{"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/end-to-end-attention-based-large-vocabulary","title":"End-to-End Attention-based Large Vocabulary Speech Recognition","arxiv_id":"1508.04395","date":"2015-08-18","proceeding":null,"authors":["Dzmitry Bahdanau","Jan Chorowski","Dmitriy Serdyuk","Philemon Brakel","Yoshua Bengio"],"abstract":"Many of the current state-of-the-art Large Vocabulary Continuous Speech\nRecognition Systems (LVCSR) are hybrids of neural networks and Hidden Markov\nModels (HMMs). Most of these systems contain separate components that deal with\nthe acoustic modelling, language modelling and sequence decoding. We\ninvestigate a more direct approach in which the HMM is replaced with a\nRecurrent Neural Network (RNN) that performs sequence prediction directly at\nthe character level. Alignment between the input features and the desired\ncharacter sequence is learned automatically by an attention mechanism built\ninto the RNN. For each predicted character, the attention mechanism scans the\ninput sequence and chooses relevant frames. We propose two methods to speed up\nthis operation: limiting the scan to a subset of most promising frames and\npooling over time the information contained in neighboring frames, thereby\nreducing source sequence length. Integrating an n-gram language model into the\ndecoding process yields recognition accuracies similar to other HMM-free\nRNN-based approaches.","url_abs":"http://arxiv.org/abs/1508.04395v2","url_pdf":"http://arxiv.org/pdf/1508.04395v2.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":"end-to-end-attention-based-large-vocabulary","repo_url":"https://github.com/rizar/attention-lvcsr","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"acoustic-modelling","task_name":"Acoustic Modelling"},{"task_slug":"language-modeling","task_name":"Language Modeling"},{"task_slug":"language-modelling","task_name":"Language Modelling"},{"task_slug":"speech-recognition","task_name":"Speech Recognition"},{"task_slug":"speech-recognition-1","task_name":"speech-recognition"}],"methods":[{"method_slug":"speed","method_name":"SPEED"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1508.04395","atlas_url":"https://app.syntology.ai/?focus=1508.04395","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}