WER

Word Error Rate Estimation for Speech Recognition: e-WER (P18-2)

Challenge:	Accents play a pivotal role in shaping human communication, a new study finds . existing ASR systems often perform inadequately, even mispronouncing African names .
Approach:	They propose a method that uses epistemic uncertainty to automate annotation to reduce costs and human labor.
Outcome:	The proposed method reduces costs and human labor by reducing data annotation and epistemic uncertainty.

Copied to clipboard

Ahmed Ali, Steve Renals

PolyNorm: Few-Shot LLM-Based Text Normalization for Text-to-Speech (2025.emnlp-industry)

Challenge:	Automatic speech recognition (ASR) systems require manual transcription of test data to compute the word error rate (WER).
Approach:	They propose an approach to estimate word error rate (e-WER) that does not require a gold-standard transcription of the test set.
Outcome:	The proposed approach achieves 16.9% WER root mean squared error across 1,400 sentences.

Copied to clipboard

Michel Wong, Ali Alshehri, Sophia Kao, Haotian He

JoeyS2T: Minimalistic Speech-to-Text Modeling with JoeyNMT (2022.emnlp-demos)

Challenge:	Text Normalization (TN) is a key preprocessing step in Text-to-Speech systems.
Approach:	They propose a prompt-based approach to TN using Large Language Models (LLMs) they propose scalable experimentation across languages to reduce the reliance on manual rules .
Outcome:	The proposed approach reduces the reliance on manual rules and enables broader linguistic applicability with minimal human intervention across eight languages.

Copied to clipboard

Mayumi Ohta, Julia Kreutzer, Stefan Riezler

Investigating data partitioning strategies for crosslinguistic low-resource ASR evaluation (2023.eacl-main)

Challenge:	JoeyS2T is a simple, lightweight NMT extension for speech-to-text tasks such as automatic speech recognition and end-to end speech translation.
Approach:	JoeyS2T is a JoeyNMT extension for automatic speech recognition and end-to-end speech translation.
Outcome:	JoeyS2T performs competitively on English speech recognition and English-to-German speech translation benchmarks.

Copied to clipboard

Zoey Liu, Justin Spence, Emily Prud’hommeaux

Making a Point: Pointer-Generator Transformers for Disjoint Vocabularies (2020.aacl-srw)

Challenge:	Automatic speech recognition data sets include a single pre-defined test set consisting of one or more speakers whose speech never appears in the training set.
Approach:	They propose to use hold-speaker(s)-out partitioning to partition data for five languages . utterance duration and intensity are more predictive factors of variability .
Outcome:	The proposed method can produce results that do not reflect model performance on unseen data or speakers.

Copied to clipboard

Nikhil Prabhu, Katharina Kann

hinglishNorm - A Corpus of Hindi-English Code Mixed Sentences for Text Normalization (2020.coling-industry)

Challenge:	Existing neural models rely on an overlap between source and target vocabularies to perform sequence-to-sequence tasks.
Approach:	They propose a pointer-generator transformer model for disjoint vocabularies that does not rely on an overlap between source and target vocs.
Outcome:	The proposed model outperforms a standard pointer-generator transformer by an average of 5.1 WER over 15 languages.

Copied to clipboard

Piyush Makhija, Ankit Kumar, Anuj Gupta

Syntactic Text Processing

Challenge:	hinglishNorm is a human annotated corpus of Hindi-English code-mixed sentences for text normalization task.
Approach:	They propose to annotate sentences in Hindi-English code-mixed sentences using a human annotated normalized form.
Outcome:	The proposed corpus contains 13494 segments annotated for text normalization.

A Corpus for Modeling Word Importance in Spoken Dialogue Transcripts (L18-1)

Copied to clipboard

Sushant Kafle, Matt Huenerfauth

NeKo: Cross-Modality Post-Recognition Error Correction with Tasks-Guided Mixture-of-Experts Language Model (2025.acl-industry)

Challenge:	a project aims to create a system that uses automatic speech recognition (ASR) to produce real-time text captions of spoken English during in-person meetings with hearing individuals.
Approach:	They propose to use automatic speech recognition to produce captions in real-time . they add word-importance annotations to a transcript of a conversational dialogue corpus .
Outcome:	The proposed system would produce captions in real-time for people who are deaf or hard-of-hearing . the best performing model has an F-score of 0.60 in an ordinal 6-class word-importance classification task with an agreement (concordance correlation coefficient) of 0.89 with the human annotators.

Copied to clipboard

Yen-Ting Lin, Zhehuai Chen, Piotr Zelasko, Zhen Wan, Xuesong Yang, Zih-Ching Chen, Krishna C Puvvada, Ke Hu, Szu-Wei Fu, Jun Wei Chiu, Jagadeesh Balam, Boris Ginsburg, Yu-Chiang Frank Wang, Chao-Han Huck Yang

Building Accurate Low Latency ASR for Streaming Voice Search in E-commerce (2023.acl-industry)

Challenge:	Existing methods to train a model on a mixture of domain datasets require separate correction language models.
Approach:	They propose a multi-task correction MoE that trains experts to become an "expert" of speech-to-text, language-totext and vision-to text datasets by learning to route each dataset’s tokens to its mapped expert.
Outcome:	The proposed model outperforms GPT-3.5 and Claude-3.5-Sonnet on the Open ASR Leaderboard and reaches an average relative 5.0% WER reduction and substantial improvements in BLEU scores.

Copied to clipboard

Abhinav Goyal, Nikesh Garera

Exploring the Effect of Dialect Mismatched Language Models in Telugu Automatic Speech Recognition (2022.naacl-srw)

Challenge:	Recent years have witnessed the popularity of end-to-end ASR models, which have demonstrated higher accuracy compared to traditional pipelines with separate acoustic, pronunciation, and language models.
Approach:	They build accurate LSTM, attention and CTC based streaming ASR models for large-scale Hinglish voice search.
Outcome:	The proposed model achieves a word error rate (WER) of 3.69% without EOS and 4.78% with EOS, with 1300 ms (46.64%) reduction in latency.

Copied to clipboard

Aditya Yadavalli, Ganesh Sai Mirishkar, Anil Vuppala

Synthetic Doctor-Patient Dialogue Generation for Robust Medical ASR: A Scalable Pipeline for Vocabulary Expansion and Privacy Preservation (2026.eacl-industry)

Challenge:	Existing studies have found that the ASR system is susceptible to dialect variations within a language, thereby adversely affecting the APR.
Approach:	They propose to build a dialect-specific AM while keeping the Language Model constant for all the dialects and to reduce the degradation by 9% and 15%.
Outcome:	The proposed model can be built for three different Telugu regional dialects while keeping the Language Model constant for all the dialects.

Copied to clipboard

Kefei Liu, Meizhu Liu

Transformer versus LSTM Language Models trained on Uncertain ASR Hypotheses in Limited Data Scenarios (2022.lrec-1)

Challenge:	Existing ASR models struggle with high word error rates (WER) on clinical vocabulary, especially medication names.
Approach:	They propose to generate doctor-patient dialogues in both text and audio formats using a curated set of over 124,000 medical terms.
Outcome:	The proposed pipeline generated over 1 billion audios with ground truth transcriptions.

Copied to clipboard

Imran Sheikh, Emmanuel Vincent, Irina Illina

That doesn’t sound right: Evaluating speech transcription quality in field linguistics corpora (2025.acl-short)

Challenge:	Existing studies show that domain-specific LMs can only rely on limited in-domain speech data . a qualitative analysis reveals that Transformer LM can predict less frequent words .
Approach:	They propose a method to train Transformer LMs on ASR confusion networks . they find they are better at exploiting alternate uncertain ASR hypotheses .
Outcome:	The proposed method reduces perplexity by 3-6% on AMI scenarios but performs similar to LSTM LMs on Verbmobil conversational corpus.

Copied to clipboard

Eric Le Ferrand, Bo Jiang, Joshua Hartshorne, Emily Prud’hommeaux

Efficient Sample-Specific Encoder Perturbations (2024.naacl-short)

Challenge:	Automated speech recognition (ASR) is a popular tool for documenting languages, but field linguists do not have the data to train robust models.
Approach:	They propose to use fieldwork data to identify speech transcriptions that may be unsuitable for training ASR models.
Outcome:	The proposed measures can be used to identify transcriptions with characteristics common in field data but could be detrimental to ASR training.

Copied to clipboard

Yassir Fathullah, Mark Gales

Pseudo2Real: Task Arithmetic for Pseudo-Label Correction in Automatic Speech Recognition (2026.findings-acl)

Challenge:	Existing approaches to training encoder-decoder systems often depend on teacher-forcing with the likelihood criteria, e.g. next token prediction of the reference sequence.
Approach:	They propose an inference-efficient way to modify the behaviour of an encoder-decoder system according to a specific attribute of interest by using a small proxy network.
Outcome:	The proposed framework improves the COMET performance of Flan-T5 on Machine Translation and the WER of Whisper foundation models on Speech Recognition.

Copied to clipboard

Yi-Cheng Lin, Yu-Hsuan Li Liang, Hsuan Su, Tzu-Quan Lin, Shang-Tse Chen, Yun-Nung Chen, Hung-yi Lee

When Good OCR Is Not Enough: Benchmarking OCR Robustness for Retrieval-Augmented Generation (2026.acl-industry)

Challenge:	ASR models can be used to correct accent-specific errors without ground truth . pseudo-labels inherit the teacher model's systematic biases, authors say .
Approach:	They propose a parameter-space correction technique that captures pseudo-label biases . they propose achieving up to 35% relative WER reduction on a pseudo-labeled target model .
Outcome:	The proposed model achieves 35% relative WER reduction on ten African accents with the Whisper tiny model.

Copied to clipboard

Lin Sun, null Wangdexian, Jingang Huang, Linglin Zhang, Change Jia, Zhengwei Cheng, Xiangzheng Zhang

Responsible & Trustworthy NLP

Challenge:	Existing OCR benchmarks rely on character-level metrics to measure downstream performance . high OCR accuracy does not translate into strong downstream performance, authors say .
Approach:	They propose an OCR benchmark for industrial RAG systems that measures character-level metrics . they find that high OCR accuracy does not translate into strong downstream RAG performance .
Outcome:	The proposed benchmark shows that high OCR accuracy does not translate into strong downstream performance . structural and semantic errors can cause substantial retrieval failures even when WER/CER remains low.

Retrieve and Copy: Scaling ASR Personalization to Large Catalogs (2023.emnlp-industry)

Copied to clipboard

Sai Muralidhar Jayanthi, Devang Kulshreshtha, Saket Dingliwal, Srikanth Ronanki, Sravan Bodapati

Information Retrieval

Challenge:	End-to-end ASR models struggle to recognize uncommon domain-specific words due to limited audio context.
Approach:	They propose a "Retrieve and Copy" mechanism to improve latency while retaining the accuracy even when scaled to a large catalog.
Outcome:	The proposed method achieves 6% more word error rate reduction and 3.6% improvement in F1 when scaled to a large catalog size while retaining the accuracy.

An Application for Building a Polish Telephone Speech Corpus (L18-1)

Copied to clipboard

Bartosz Ziółko, Piotr Żelasko, Ireneusz Gawlik, Tomasz Pędzimąż, Tomasz Jadczyk

The Norwegian Parliamentary Speech Corpus (2022.lrec-1)

Challenge:	Specifically, we describe a tool designed to improve our Automatic Speech Recognition system performance.
Approach:	They propose to build a tool for speech corpus collection of a specific domain content.
Outcome:	The proposed tool can be used to gather 63 hours of speech recordings across several domains and achieve lower WER in two grammar-based speech recognition tasks.

Copied to clipboard

Per Erik Solberg, Pablo Ortiz

Towards an Open-Source Dutch Speech Recognition System for the Healthcare Domain (2022.lrec-1)

Challenge:	the dataset contains recordings of meetings at the Norwegian parliament . it is the first publicly available dataset containing unscripted, Norwegian speech .
Approach:	the Norwegian Parliamentary Speech Corpus is a publicly available speech dataset . it contains recordings of meetings from the Norwegian parliament with orthographic transcriptions . the dataset is intended to fill a gap in the available unscripted speech data .
Outcome:	the dataset contains recordings of meetings at the Norwegian parliament with orthographic transcriptions in Norwegian Bokml and Norwegian Nynorsk.

Copied to clipboard

Cristian Tejedor-García, Berrie van der Molen, Henk van den Heuvel, Arjan van Hessen, Toine Pieters

SEAM: Bridging the Temporal-Semantic Granularity Gap for LLM-based Speech Recognition (2026.findings-eacl)

Challenge:	Existing generic speech recognition systems do not include healthcare jargon in the lexicon and do not safeguard privacy of sensitive data.
Approach:	They propose to use a language model to train Dutch doctors to use medicines in their audiovisual recordings.
Outcome:	The proposed method reduces the word error rate (WER) by 5.2% on the use of medicines in the Netherlands.

Copied to clipboard

Junseok Oh, Ji-Hwan Kim

Role-specific Language Models for Processing Recorded Neuropsychological Exams (N18-2)

Challenge:	Existing duration-based methods generate embeddings at fixed rates, creating distributional mismatch with LLM pre-training.
Approach:	They propose an encoder-decoder architecture that generates embeddings at variable rates through cross-attention between speech features and text embeddables.
Outcome:	The proposed architecture achieves competitive performance on LibriSpeech (2.6%/5.2% WER) and 4.7% WER on TED-LIUM-v2 with a multi-stage training strategy and First Token Guidance.

Copied to clipboard

Tuka Al Hanai, Rhoda Au, James Glass

PortOldBERT: Portuguese Historical Language Models (2026.eacl-long)

Challenge:	Neuropsychological examinations are an important screening tool for the presence of cognitive conditions such as Alzheimer's, Parkinson's and spinal-cord injuries.
Approach:	They propose to use audio recordings to determine the cognitive health of 92 subjects from audio that was diarized using an automatic speech recognition system trained on TED talks and on structured language used by testers and subjects.
Outcome:	The proposed method can determine the cognitive health of 92 subjects from audio that was diarized using an automatic speech recognition system trained on TED talks and on the structured language used by testers and subjects.

Copied to clipboard

Tomas Freitas Osorio, Henrique Lopes Cardoso

Can Visual Context Improve Automatic Speech Recognition for an Embodied Agent? (2022.emnlp-main)

Challenge:	Historical language models are important for many tasks, but have limited application in historical Portuguese.
Approach:	They propose a Portuguese encoder language model that integrates historical linguistic contexts while retaining the ability to process contemporary text.
Outcome:	The proposed model outperforms Albertina on multiple tasks and for different historical periods.

Copied to clipboard

Pradip Pramanick, Chayan Sarkar

SAC: Neural Speech Codec with Semantic-Acoustic Dual-Stream Quantization (2026.acl-long)

Challenge:	ASR systems are often unable to recognize speech due to generic datasets and open-vocabulary modeling.
Approach:	They propose to incorporate a robot’s visual information into an ASR system and improve the recognition of a spoken utterance containing a visible entity.
Outcome:	The proposed method achieves a 59% relative reduction in WER from an unmodified ASR system.

Copied to clipboard

Wenxi Chen, Ruiqi Yan, Yushen Chen, Zhikang Niu, Ziyang Ma, Xiquan Li, Yuzhe Liang, null Wenhanlin, Shunshun Yin, Ming Tao, Xinsheng Wang, Xie Chen

AccentFold: A Journey through African Accents for Zero-Shot ASR Adaptation to Target Accents (2024.findings-eacl)

Challenge:	Existing speech codecs struggle to balance high-quality reconstruction with semantically rich representations, limiting their effectiveness in both generative and understanding tasks.
Approach:	They propose a neural speech codec with semantic-acoustic dual-stream quantization that disentangles semantic and acousian modeling into two dedicated streams.
Outcome:	The proposed codec outperforms state-of-the-art speech tokenizers in auto-propagating text-to-speech models.

Copied to clipboard

Abraham Owodunni, Aditya Yadavalli, Chris Emezue, Tobi Olatunji, Clinton Mbataku

AssameseBackTranslit: Back Transliteration of Romanized Assamese Social Media Text (2024.lrec-main)

Challenge:	AccentFold uses spatial relationships to improve speech recognition for accented speech . existing methods for accent recognition have been limited due to data scarcity and budget constraints .
Approach:	They propose a method that exploits spatial relationships between learned accent embeddings to improve downstream automatic speech recognition.
Outcome:	The proposed method outperforms baseline methods in accented speech training.

Copied to clipboard

Hemanta Baruah, Sanasam Ranbir Singh, Priyankoo Sarmah

AudioSAE: Towards Understanding of Audio-Processing Models with Sparse AutoEncoders (2026.eacl-long)

Challenge:	a novel dataset capturing native text composed in the Roman/Latin script is presented . the dataset comprises 60,312 Roman-native parallel transliterated sentences .
Approach:	They propose a back transliteration dataset capturing native text composed in the Roman/Latin script and its corresponding representation in the native Assamese script.
Outcome:	The proposed dataset outperforms baseline models in terms of word-level transliteration evaluation benchmarks and performance assessments.

Copied to clipboard

Georgii Aparin, Tasnima Sadekova, Alexey Rukhovich, Assel Yermekova, Laida Kushnareva, Vadim Popov, Kristian Kuznetsov, Irina Piontkovskaya

Large Margin Neural Language Model (D18-1)

Challenge:	Feature steering reduces Whisper’s false speech detections by 70% with negligible WER increase, demonstrating real-world applicability.
Approach:	They train Sparse Autoencoders across all encoder layers of Whisper and HuBERT and evaluate their stability, interpretability, and practical utility.
Outcome:	The proposed models capture general acoustic and semantic information as well as specific events, including environmental noises and paralinguistic sounds, and disentangle them effectively.

Copied to clipboard

Jiaji Huang, Yi Li, Wei Ping, Liang Huang

Align-Refine: Non-Autoregressive Speech Recognition via Iterative Realignment (2021.naacl-main)

Challenge:	Conventionally, neural language models are trained by minimizing perplexity (PPL) on grammatical sentences.
Approach:	They propose a large margin criterion for training neural language models by minimizing perplexity on grammatical sentences and propose enlarged margins for task-specific training.
Outcome:	The proposed method gains up to 1.1 WER reduction for speech recognition and 1.0 BLEU increase for machine translation.

Copied to clipboard

Ethan A. Chi, Julian Salazar, Katrin Kirchhoff

Improving End-to-End Bangla Speech Recognition with Semi-supervised Training (2020.findings-emnlp)

Challenge:	Non-autoregressive encoder-decoder models improve decoding speed, but generation quality suffers . editing at the level of output sequences limits model flexibility.
Approach:	They propose iterative realignment which iteratively realigns connectionist temporal alignments.
Outcome:	The proposed model matches an autoregressive baseline with a 14x speedup on the WSJ dataset; on LibriSpeech, it achieves an LM-free test-other WER of 9.0% (19% relative improvement on comparable work).

Copied to clipboard

Nafis Sadeq, Nafis Tahmid Chowdhury, Farhan Tanvir Utshaw, Shafayat Ahmed, Muhammad Abdullah Adnan

Automatic Speech Recognition System-Independent Word Error Rate Estimation (2024.lrec-main)

Challenge:	Existing methods to train speech recognition systems require large annotated corpus.
Approach:	They propose a semi-supervised training approach that exploits large unpaired audio and text data to improve the performance of an automatic speech recognition system.
Outcome:	The proposed method reduces the WER of the system from 37% to 31.9%.

Copied to clipboard

Chanho Park, Mingjie Chen, Thomas Hain

MetaMixSpeech: Meta Task Augmentation for Low-Resource Speech Recognition (2025.findings-emnlp)

Challenge:	Word error rate (WER) is a metric used to evaluate the quality of transcriptions produced by Automatic Speech Recognition systems.
Approach:	They propose a hypothesis generation method for ASR system-dependent WER estimation . they use phonetically similar or linguistically more likely alternative words to generate hypotheses .
Outcome:	The proposed method outperforms baseline estimators on in-domain data and out-of-domain on Switchboard and CALLHOME.

Copied to clipboard

Yaqi Chen, Hao Zhang, Wenlin Zhang, XuKui Yang, Dan Qu, Yunpeng Liu

RECAST: Retrieval-Augmented Contextual ASR via Decoder-State Keyword Spotting (2025.findings-emnlp)

Challenge:	Meta-learning has proven to be a powerful paradigm for improving speech recognition performance . however, multilingual meta learning also faces challenges such as task overfitting and learner overfit .
Approach:	a new method is proposed to augment meta-training tasks with "more data" the method incorporates both support and query augmentations .
Outcome:	The proposed method achieves a 6.35% improvement in the word error rate on FLEURS and Common Voice datasets.

Copied to clipboard

Ashish Mittal, Sunita Sarawagi, Preethi Jyothi

Multimodal and Multiresolution Speech Recognition with Transformers (2020.acl-main)

Challenge:	RECAST is a lightweight retrieval-augmented approach for contextual ASR . it repurposes decoder states of a pretrained ASR model to retrieve relevant keywords .
Approach:	RECAST is a retrieval-augmented approach that repurposes decoder states of a pretrained ASR model to retrieve relevant keywords without requiring audio exemplars.
Outcome:	RECAST outperforms full-list prompt biasing and strong phonetic/text baselines on 4,000 keywords across diverse domains.

Copied to clipboard

Georgios Paraskevopoulos, Srinivas Parthasarathy, Aparna Khare, Shiva Sundaram

DyPCL: Dynamic Phoneme-level Contrastive Learning for Dysarthric Speech Recognition (2025.naacl-long)

Challenge:	Existing audio visual automatic speech recognition systems rely on audio input to produce transcriptions.
Approach:	They propose an audio visual automatic speech recognition system using a transformer-based architecture and incorporate a multitask training criterion for multiresolution ASR.
Outcome:	The proposed system can generate character and subword transcriptions with visual information.

Copied to clipboard

Wonjun Lee, Solee Im, Heejin Do, Yunsu Kim, Jungseul Ok, Gary Lee

Masked Language Model Scoring (2020.acl-main)

Challenge:	Existing studies have focused on data augmentation and feature extraction methods to improve dysarthric speech recognition.
Approach:	They propose a Dynamic Phoneme-level Contrastive Learning method which decomposes the speech utterance into phoneme segments for phoneme- level contrastive learning.
Outcome:	The proposed method outperforms baseline models and achieves an average 22.10% reduction in word error rate (WER) across the overall dysarthria group.

Copied to clipboard

Julian Salazar, Davis Liang, Toan Q. Nguyen, Katrin Kirchhoff

TrInk: Ink Generation with Transformer Network (2025.emnlp-main)

Challenge:	Pretrained masked language models require finetuning for most tasks.
Approach:	They evaluate pretrained masked language models out of the box via their pseudo-log-likelihood scores (PLLs) they attribute this success to PLL’s unsupervised expression of linguistic acceptability without a left-to-right bias, greatly improving on scores from GPT-2 .
Outcome:	The proposed model outperforms autoregressive language models in a variety of tasks.

Copied to clipboard

Zezhong Jin, Shubhang Desai, Xu Chen, Biyi Fang, Zhuoyi Huang, Zhe Li, Chong-Xin Gan, Xiao Tu, Man-Wai Mak, Yan Lu, Shujie Liu

Highly Effective Arabic Diacritization using Sequence to Sequence Modeling (N19-1)

Challenge:	Existing methods for handwriting generation capture global dependencies and can generate high-quality handwritten samples.
Approach:	They propose a Transformer-based model for ink generation, TrInk, which captures global dependencies.
Outcome:	The proposed model reduces character error rate and word error rate by 35.56% on the IAM-OnDB dataset compared to previous models.

Copied to clipboard

Hamdy Mubarak, Ahmed Abdelali, Hassan Sajjad, Younes Samih, Kareem Darwish

Lost in Transcription: Identifying and Quantifying the Accuracy Biases of Automatic Speech Recognition Systems Against Disfluent Speech (2024.naacl-long)

Challenge:	Arabic text is written without short vowels (or diacritics) their presence is essential for properly verbalizing Arabic .
Approach:	They propose a character-level sequence-to-sequence deep learning model that recovers both types of diacritics without the use of explicit feature engineering.
Outcome:	The proposed model outperforms all previous state-of-the-art models on overlapping windows of words . it achieves a word error rate (WER) of 4.49% compared to the state- of-the art systems .

Copied to clipboard

Dena Mujtaba, Nihar Mahapatra, Megan Arney, J Yaruss, Hope Gerlach-Houck, Caryn Herring, Jia Bin

Advocating Character Error Rate for Multilingual ASR Evaluation (2025.findings-naacl)

Challenge:	Automatic speech recognition systems fail to accurately interpret speech patterns deviating from typical fluency, leading to critical usability issues and misinterpretations.
Approach:	They evaluate six leading automatic speech recognition systems based on a real-world dataset and a synthetic dataset derived from the widely-used LibriSpeech benchmark.
Outcome:	The six leading speech recognition systems were evaluated on a real-world dataset and a synthetic dataset derived from the widely-used LibriSpeech benchmark.

Copied to clipboard

Thennal D K, Jesin James, Deepa Padmini Gopinath, Muhammed Ashraf K

Class-based LSTM Russian Language Model with Linguistic Information (2020.lrec-1)

Challenge:	Word error rate (WER) has been used for automatic speech recognition (ASR) evaluations for English datasets for many years.
Approach:	They propose to use the character error rate as the primary metric in multilingual ASR evaluation to account for morphologically complex languages.
Outcome:	The character error rate (CER) is the primary evaluation metric in multilingual ASR evaluation.

Copied to clipboard

Irina Kipyatkova, Alexey Karpov

WER-BERT: Automatic WER Estimation with BERT in a Balanced Ordinal Classification Paradigm (2021.eacl-main)

Challenge:	LSTM models can be used in speech recognition systems at N-best or lattice rescoring stage.
Approach:	They propose to use word frequency and linguistic information to generate class-based LSTM Russian language models with various numbers of classes.
Outcome:	The proposed models outperform word-based models and word2vec models in terms of perplexity, training time, and word error rate.

Copied to clipboard

Akshay Krishna Sheshadri, Anvesh Rao Vijjini, Sukhdeep Kharbanda

Information Extraction & Text Mining

Challenge:	Automatic Speech Recognition (ASR) systems are evaluated using Word Error Rate (WER) a higher WER means a lower percentage of errors between the ground truth and the transcription of the system.
Approach:	They propose a new balanced paradigm for automatic Word Error Rate estimation using a Librispeech dataset and a Google Cloud's Speech-to-Text API.
Outcome:	The proposed approach is more effective than regression in a classification setting, but suffers from heavy class imbalance.

Evaluating Open-Source ASR Systems: Performance Across Diverse Audio Conditions and Error Correction Methods (2025.coling-main)

Copied to clipboard

Saki Imai, Tahiya Chowdhury, Amanda J. Stent

A Benchmark of French ASR Systems Based on Error Severity (2025.coling-main)

Challenge:	Automated speech recognition (ASR) systems are able to transcribe spontaneous human conversations with high accuracy.
Approach:	They evaluate the accuracy of open source automatic speech recognition systems across conversational speech datasets and explore the potential of ASR ensembling and post-ASR correction methods to improve transcription accuracy.
Outcome:	The proposed methods highlight the need for robust error correction techniques and address demographic biases to enhance ASR performance and inclusivity.

Copied to clipboard

Antoine Tholly, Jane Wottawa, Mickael Rouvier, Richard Dufour

SDS-200: A Swiss German Speech to Standard German Text Corpus (2022.lrec-1)

Challenge:	Automatic Speech Recognition (ASR) transcription errors are often assessed using metrics that compare them with a reference transcription.
Approach:	They propose to categorize transcription errors into four levels of severity based on objective linguistic criteria, contextual patterns, and the use of content words as the unit of analysis.
Outcome:	The proposed evaluation categorizes errors into four levels of severity based on objective linguistic criteria, contextual patterns, and the use of content words as the unit of analysis.

Copied to clipboard

Michel Plüss, Manuela Hürlimann, Marc Cuny, Alla Stöckli, Nikolaos Kapotis, Julia Hartmann, Malgorzata Anna Ulasik, Christian Scheller, Yanick Schraner, Amit Jain, Jan Deriu, Mark Cieliebak, Manfred Vogel

PolyWER: A Holistic Evaluation Framework for Code-Switched Speech Recognition (2024.findings-emnlp)

Challenge:	Using a web recording tool, participants were asked to translate their Swiss German text to their own dialect before recording it.
Approach:	They present a corpus of Swiss German dialectal speech with Standard German text translations . the dataset allows for training speech translation, dialect recognition, and speech synthesis systems .
Outcome:	The dataset allows for training speech translation, dialect recognition, and speech synthesis systems.

Copied to clipboard

Karima Kadaoui, Maryam Ali, Hawau Toyin, Ibrahim Mohammed, Hanan Aldarmaki

FastCorrect 2: Fast Error Correction on Multiple Candidates for Automatic Speech Recognition (2021.findings-emnlp)

Challenge:	Existing methods for measuring accuracy, such as Word Error Rate (WER), are too strict to address this challenge.
Approach:	They propose a framework for evaluating speech recognition systems to handle language-mixing by appending annotations to a publicly available Arabic-English code-switched dataset.
Outcome:	The proposed framework evaluates speech recognition systems against human judgement and a publicly available Arabic-English code-switched dataset.

Copied to clipboard

Yichong Leng, Xu Tan, Rui Wang, Linchen Zhu, Jin Xu, Wenjie Liu, Linquan Liu, Xiang-Yang Li, Tao Qin, Edward Lin, Tie-Yan Liu

WER We Stand: Benchmarking Urdu ASR Models (2025.coling-main)

Challenge:	Error correction is widely used in automatic speech recognition (ASR) to post-process the generated sentence.
Approach:	They propose a fast correction model that takes multiple ASR candidates as input for better correction accuracy.
Outcome:	The proposed model can reduce the word error rate (WER) with multiple candidates by 3.2% and 2.6%.

Copied to clipboard

Samee Arif, Aamina Jamal Khan, Mustafa Abbas, Agha Ali Raza, Awais Athar

Joint Audio/Text Training for Transformer Rescorer of Streaming Speech Recognition (2022.findings-emnlp)

Challenge:	This paper analyzes the performance of three ASR models for low-resource languages like Urdu . low-rural languages like urdu have significant gaps in accuracy and reliability .
Approach:	They evaluate the performance of three ASR models: Whisper, MMS, and Seamless-M4T . they present the first conversational speech dataset for benchmarking Urdu ASR systems .
Outcome:	The proposed model families outperform Whisper, MMS, and Seamless-M4T on two types of speech datasets.

Copied to clipboard

Suyoun Kim, Ke Li, Lucas Kabela, Ron Huang, Jiedan Zhu, Ozlem Kalinli, Duc Le

Accented Speech Recognition With Accent-specific Codebooks (2023.emnlp-main)

Challenge:	Recent studies have shown that streaming end-to-end speech recognition models suffer from higher word error rates (WER) compared to non-streaming models, streaming endto-ended ASR models are limited to short audio context or not use future context to satisfy low latency constraints.
Approach:	They propose a 2nd-pass rescoring model on top of the 1st-pass streaming model to improve recognition accuracy while keeping latency low.
Outcome:	The proposed method improves word error rate significantly compared to the existing model without adding any additional parameters or latency.

Copied to clipboard

Darshan Prabhu, Preethi Jyothi, Sriram Ganapathy, Vinit Unni

Automatic Speech Recognition in Sanskrit: A New Speech Corpus and Modelling Insights (2021.findings-acl)

Challenge:	Degradation in performance across underrepresented accents is a severe deterrent to inclusive adoption of ASR.
Approach:	They propose an approach to adapt speech accents to unseen accents by using cross-attention with a trainable set of codebooks.
Outcome:	The proposed approach yields significant performance gains on the seen English accents and unseen accents on the Mozilla Common Voice dataset.

Copied to clipboard

Devaraja Adiga, Rishabh Kumar, Amrith Krishna, Preethi Jyothi, Ganesh Ramakrishnan, Pawan Goyal

Adapting Language Models When Training on Privacy-Transformed Data (2022.lrec-1)

Challenge:	In this paper, we propose the first large scale study of automatic speech recognition in Sanskrit . we focus on the impact of unit selection in San's ASR systems .
Approach:	They propose a large scale study of automatic speech recognition in Sanskrit . they propose syllable level unit selection that captures character sequences .
Outcome:	The proposed model captures character sequences from one vowel in the word to the next vowela.

Copied to clipboard

Tugtekin Turan, Dietrich Klakow, Emmanuel Vincent, Denis Jouvet

Open ASR for Icelandic: Resources and a Baseline System (L18-1)

Challenge:	Using data sanitization methods to remove personal information from spoken messages is not effective because privacy-transformed data is unlikely to match the test distribution.
Approach:	They propose to use a data sanitization approach to remove personal information from spoken messages by replacing named entities with other words from the same class.
Outcome:	The proposed approach removes personal information from the spoken messages using an automatic named entity recognition method.

Copied to clipboard

Anna Björk Nikulásdóttir, Inga Rún Helgadóttir, Matthías Pétursson, Jón Guðnason

Uni-Dubbing: Zero-Shot Speech Synthesis from Visual Articulation (2024.acl-long)

Challenge:	Existing language resources are not sufficient for less-resourced languages, but a system with sufficient resources is needed.
Approach:	They describe available language resources and their preparation for use in a large vocabulary speech recognition system for Icelandic.
Outcome:	The proposed system improves on acoustic training sets and a speech corpus with a pronunciation dictionary.

Copied to clipboard

Songju Lei, Xize Cheng, Mengjiao Lyu, Jianqiao Hu, Jintao Tan, Runlin Liu, Lingyu Xiong, Tao Jin, Xiandong Li, Zhou Zhao

Ensembles of Hybrid and End-to-End Speech Recognition. (2024.lrec-main)

Challenge:	Multimodal speech synthesis is a key challenge due to the scarcity of datasets that pair audio with corresponding video.
Approach:	They propose a method that incorporates modality alignment during the pre-training phase on multimodal datasets and freezes the video modality extraction component and the encoder module within the pretrained weights.
Outcome:	The proposed method achieves a reduced word error rate (WER) of 31.73%, surpassing the previous best of 33.9% with single-modality audio.

Copied to clipboard

Aditya Kamlesh Parikh, Louis ten Bosch, Henk van den Heuvel

GenPTQ: Green Post-Training Quantization for Large-Scale ASR Models with Mixed-Precision Bit Allocation (2025.findings-emnlp)

Challenge:	Existing methods to combine hybrid and end-to-end ASRs with confidence measures are limited and neither can achieve optimal performance.
Approach:	They propose to combine the hybrid Kaldi-based Automatic Speech Recognition system with the end-to-end wav2vec 2.0 XLS-R ASR using confidence measures.
Outcome:	The proposed method reduces the word error rate by 14% on the primary test set and 20% on other noisy and imbalanced data.

Copied to clipboard

Beom Jin Kang, Hyun Kim

Whispering LLaMA: A Cross-Modal Generative Error Correction Framework for Speech Recognition (2023.emnlp-main)

Challenge:	Large-scale models with high memory and computation demands face significant challenges for deployment.
Approach:	They propose a weight-only quantization method that optimizes for weights and QAT to restore accuracy.
Outcome:	The proposed method achieves 89.1% model size reduction with only 0.8% increase in WER and completes optimization in just 15 seconds.

Copied to clipboard

Srijith Radhakrishnan, Chao-Han Yang, Sumeer Khan, Rohit Kumar, Narsis Kiani, David Gomez-Cabrero, Jesper Tegnér

Language-specific Effects on Automatic Speech Recognition Errors for World Englishes (2022.coling-1)

Challenge:	Existing methods for generative error correction in automatic speech recognition (ASR) use a two-pass reranking paradigm to generate n-best hypotheses.
Approach:	They propose a cross-modal fusion technique for generative error correction in automatic speech recognition.
Outcome:	The proposed technique shows a 37.66% improvement in word error rate relative to the n-best Oracle.

Copied to clipboard

June Choe, Yiran Chen, May Pik Yu Chan, Aini Li, Xin Gao, Nicole Holliday

To Distill or Not to Distill? On the Robustness of Robust Knowledge Distillation (2024.acl-long)

Challenge:	Existing systems are not able to meet the needs of speakers of different demographic groups.
Approach:	They propose to analyze the performance of Otter’s automatic captioning system on native and non-native English speakers of different language background through a linguistic analysis of segment-level errors.
Outcome:	The proposed system predicts certain errors from the phonological structure of a speaker’s native language.

Copied to clipboard

Abdul Waheed, Karima Kadaoui, Muhammad Abdul-Mageed

Responsible & Trustworthy NLP

Challenge:	Existing models for multilingual automatic speech recognition (ASR) are computationallyintensive and lack proper comprehensive evaluations.
Approach:	They propose to distill knowledge from large teacher models into smaller student variants that are more efficient.
Outcome:	The proposed model outperforms existing models on standard benchmarks and dialectal data.

Unsupervised Multi-View Post-OCR Error Correction With Language Models (2021.emnlp-main)

Copied to clipboard

Harsh Gupta, Luciano Del Corro, Samuel Broscheit, Johannes Hoffart, Eliot Brenner

Masked Audio Text Encoders are Effective Multi-Modal Rescorers (2023.findings-acl)

Challenge:	Prior work used text generation techniques or redundancy in similar passages for OCR error correction, which is not appropriate in cases of low corpus redundancies or weak document contextual information.
Approach:	They propose to use a pretrained language model to reconcile different OCR views in unsupervised way so that their combination contains fewer errors than each individual view.
Outcome:	The proposed model can reconcile multiple OCR views so that their combined version contains fewer errors than the best OCR view.

Copied to clipboard

Jinglun Cai, Monica Sunkara, Xilai Li, Anshu Bhatia, Xiao Pan, Sravan Bodapati

Progress in Multilingual Speech Recognition for Low Resource Languages Kurmanji Kurdish, Cree and Inuktut (2022.lrec-1)

Challenge:	Masked Language Models (MLMs) have proven to be effective for second-pass rescoring in Automatic Speech Recognition systems.
Approach:	They propose a multi-modal masked language model rescorer which integrates acoustic representations into the input space of MLM.
Outcome:	The proposed model reduces word error rate (WER) by 4%-16% on in-domain and 3%-7% on out-of-domain datasets over the text-only baseline.

Copied to clipboard

Vishwa Gupta, Gilles Boulianne

XLAVS-R: Cross-Lingual Audio-Visual Speech Representation Learning for Noise-Robust Speech Perception (2024.acl-long)

Challenge:	Using acoustic data, we develop automatic speech recognition systems for three low resource languages.
Approach:	They develop automatic speech recognition systems for three low resource languages using acoustic training data from 12 different languages in the hybrid DNN/HMM framework.
Outcome:	The proposed models are for three low resource languages: Kurmanji Kurdish, Cree and Inuktut.

Copied to clipboard

HyoJung Han, Mohamed Anwar, Juan Pino, Wei-Ning Hsu, Marine Carpuat, Bowen Shi, Changhan Wang

NeuTral Rewriter: A Rule-Based and Neural Approach to Automatic Rewriting into Gender Neutral Alternatives (2021.emnlp-main)

Challenge:	Speech recognition and translation systems perform poorly on noisy inputs, which are frequent in realistic environments.
Approach:	They propose a cross-lingual audio-visual speech representation model for noise-robust speech recognition and translation in over 100 languages.
Outcome:	The proposed model outperforms the previous state-of-the-art by 18.5% WER and 4.7 BLEU on downstream audio-visual speech recognition and translation tasks.

Copied to clipboard

Eva Vanmassenhove, Chris Emmery, Dimitar Shterionov

Paraphrasing

Challenge:	Recent years have seen an increasing need for gender-neutral and inclusive language.
Approach:	They propose a rule-based and a neural approach to gender-neutral rewriting for English . they use manually curated synthetic and natural data to train a rewriter .
Outcome:	The proposed approach improves on the rule-based approach with word error rates below 0.18% on synthetic, in-domain and out-domain test sets.

Listening Like Humans: Semantics-Guided Noise-Robust Multimodal Speech Recognition (2026.acl-long)

Copied to clipboard

Yan Fang, Jun Chen, Yian Yao, Shuxin Zhong, Min Sun, Kaishun Wu

A Unified Speaker Adaptation Approach for ASR (2021.emnlp-main)

Challenge:	Severe acoustic degradation results in unreliable ASR outputs . et al., 2024b): critical concerns regarding reliability and fairness of ASR .
Approach:	They propose a multimodal framework that reframes ASR as semantics-guided speech reconstruction.
Outcome:	The proposed framework achieves an average reduction in WER while also attaining 98.71% BERTScore and 96.7% USE over advanced baselines.

Copied to clipboard

Yingzhu Zhao, Chongjia Ni, Cheung-Chi Leung, Shafiq Joty, Eng Siong Chng, Bin Ma

Retrieval Augmented Generation based context discovery for ASR (2025.findings-emnlp)

Challenge:	Adapting a model to target speakers requires a lot of compute and may cause catastrophic forgetting to the existing speakers.
Approach:	They propose a unified speaker adaptation approach consisting of feature adaptation and model adaptation.
Outcome:	The proposed model outperforms baseline models with 20.58% relative WER reduction and surpasses finetuning method by 2.54% on target speaker adaptation.

Copied to clipboard

Siskos Dimitrios, Stavros Papadopoulos, Pablo Peso Parada, Jisi Zhang, Karthikeyan Saravanan, Anastasios Drosou

Exploring the Potential of Multimodal LLM with Knowledge-Intensive Multimodal ASR (2024.findings-emnlp)

Challenge:	Existing strategies for automatic context discovery remain a challenge . embedding-based retrieval reduces WER by up to 17% relative to using no-context .
Approach:	They propose an embedding-based retrieval approach for automatic context discovery . they propose two alternatives based on large language models to contextualize their effectiveness .
Outcome:	The proposed method reduces WER by up to 17% relative to using no-context .

Copied to clipboard

Minghan Wang, Yuxia Wang, Thuy-Trang Vu, Ehsan Shareghi, Reza Haf

DeRAGEC: Denoising Named Entity Candidates with Synthetic Rationale for ASR Error Correction (2025.findings-acl)

Challenge:	Recent advances in multimodal large language models have made significant progress in integrating information across various modalities, yet real-world applications in educational and scientific domains remain challenging.
Approach:	They propose a task that focuses on transcribing scientific conference videos by leveraging visual information from slides to enhance the accuracy of technical terminologies.
Outcome:	The proposed framework improves transcript quality through post-editing and improves performance over speech-only baselines.

Copied to clipboard

Solee Im, Wonjun Lee, JinMyeong An, Yunsu Kim, Jungseul Ok, Gary Lee

BembaSpeech: A Speech Recognition Corpus for the Bemba Language (2022.lrec-1)

Challenge:	Recent studies have demonstrated that postprocessing speech recognition transcriptions with large language models can significantly enhance the accuracy of Automatic Speech Recognition (ASR).
Approach:	They propose a method to improve Named Entity (NE) correction in Automatic Speech Recognition systems by leveraging phonetic similarity and augmented definitions.
Outcome:	The proposed method outperforms baseline methods on common voice and STOP datasets and achieves a 28% reduction in WER and NE hit ratio.

Copied to clipboard

Claytone Sikasote, Antonios Anastasopoulos

Automatic Speech Recognition for Uyghur through Multilingual Acoustic Modeling (2020.lrec-1)

Challenge:	Existing speech recognition systems for African languages are very low . lack of resources (speech and text) can be attributed to poor quality of speech.
Approach:	They present a preprocessed, ready-to-use automatic speech recognition corpus, BembaSpeech, consisting of 24 hours of read speech in the Bemba language.
Outcome:	The proposed model achieves a word error rate (WER) of 32.91% on the Bemba language . the 1 billion XLS-R parameter model achieve better performance than the monolingual pre-trained English model on the corpus.

Copied to clipboard

Ayimunishagu Abulimiti, Tanja Schultz

CEASR: A Corpus for Evaluating Automatic Speech Recognition (2020.lrec-1)

Challenge:	Low-resource languages suffer from lower performance of Automatic Speech Recognition (ASR) due to the lack of data.
Approach:	They propose to use Turkish as donor language to train acoustic models using multilingual training to achieve more context coverage.
Outcome:	The proposed system performs better with multilingual training for the under-resourced Uyghur language.

Copied to clipboard

Malgorzata Anna Ulasik, Manuela Hürlimann, Fabian Germann, Esin Gedik, Fernando Benites, Mark Cieliebak

Improving Speech Recognition for the Elderly: A New Corpus of Elderly Japanese Speech and Investigation of Acoustic Modeling for Speech Recognition (2020.lrec-1)

Challenge:	Automatic Speech Recognition (ASR) systems are increasingly needed for research and practical applications.
Approach:	They propose to use public speech corpora to evaluate the quality of automatic speech recognition (ASR) they calculate an average Word Error Rate (WER) per corpus, per system and per corpor-system pair .
Outcome:	The proposed corpus evaluates the quality of automatic speech recognition systems using public speech corpora and transcripts generated by state-of-the-art systems.

Copied to clipboard

Meiko Fukuda, Hiromitsu Nishizaki, Yurie Iribe, Ryota Nishimura, Norihide Kitaoka

ÌròyìnSpeech: A Multi-purpose Yorùbá Speech Corpus (2024.lrec-main)

Challenge:	In an aging society, a highly accurate speech recognition system is needed for use in electronic devices for the elderly but this cannot be achieved using conventional speech recognition systems due to the unique features of the speech of elderly people.
Approach:	They construct a new corpus of elderly Japanese speech from existing Japanese speech corpora and train them using existing data.
Outcome:	The proposed models achieve word error rates (WER) as low as 13.38%, exceeding the results of the previous study.

Copied to clipboard

Tolulope Ogunremi, Kola Tubosun, Anuoluwapo Aremu, Iroro Orife, David Ifeoluwa Adelani

On Construction of the ASR-oriented Indian English Pronunciation Dictionary (2020.lrec-1)

Challenge:	rynSpeech corpus is a dataset that can be used for both Text-to-Speecher (TTS) and Automatic Speech Recognition (ASR) speakers of many African languages have no access to voice-enabled applications in their native languages.
Approach:	They propose a dataset to collect Yorùbá speech data that can be used for both TTS and ASR tasks.
Outcome:	The proposed dataset can generate a good quality model with as little as 5 hours of speech . the results are consistent with previous studies on the Yorùbá language .

Copied to clipboard

Xian Huang, Xin Jin, Qike Li, Keliang Zhang

Advancing Arabic Diacritization: Improved Datasets, Benchmarking, and State-of-the-Art Models (2025.emnlp-main)

Challenge:	Indian English (IE) has distinctive characteristics, especially phonologically, from other varieties of English.
Approach:	They build a small IE spontaneous speech corpus and use a linguistically-guided IE pronunciation dictionary to apply it to IE.
Outcome:	The proposed system performs better on IE spontaneous speech data than the one trained with CMUdict.

Copied to clipboard

Abubakr Mohamed, Hamdy Mubarak

Challenge:	Arabic diacritics are typically omitted in written Arabic, leading to ambiguity . authors propose a methodology to analyze and refine a large diacritized corpus .
Approach:	They propose a methodology to analyze and refine a large diacritized corpus to improve training quality.
Outcome:	The proposed model achieves state-of-the-art results with 3.12% and 2.70% WER on WikiNews-2014 and Wikinews-2024.

LegoSLM: Connecting LLM with Speech Encoder using CTC Posteriors (2025.findings-emnlp)

Copied to clipboard

Rao Ma, Tongzhou Chen, Kartik Audhkhasi, Bhuvana Ramabhadran

Multi-Stage Multi-Modal Pre-Training for Automatic Speech Recognition (2024.lrec-main)

Challenge:	Recent studies show that pre-trained speech encoders and large language models can perform suboptimal performance on a range of spoken language processing tasks.
Approach:	They propose to combine large-scale pre-trained speech encoders and large-language models for better performance on automatic speech recognition tasks.
Outcome:	The proposed model can get an average of 49% WER reduction over the baseline model on 8 MLS testsets.

Copied to clipboard

Yash Jain, David M. Chan, Pranav Dheram, Aparna Khare, Olabanji Shonibare, Venkatesh Ravichandran, Shalini Ghosh

myMediCon: End-to-End Burmese Automatic Speech Recognition for Medical Conversations (2024.lrec-main)

Challenge:	Existing methods for pre-training for automatic speech recognition (ASR) focus on single-stage pre-train followed by fine-tuning on downstream task.
Approach:	They propose a multi-modal pre-training method that combines unsupervised pre-training with translation-based supervised mid-training.
Outcome:	The proposed method improves WERs by 38.45% over baselines on both Librispeech and SUPERB.

Copied to clipboard

Hay Man Htun, Ye Kyaw Thu, Hutchatai Chanlekha, Kotaro Funakoshi, Thepchai Supnithi

NB Uttale: A Norwegian Pronunciation Lexicon with Dialect Variation (2024.lrec-main)

Challenge:	Existing medical conversation speech corpora for Burmese are limited, despite advances in ASR.
Approach:	They propose to use a manually curated medical conversation speech corpus for Burmese to examine the performance of ASR models.
Outcome:	The proposed model outperforms the Transformer model and the Recurrent Neural Network (RNN) models.

Copied to clipboard

Marie Iversdatter Røsok, Ingerid Løyning Dale

Syntactic Text Processing

Challenge:	lexicon is based on the NST Bokml lexiconic for East Norwegian . lexica are an essential linguistic resource in speech recognition and speech synthesis systems .
Approach:	They propose to use Bokml orthographic word forms and up to eight alternate phonological transcriptions per word form to generate a Norwegian pronunciation lexicon.
Outcome:	The proposed model improves the accuracy of the proposed model and its outputs with word- and phoneme-error-rate metrics.

MMS-LLaMA: Efficient LLM-based Audio-Visual Speech Recognition with Minimal Multimodal Speech Tokens (2025.findings-acl)

Copied to clipboard

Jeong Hun Yeo, Hyeongseop Rha, Se Jin Park, Yong Man Ro

On the Robust Approximation of ASR Metrics (2025.findings-acl)

Challenge:	Recent Large Language Model (LLM) based AVSR systems incur high computational costs due to high temporal resolution of audio-visual speech.
Approach:	They propose an efficient multimodal speech LLM framework that minimizes token length while preserving essential linguistic content.
Outcome:	The proposed approach reduces token usage by 86% while using only 3.5 tokens per second.

Copied to clipboard

Abdul Waheed, Hanin Atwany, Rita Singh, Bhiksha Raj

Lost in Transcription, Found in Distribution Shift: Demystifying Hallucination in Speech Foundation Models (2025.findings-acl)

Challenge:	Existing methods for estimating speech recognition metrics depend on ground truth labels.
Approach:	They propose a label-free approach to approximating ASR performance metrics . they embed multimodal embeddings in a unified space for speech and transcription representations .
Outcome:	The proposed method outperforms baseline models on speech recognition benchmarks by 50%.

Copied to clipboard

Hanin Atwany, Abdul Waheed, Rita Singh, Monojit Choudhury, Bhiksha Raj

LASER: An LLM-based ASR Scoring and Evaluation Rubric (2025.emnlp-main)

Challenge:	Automatic speech recognition systems have seen remarkable improvements in recent years, but evaluation of performance remains dependent on word and character error rate (WER/CER).
Approach:	They investigate how distribution shifts, model size and model architecture influence hallucination error rate (HER) HER is a metric used to quantify hallucinosity in automatic speech recognition systems.
Outcome:	The proposed model can be used to measure hallucination errors in high-stakes domains such as healthcare, legal, and aviation.

Copied to clipboard

Amruta Parulekar, Preethi Jyothi

Let’s Fuse Step by Step: A Generative Fusion Decoding Algorithm with LLMs for Robust and Instruction-Aware ASR and OCR (2025.findings-acl)

Challenge:	Standard ASR evaluation metrics like word error rate penalize morphological and syntactic nuances that do not significantly alter sentence semantics.
Approach:	They propose an LLM-based scoring rubric LASER that leverages state-of-the-art LLMs’ in-context learning abilities to learn from prompts with detailed examples.
Outcome:	The proposed scoring rubric combines state-of-the-art learning capabilities with a high correlation score with human annotations.

Copied to clipboard

Chan-Jan Hsu, Yi-Chang Chen, Feng-Ting Liao, Pei-Chen Ho, Yu-Hsiang Wang, Po-Chun Hsu, Da-shan Shiu

UniSonate: A Unified Model for Speech, Music, and Sound Effect Generation with Text Instructions (2026.acl-long)

Challenge:	Various fusion strategies have been explored for integration of large language models into multi-modal systems.
Approach:	They propose a framework for deep fusion decoding that integrates large language models into cross-modal text recognition systems.
Outcome:	The proposed framework surpasses cascaded methods in English and Mandarin, and significantly reduces WERs by 17.7%.

Copied to clipboard

Chunyu Qiang, Xiaopeng Wang, Kang Yin, Yuzhe Liang, Yuxin Guo, Teng Ma, Ziyu Zhang, Tianrui Wang, Cheng Gong, Yushen Chen, Ruibo Fu, Longbiao Wang, Jianwu Dang

Zero-Shot Context-Aware ASR for Diverse Arabic Varieties (2026.findings-acl)

Challenge:	Generative audio modeling has been fragmented into specialized tasks such as text-to-speech (TTS), text- to-music (TTM), and text-ta (TTA) specialized models require reference audio for timbre cloning and strict phoneme alignment, whereas TTA models generate unstructured textures from open-ended captions.
Approach:	They propose a unified flow-matching framework capable of synthesizing speech, music, sound effects . they propose 'token injection mechanism' that projects unstructured environmental sounds into structured temporal latent space .
Outcome:	The proposed framework achieves state-of-the-art performance in instruction-based TTS and TTM while maintaining competitive fidelity in TTA.

Copied to clipboard

Bashar Talafha, Amin Abu Alhassan, Muhammad Abdul-Mageed

Speech Recognition Corpus of the Khinalug Language for Documenting Endangered Languages (2024.lrec-main)

Challenge:	Large-scale multilingual ASR has substantially improved recognition for high-resource languages.
Approach:	They propose a proxy-guided -best selection paradigm that conditions inference on external side information without parameter updates.
Outcome:	The proposed model reduces WER by 15.6% relative and recovers a fraction of oracle n-best gains on the common voice MSA testbed.

Copied to clipboard

Zhaolin Li, Monika Rind-Pawlowski, Jan Niehues

The Influence of Automatic Speech Recognition on Linguistic Features and Automatic Alzheimer’s Disease Detection from Spontaneous Speech (2024.lrec-main)

Challenge:	Existing tools to document endangered languages are limited due to data scarcity and the need for training.
Approach:	They propose to use a speech corpus for Khinalug, an endangered language spoken in northern Azerbaijan, to create a model that can be used in language documentation scenarios.
Outcome:	The proposed model achieves 6.65 CER points and 25.53 WER points in low-resource scenarios.

Copied to clipboard

Jonathan Heitz, Gerold Schneider, Nicolas Langer

DM-Codec: Distilling Multimodal Representations for Speech Tokenization (2025.findings-emnlp)

Challenge:	Existing biomarkers for AD diagnosis can only be applied to relatively small sample sizes due to limited availability, excessive costs and invasive nature.
Approach:	They compare automatic speech recognition systems in terms of Word Error Rate (WER) using a publicly available benchmark dataset of speech recordings of AD patients and controls.
Outcome:	The proposed method improves classification performance by replacing manual transcriptions with ASR output.

Copied to clipboard

Md Mubtasim Ahasan, Md Fahim, Tasnim Mohiuddin, Akmmahbubur Rahman, Aman Chadha, Tariq Iqbal, M Ashraful Amin, Md Mofijul Islam, Amin Ahsan Ali

From Naturalness to Norms: Interactional Cultural Competence for SpeechLMs (2026.acl-long)

Challenge:	Existing speech tokenization models lack contextual representations for speech synthesis . absence of contextual representation results in elevated WER and WIL scores .
Approach:	They propose a language model-guided distillation method that incorporates contextual information into a comprehensive speech tokenizer.
Outcome:	The proposed method outperforms state-of-the-art tokenization models in reducing WER and WIL scores.

Copied to clipboard

Santosh T.y.s.s

SpeechIQ: Speech-Agentic Intelligence Quotient Across Cognitive Levels in Voice Understanding by Large Language Models (2025.acl-long)

Challenge:	Spoken language models are increasingly real-time conversational actors.
Approach:	They propose a speech-first view of cultural competence as interactional competence . they synthesize social-science foundations into a taxonomy of culture-bearing signals in speech .
Outcome:	The proposed model is based on a theory-derived taxonomy of culture-bearing signals in speech . it shows that cultural appropriateness is not a generic human-likeness .

Copied to clipboard

Zhen Wan, Chao-Han Huck Yang, Yahan Yu, Jinchuan Tian, Sheng Li, Ke Hu, Zhehuai Chen, Shinji Watanabe, Fei Cheng, Chenhui Chu, Sadao Kurohashi

Unraveling Spontaneous Speech Dimensions for Cross-Corpus ASR System Evaluation for French (2024.lrec-main)

Challenge:	SIQ quantifies voice understanding abilities and provides unified comparisons between cascaded methods and end-to-end models.
Approach:	They propose a human cognition-inspired evaluation pipeline for voice understanding large language models (LLM_Voice) that quantifies voice understanding abilities and provides unified comparisons between cascaded methods and end-to-end models.
Outcome:	The proposed framework quantifies voice understanding abilities and provides unified comparisons between cascaded methods and end-to-end models, identifies annotation errors in existing benchmarks, and detects hallucinations in LLM_Voice.

Copied to clipboard

Solene Virginie Evain, Solange Rossato, François Portet

VietMed: A Dataset and Benchmark for Automatic Speech Recognition of Vietnamese in the Medical Domain (2024.lrec-main)

Challenge:	'spontaneous speech' is a catch-all term used for situations like speaking with a friend, being interviewed on radio/TV or giving a lecture.
Approach:	They propose to use four dimensions to describe spontaneous speech variation in automatic speech recognition systems.
Outcome:	The proposed system can be used to predict the WER of speech recognition systems on face-to-face interactions.

Copied to clipboard

Khai Le-Duc

Idiosyncratic Versus Normative Modeling of Atypical Speech Recognition: Dysarthric Case Studies (2025.emnlp-main)

Challenge:	Currently, there are no publicly available speech recognition datasets in the medical domain due to privacy restrictions.
Approach:	They present a Vietnamese speech recognition dataset in the medical domain comprising 16h of labeled medical speech, 1000h of unlabeled medical and 1200h of general-domain speech.
Outcome:	The proposed model outperforms state-of-the-art models from 51.8% to 29.6% WER on test set.

Copied to clipboard

Vishnu Raja, Adithya V Ganesan, Anand Syamkumar, Ritwik Banerjee, H. Schwartz

Speech-Hands: A Self-Reflection Voice Agentic Approach to Speech Recognition and Audio Reasoning with Omni Perception (2026.acl-long)

Challenge:	Past studies have focused on fully personalized (or idiosyncratic) models for atypical speech . past studies focused on idiotic models, but current approaches focus on generalizing and handling idiomatic patterns .
Approach:	They compare four models that generalize and handle idiosyncrasy to find atypical speech . they find the dysarthric-idios-ync model performs better than the idioconic approach .
Outcome:	The proposed model generalizes and handles idiosyncrasy better than the idiocy model . the model requires less personalized data and reduces word error rate from 71% to 32% .

Copied to clipboard

Zhen Wan, Chao-Han Huck Yang, Jinchuan Tian, Hanrong Ye, Ankita Pasad, Szu-Wei Fu, Arushi Goel, Ryo Hachiuma, Shizhe Diao, Kunal Dhawan, Sreyan Ghosh, Yusuke Hirota, Zhehuai Chen, Rafael Valle, Chenhui Chu, Shinji Watanabe, Boris Ginsburg, Yu-Chiang Frank Wang