Publishing (2019)

work page 2019
[2]

Ao, J., Wang, R., Zhou, L., Wang, C., Ren, S., Wu, Y., Liu, S., Ko, T., Li, Q., Zhang, Y., Wei, Z., Qian, Y., Li, J., Wei, F.: Speecht5: Unified-modal encoder- decoder pre-training for spoken language processing (2022)

work page 2022
[3]

International Journal of Emerging Technology and Advanced Engineering1(2) (2011)

Atayero, A.A., Alatishe, A.S., Iruemi, J.O.: Development of ispeak: A voice- activated relationship management system. International Journal of Emerging Technology and Advanced Engineering1(2) (2011)

work page 2011
[4]

Baevski, A., Zhou, H., Mohamed, A., Auli, M.: wav2vec 2.0: A framework for self-supervised learning of speech representations (2020)

work page 2020
[5]

Bartelds, M., San, N., McDonnell, B., Jurafsky, D., Wieling, M.: Making more of little data: Improving low-resource automatic speech recognition using data augmentation (2023)

work page 2023
[6]

DeepSeek-AI: Deepseek-v2: A strong, economical, and efficient mixture-of-experts language model (2024)

2024
[7]

arXiv preprint arXiv:2403.14608 (2024)

Han, Z., Gao, C., Liu, J., Zhang, S.Q., et al.: Parameter-efficient fine-tuning for large models: A comprehensive survey. arXiv preprint arXiv:2403.14608 (2024)

arXiv 2024
[8]

In: Proc

He, H., Choi, J.D.: The stem cell hypothesis: Dilemma behind multi-task learning with transformer encoders. In: Proc. EMNLP. pp. 5555–5577. Online and Punta Cana, Dominican Republic (Nov 2021)

work page 2021

Show all 22 references

[9]

E Fendji, Diane C

Jean Louis K. E Fendji, Diane C. M. Tala, B.O.Y., Atemkeng, M.: Automatic speech recognition using limited vocabulary: A survey. Applied Artificial Intelli- gence36(1), 2095039 (2022)

2022
[10]

IJMELR (2023)

Kaliuta, K.: Implementing voice recognition and natural language processing in salesforce. IJMELR (2023)

2023
[11]

Uni- versity of Stirling (2016)

Mustapha, M.: Implementing speech-to-text technologies for on-the-go crm. Uni- versity of Stirling (2016)

2016
[12]

arXiv preprint arXiv:2404.16743 (2024) Title Suppressed Due to Excessive Length 15

Park, C., Chen, M., Hain, T.: Automatic speech recognition system-independent word error rate estimatio. arXiv preprint arXiv:2404.16743 (2024) Title Suppressed Due to Excessive Length 15

2024 arXiv
[13]

In: Interspeech

Park, D.S., Zhang, Y., Jia, Y., Han, W., Chiu, C.C., Li, B., Wu, Y., Le, Q.V.: Improved noisy student training for automatic speech recognition. In: Interspeech
[14]

Radford, A., Kim, J.W., Xu, T., Brockman, G., McLeavey, C., Sutskever, I.: Robust speech recognition via large-scale weak supervision (2022), https://arxiv.org/abs/2212.04356

2022 arXiv
[15]

In: International conference on machine learning

Radford,A.,Kim,J.W.,Xu,T.,Brockman,G.,McLeavey,C.,Sutskever,I.:Robust speech recognition via large-scale weak supervision. In: International conference on machine learning. pp. 28492–28518. PMLR (2023)

2023
[16]

Ren, Y., Tan, X., Qin, T., Zhao, S., Zhao, Z., Liu, T.Y.: Almost unsupervised text to speech and automatic speech recognition (2020)

2020
[17]

arXiv preprint arXiv:1904.05862 (2019)

Schneider, S., Baevski, A., Collobert, R., Auli, M.: wav2vec: Unsupervised pre- training for speech recognition. arXiv preprint arXiv:1904.05862 (2019)

2019 arXiv
[18]

Singh, S., Hou, F., Wang, R.: A novel self-training approach for low-resource speech recognition (2023)

2023
[19]

In: Proc.ICDAR

Wigington, C., Stewart, S., Davis, B., Barrett, B., Price, B., Cohen, S.: Data augmentation for recognition of handwritten words and lines using a cnn-lstm network. In: Proc.ICDAR. vol. 1, pp. 639–645. IEEE (2017)

2017
[20]

In: SIGKDD

Xu, J., Tan, X., Ren, Y., Qin, T., Li, J., Zhao, S., Liu, T.Y.: Lrspeech: Extremely low-resource speech synthesis and recognition. In: SIGKDD. p. 2802–2812. KDD ’20, Association for Computing Machinery, New York, NY, USA (2020)

2020
[21]

E3S Web Conf

Yunlong, Z., Xiangyu, L., Hongyan, X., Yingzhe, H., Jialiang, Q.: Design of intelligent customer service report system based on auto- matic speech recognition and text classification. E3S Web Conf. (2021). https://doi.org/10.1051/e3sconf/202129501064

2021
[2020]

interspeech_2020 (2020)

2020

Pith tools

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

[1] [1]

IOP Publishing (2019)

Adlin, F.N., Ferdiana, R., Fauziati, S.: Current trend and literature on electronic crmadoptionreview.In:JournalofPhysics:ConferenceSeries.vol.1201,p.012058. IOP Publishing (2019)

work page 2019

[2] [2]

Ao, J., Wang, R., Zhou, L., Wang, C., Ren, S., Wu, Y., Liu, S., Ko, T., Li, Q., Zhang, Y., Wei, Z., Qian, Y., Li, J., Wei, F.: Speecht5: Unified-modal encoder- decoder pre-training for spoken language processing (2022)

work page 2022

[3] [3]

International Journal of Emerging Technology and Advanced Engineering1(2) (2011)

Atayero, A.A., Alatishe, A.S., Iruemi, J.O.: Development of ispeak: A voice- activated relationship management system. International Journal of Emerging Technology and Advanced Engineering1(2) (2011)

work page 2011

[4] [4]

Baevski, A., Zhou, H., Mohamed, A., Auli, M.: wav2vec 2.0: A framework for self-supervised learning of speech representations (2020)

work page 2020

[5] [5]

Bartelds, M., San, N., McDonnell, B., Jurafsky, D., Wieling, M.: Making more of little data: Improving low-resource automatic speech recognition using data augmentation (2023)

work page 2023

[6] [6]

DeepSeek-AI: Deepseek-v2: A strong, economical, and efficient mixture-of-experts language model (2024)

2024

[7] [7]

arXiv preprint arXiv:2403.14608 (2024)

Han, Z., Gao, C., Liu, J., Zhang, S.Q., et al.: Parameter-efficient fine-tuning for large models: A comprehensive survey. arXiv preprint arXiv:2403.14608 (2024)

arXiv 2024

[8] [8]

In: Proc

He, H., Choi, J.D.: The stem cell hypothesis: Dilemma behind multi-task learning with transformer encoders. In: Proc. EMNLP. pp. 5555–5577. Online and Punta Cana, Dominican Republic (Nov 2021)

work page 2021