{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/bidirectional-quaternion-long-short-term","title":"Bidirectional Quaternion Long-Short Term Memory Recurrent Neural Networks for Speech Recognition","arxiv_id":"1811.02566","date":"2018-11-06","proceeding":null,"authors":["Titouan Parcollet","Mohamed Morchid","Georges Linarès","Renato de Mori"],"abstract":"Recurrent neural networks (RNN) are at the core of modern automatic speech\nrecognition (ASR) systems. In particular, long-short term memory (LSTM)\nrecurrent neural networks have achieved state-of-the-art results in many speech\nrecognition tasks, due to their efficient representation of long and short term\ndependencies in sequences of inter-dependent features. Nonetheless, internal\ndependencies within the element composing multidimensional features are weakly\nconsidered by traditional real-valued representations. We propose a novel\nquaternion long-short term memory (QLSTM) recurrent neural network that takes\ninto account both the external relations between the features composing a\nsequence, and these internal latent structural dependencies with the quaternion\nalgebra. QLSTMs are compared to LSTMs during a memory copy-task and a realistic\napplication of speech recognition on the Wall Street Journal (WSJ) dataset.\nQLSTM reaches better performances during the two experiments with up to $2.8$\ntimes less learning parameters, leading to a more expressive representation of\nthe information.","url_abs":"http://arxiv.org/abs/1811.02566v1","url_pdf":"http://arxiv.org/pdf/1811.02566v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"bidirectional-quaternion-long-short-term","repo_url":"https://github.com/mravanelli/pytorch-kaldi","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"bidirectional-quaternion-long-short-term","repo_url":"https://github.com/Orkis-Research/Pytorch-Quaternion-Neural-Networks/tree/28caa7cde240e354fd7b87280450fd233cd494c3/exp/icassp_2019","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"GPL-3.0"}}],"tasks":[{"task_slug":"automatic-speech-recognition-2","task_name":"Automatic Speech Recognition"},{"task_slug":"automatic-speech-recognition","task_name":"Automatic Speech Recognition (ASR)"},{"task_slug":"speech-recognition","task_name":"Speech Recognition"},{"task_slug":"speech-recognition-1","task_name":"speech-recognition"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}