{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,25]],"date-time":"2026-10-25T05:35:58Z","timestamp":1792906558508,"version":"4.5.0"},"reference-count":99,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","license":[{"start":{"date-parts":[[2021,1,1]],"date-time":"2021-01-01T00:00:00Z","timestamp":1609459200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2021,1,1]],"date-time":"2021-01-01T00:00:00Z","timestamp":1609459200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2021,1,1]],"date-time":"2021-01-01T00:00:00Z","timestamp":1609459200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"name":"Hong Kong Research Grants Council GRF","award":["14200218"],"award-info":[{"award-number":["14200218"]}]},{"name":"Hong Kong Research Grants Council GRF","award":["14200220"],"award-info":[{"award-number":["14200220"]}]},{"name":"Hong Kong Research Grants Council GRF","award":["14200021"],"award-info":[{"award-number":["14200021"]}]},{"DOI":"10.13039\/501100010428","name":"Innovation and Technology Fund","doi-asserted-by":"publisher","award":["ITS\/254\/19"],"award-info":[{"award-number":["ITS\/254\/19"]}],"id":[{"id":"10.13039\/501100010428","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100006355","name":"Shun Hing Institute of Advanced Engineering","doi-asserted-by":"publisher","award":["MMT-p1-19"],"award-info":[{"award-number":["MMT-p1-19"]}],"id":[{"id":"10.13039\/501100006355","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE\/ACM Trans. Audio Speech Lang. Process."],"published-print":{"date-parts":[[2021]]},"DOI":"10.1109\/taslp.2021.3129357","type":"journal-article","created":{"date-parts":[[2021,11,19]],"date-time":"2021-11-19T15:36:00Z","timestamp":1637336160000},"page":"3679-3693","source":"Crossref","is-referenced-by-count":16,"title":["Mixed Precision Low-Bit Quantization of Neural Network Language Models for Speech Recognition"],"prefix":"10.1109","volume":"29","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-9650-0290","authenticated-orcid":false,"given":"Junhao","family":"Xu","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2449-1436","authenticated-orcid":false,"given":"Jianwei","family":"Yu","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-3345-6923","authenticated-orcid":false,"given":"Shoukang","family":"Hu","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6725-1160","authenticated-orcid":false,"given":"Xunying","family":"Liu","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Helen","family":"Meng","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1109\/TASSP.1987.1165125"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1006\/csla.1999.0128"},{"key":"ref3","first-page":"1137","article-title":"A neural probabilistic language model","volume":"3","author":"Bengio","year":"2003","journal-title":"J. Mach. Learn. Res."},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1016\/j.csl.2006.09.003"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.21437\/Interspeech.2010-343"},{"key":"ref6","first-page":"20","article-title":"Deep neural network language models","volume-title":"Proc. NAACL-HLT 2012 Workshop","author":"Arisoy","year":"2012"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/TASL.2012.2215599"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1109\/TASLP.2015.2400218"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1109\/TASLP.2016.2598304"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1109\/TASLP.2019.2922048"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.21437\/interspeech.2019-2225"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP40776.2020.9053399"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.21437\/Interspeech.2020-1164"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.21437\/Interspeech.2020-2770"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1162\/neco.1997.9.8.1735"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1706.03762"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP39728.2021.9414076"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/d16-1053"},{"key":"ref19","article-title":"A structured self-attentive sentence embedding","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Lin","year":"2017"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D16-1244"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref22","article-title":"Layer normalization","author":"Ba","year":"2016"},{"key":"ref23","first-page":"1243","article-title":"Convolutional sequence to sequence learning","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Gehring","year":"2017"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1109\/ASRU46091.2019.9004025"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP40776.2020.9054345"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP40776.2020.9053896"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/ASRU46091.2019.9003750"},{"key":"ref28","article-title":"Sequence transduction with recurrent neural networks","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Graves","year":"2012"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP.2016.7472621"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.21437\/interspeech.2020-1488"},{"key":"ref31","first-page":"963","article-title":"Expectation backpropagation: Parameter-free training of multilayer neural networks with continuous or discrete weights","volume":"1","author":"Soudry","year":"2014","journal-title":"Proc. Neural Inf. Process. Syst."},{"key":"ref32","first-page":"3123","article-title":"Binaryconnect: Training deep neural networks with binary weights during propagations","volume-title":"Proc. Neural Inf. Process. Syst.","author":"Courbariaux","year":"2015"},{"issue":"1","key":"ref33","first-page":"6869","article-title":"Quantized neural networks: Training neural networks with low precision weights and activations","volume":"18","author":"Hubara","year":"2017","journal-title":"J. Mach. Learn. Res."},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46493-0_32"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00038"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00881"},{"key":"ref37","article-title":"Hawq-v2: Hessian aware trace-weighted quantization of neural networks","author":"Dong","year":"2019","journal-title":"Proc. Adv. Neural Inf. Process. Syst."},{"key":"ref38","first-page":"1135","article-title":"Learning both weights and connections for efficient neural networks","volume-title":"Proc. Neural Inf. Process. Syst.","author":"Han","year":"2015"},{"key":"ref39","article-title":"Deep compression: Compressing deep neural networks with pruning, trained quantization and huffman coding","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Han","year":"2015"},{"key":"ref40","first-page":"1927","article-title":"Exploring the regularity of sparse structure in convolutional neural networks","volume-title":"Proc. IEEE Conf. Comput. Vis. Pattern Recognit.","author":"Mao","year":"2017"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2015.7298681"},{"key":"ref42","first-page":"2074","article-title":"Learning structured sparsity in deep neural networks","volume-title":"Neural Inf. Process. Syst.","volume":"29","author":"Wen","year":"2016"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.308"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.5555\/3298023.3298188"},{"key":"ref45","article-title":"Distilling the knowledge in a neural network","author":"Hinton","year":"2014","journal-title":"Neural Inf. Process. Syst."},{"key":"ref46","article-title":"Fitnets: Hints for thin deep nets","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Romero","year":"2015"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.21437\/Interspeech.2016-1190"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.5244\/C.28.88"},{"key":"ref49","article-title":"Speeding-up convolutional neural networks using fine-tuned CP-decomposition","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Lebedev","year":"2015"},{"key":"ref50","article-title":"Convolutional neural networks with low-rank regularization","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Tai","year":"2016"},{"key":"ref51","first-page":"3088","article-title":"Structured transforms for small-footprint deep learning","volume-title":"Proc. Neural Inf. Process. Syst.","author":"Sindhwani","year":"2015"},{"key":"ref52","article-title":"Squeezenet: Alexnet-level accuracy with 50x fewer parameters and$< $ 0.5 mb model size","author":"Iandola","year":"2016"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2018.00716"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-01264-9_8"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP.2013.6638949"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP.2015.7178919"},{"key":"ref57","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/N18-1192"},{"key":"ref58","doi-asserted-by":"publisher","DOI":"10.1109\/TASLP.2020.3015659"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP.2019.8682336"},{"key":"ref60","doi-asserted-by":"publisher","DOI":"10.21437\/Interspeech.2019-2811"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.21437\/Interspeech.2019-1752"},{"key":"ref62","doi-asserted-by":"publisher","DOI":"10.1631\/FITEE.1800469"},{"key":"ref63","doi-asserted-by":"publisher","DOI":"10.1162\/106365602320169811"},{"key":"ref64","first-page":"2020","article-title":"Neural architecture search with Bayesian optimisation and optimal transport","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Kandasamy","year":"2018"},{"key":"ref65","article-title":"Neural architecture search with reinforcement learning","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Zoph","year":"2017"},{"key":"ref66","article-title":"Designing neural network architectures using reinforcement learning","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Baker","year":"2017"},{"key":"ref67","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v32i1.11709"},{"key":"ref68","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2018.00257"},{"key":"ref69","article-title":"Darts: Differentiable architecture search","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Liu","year":"2018"},{"key":"ref70","article-title":"Snas: Stochastic neural architecture search","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Xie","year":"2019"},{"key":"ref71","article-title":"Proxylessnas: Direct neural architecture search on target task and hardware","author":"Cai","year":"2019","journal-title":"Proc. Int. Conf. Learn. Representations"},{"key":"ref72","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR42600.2020.00242"},{"key":"ref73","doi-asserted-by":"publisher","DOI":"10.1561\/2200000016"},{"key":"ref74","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v32i1.11713"},{"key":"ref75","doi-asserted-by":"publisher","DOI":"10.1109\/ASRU46091.2019.9003744"},{"key":"ref76","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP40776.2020.9054724"},{"key":"ref77","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP40776.2020.9053483"},{"key":"ref78","doi-asserted-by":"publisher","DOI":"10.21437\/interspeech.2016-595"},{"key":"ref79","doi-asserted-by":"publisher","DOI":"10.21437\/Interspeech.2011-328"},{"key":"ref80","doi-asserted-by":"publisher","DOI":"10.1109\/SLT.2014.7078569"},{"key":"ref81","doi-asserted-by":"publisher","DOI":"10.1109\/ICASSP.1992.225858"},{"key":"ref82","doi-asserted-by":"publisher","DOI":"10.1007\/11965152_37"},{"key":"ref83","first-page":"473","article-title":"LSTM can solve hard long time lag problems","author":"Hochreiter","year":"1997","journal-title":"Proc. Neural Inf. Process. Syst."},{"key":"ref84","article-title":"Gaussian error linear units (gelus","author":"Hendrycks","year":"2016"},{"key":"ref85","doi-asserted-by":"publisher","DOI":"10.1109\/icassp49357.2023.10094626"},{"issue":"4","key":"ref86","first-page":"747","article-title":"An extragradient method for finding saddle points and for other problems, konom. I mat","volume-title":"Ekonom. i Mat. Metody","author":"Korpelevi","year":"1976"},{"key":"ref87","article-title":"Entropy-based pruning of backoff language models","volume-title":"Proc. DARPA Broadcast. News Transcription Understanding Workshop","author":"Stolcke","year":"1998"},{"key":"ref88","doi-asserted-by":"publisher","DOI":"10.1109\/ICSLP.1996.607084"},{"key":"ref89","doi-asserted-by":"publisher","DOI":"10.1145\/1944345.1944349"},{"key":"ref90","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-05318-5_3"},{"key":"ref91","first-page":"6758","article-title":"Neural architecture search for LF-MMI trained time delay neural networks","author":"Hu","year":"2020","journal-title":"Proc. IEEE Int. Conf. Acoust., Speech, Signal Process."},{"key":"ref92","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR42600.2020.01210"},{"key":"ref93","first-page":"4095","article-title":"Efficient neural architecture search via parameters sharing","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Pham","year":"2018"},{"key":"ref94","doi-asserted-by":"publisher","DOI":"10.21437\/Interspeech.2021-1962"},{"key":"ref95","article-title":"The Kaldi speech recognition toolkit","volume-title":"Proc. IEEE Workshop Autom. Speech Recognit. Understanding","author":"Povey","year":"2011"},{"key":"ref96","doi-asserted-by":"publisher","DOI":"10.1109\/TASLP.2021.3084072"},{"key":"ref97","doi-asserted-by":"publisher","DOI":"10.21437\/Interspeech.2020-2846"},{"key":"ref98","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-01246-5_2"},{"key":"ref99","article-title":"The concrete distribution: A continuous relaxation of discrete random variables","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Maddison","year":"2017"}],"container-title":["IEEE\/ACM Transactions on Audio, Speech, and Language Processing"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx7\/6570655\/9289074\/09622190.pdf?arnumber=9622190","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,25]],"date-time":"2026-10-25T05:01:30Z","timestamp":1792904490000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/9622190\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2021]]},"references-count":99,"URL":"https:\/\/doi.org\/10.1109\/taslp.2021.3129357","relation":{},"ISSN":["2329-9290","2329-9304"],"issn-type":[{"value":"2329-9290","type":"print"},{"value":"2329-9304","type":"electronic"}],"subject":[],"published":{"date-parts":[[2021]]}}}