{"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/multimodal-locally-enhanced-transformer-for","title":"Multimodal Locally Enhanced Transformer for Continuous Sign Language Recognition","arxiv_id":null,"date":"2023-08-22","proceeding":"Conference of the International Speech Communication Association (INTERSPEECH) 2023 8","authors":["Katerina Papadimitriou","Gerasimos Potamianos"],"abstract":"In this paper, we propose a novel Transformer-based approach for continuous sign language recognition (CSLR) from videos, aiming to address the shortcomings of traditional Transformers in learning local semantic context of SL. Specifically, the proposed relies on two distinct components: (a) a window-based RNN module to capture local temporal context and (b) a Transformer encoder, enhanced with local modeling via Gaussian bias and relative position information, as well as with global structure modeling through multi-head attention. To further improve model performance, we design a multimodal framework that applies the proposed to both appearance and motion signing streams, aligning their posteriors through a guiding CTC technique. Further, we achieve visual feature and gloss sequence alignment by incorporating a knowledge distillation loss. Experimental evaluation on two popular German CSLR datasets, demonstrates the superiority of our model.","url_abs":"https://www.isca-archive.org/interspeech_2023/papadimitriou23_interspeech.html","url_pdf":"https://www.isca-archive.org/interspeech_2023/papadimitriou23_interspeech.html","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":[],"tasks":[{"task_slug":"knowledge-distillation","task_name":"Knowledge Distillation"},{"task_slug":null,"task_name":"Position"},{"task_slug":"sign-language-recognition","task_name":"Sign Language Recognition"}],"methods":[{"method_slug":"absolute-position-encodings","method_name":"Absolute Position Encodings"},{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"knowledge-distillation","method_name":"Knowledge Distillation"},{"method_slug":"label-smoothing","method_name":"Label Smoothing"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"position-wise-feed-forward-layer","method_name":"Position-Wise Feed-Forward Layer"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"transformer","method_name":"Transformer"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/sign-language-recognition-on-rwth-phoenix","task":"Sign Language Recognition","dataset":"RWTH-PHOENIX-Weather 2014","model":"WRNN + LET","rank_in_archive_order":12,"of":22,"metrics":{"Word Error Rate (WER)":"20.89"},"uses_additional_data":false},{"leaderboard":"/sota/sign-language-recognition-on-rwth-phoenix-1","task":"Sign Language Recognition","dataset":"RWTH-PHOENIX-Weather 2014 T","model":"WRNN + LET","rank_in_archive_order":9,"of":15,"metrics":{"Word Error Rate (WER)":"20.73"},"uses_additional_data":false}],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}