{"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/reusing-weights-in-subword-aware-neural","title":"Reusing Weights in Subword-aware Neural Language Models","arxiv_id":"1802.08375","date":"2018-02-23","proceeding":"NAACL 2018 6","authors":["Zhenisbek Assylbekov","Rustem Takhanov"],"abstract":"We propose several ways of reusing subword embeddings and other weights in\nsubword-aware neural language models. The proposed techniques do not benefit a\ncompetitive character-aware model, but some of them improve the performance of\nsyllable- and morpheme-aware models while showing significant reductions in\nmodel sizes. We discover a simple hands-on principle: in a multi-layer input\nembedding model, layers should be tied consecutively bottom-up if reused at\noutput. Our best morpheme-aware model with properly reused weights beats the\ncompetitive word-level model by a large margin across multiple languages and\nhas 20%-87% fewer parameters.","url_abs":"http://arxiv.org/abs/1802.08375v2","url_pdf":"http://arxiv.org/pdf/1802.08375v2.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":"reusing-weights-in-subword-aware-neural","repo_url":"https://github.com/zh3nis/morph-sum","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}