{"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/encoding-spatial-relations-from-natural","title":"Encoding Spatial Relations from Natural Language","arxiv_id":"1807.01670","date":"2018-07-04","proceeding":null,"authors":["Tiago Ramalho","Tomáš Kočiský","Frederic Besse","S. M. Ali Eslami","Gábor Melis","Fabio Viola","Phil Blunsom","Karl Moritz Hermann"],"abstract":"Natural language processing has made significant inroads into learning the\nsemantics of words through distributional approaches, however representations\nlearnt via these methods fail to capture certain kinds of information implicit\nin the real world. In particular, spatial relations are encoded in a way that\nis inconsistent with human spatial reasoning and lacking invariance to\nviewpoint changes. We present a system capable of capturing the semantics of\nspatial relations such as behind, left of, etc from natural language. Our key\ncontributions are a novel multi-modal objective based on generating images of\nscenes from their textual descriptions, and a new dataset on which to train it.\nWe demonstrate that internal representations are robust to meaning preserving\ntransformations of descriptions (paraphrase invariance), while viewpoint\ninvariance is an emergent property of the system.","url_abs":"http://arxiv.org/abs/1807.01670v2","url_pdf":"http://arxiv.org/pdf/1807.01670v2.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":"encoding-spatial-relations-from-natural","repo_url":"https://github.com/deepmind/slim-dataset","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"spatial-reasoning","task_name":"Spatial Reasoning"}],"methods":[],"datasets_introduced":[{"slug":"slim","name":"Slim","full_name":"Slim"}],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1807.01670","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}