{"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/grounded-textual-entailment","title":"Grounded Textual Entailment","arxiv_id":"1806.05645","date":"2018-06-14","proceeding":"COLING 2018 8","authors":["Hoa Trong Vu","Claudio Greco","Aliia Erofeeva","Somayeh Jafaritazehjan","Guido Linders","Marc Tanti","Alberto Testoni","Raffaella Bernardi","Albert Gatt"],"abstract":"Capturing semantic relations between sentences, such as entailment, is a\nlong-standing challenge for computational semantics. Logic-based models analyse\nentailment in terms of possible worlds (interpretations, or situations) where a\npremise P entails a hypothesis H iff in all worlds where P is true, H is also\ntrue. Statistical models view this relationship probabilistically, addressing\nit in terms of whether a human would likely infer H from P. In this paper, we\nwish to bridge these two perspectives, by arguing for a visually-grounded\nversion of the Textual Entailment task. Specifically, we ask whether models can\nperform better if, in addition to P and H, there is also an image\n(corresponding to the relevant \"world\" or \"situation\"). We use a multimodal\nversion of the SNLI dataset (Bowman et al., 2015) and we compare \"blind\" and\nvisually-augmented models of textual entailment. We show that visual\ninformation is beneficial, but we also conduct an in-depth error analysis that\nreveals that current multimodal models are not performing \"grounding\" in an\noptimal fashion.","url_abs":"http://arxiv.org/abs/1806.05645v1","url_pdf":"http://arxiv.org/pdf/1806.05645v1.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":"grounded-textual-entailment","repo_url":"https://github.com/claudiogreco/coling18-gte","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"natural-language-inference","task_name":"Natural Language Inference"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/natural-language-inference-on-v-snli","task":"Natural Language Inference","dataset":"V-SNLI","model":"V-BiMPM","rank_in_archive_order":2,"of":3,"metrics":{"Accuracy":"86.99"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-v-snli","task":"Natural Language Inference","dataset":"V-SNLI","model":"BiMPM","rank_in_archive_order":3,"of":3,"metrics":{"Accuracy":"86.41"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1806.05645","atlas_url":"https://app.syntology.ai/?focus=1806.05645","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}