{"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/understanding-image-and-text-simultaneously-a","title":"Understanding Image and Text Simultaneously: a Dual Vision-Language Machine Comprehension Task","arxiv_id":"1612.07833","date":"2016-12-22","proceeding":null,"authors":["Nan Ding","Sebastian Goodman","Fei Sha","Radu Soricut"],"abstract":"We introduce a new multi-modal task for computer systems, posed as a combined\nvision-language comprehension challenge: identifying the most suitable text\ndescribing a scene, given several similar options. Accomplishing the task\nentails demonstrating comprehension beyond just recognizing \"keywords\" (or\nkey-phrases) and their corresponding visual concepts. Instead, it requires an\nalignment between the representations of the two modalities that achieves a\nvisually-grounded \"understanding\" of various linguistic elements and their\ndependencies. This new task also admits an easy-to-compute and well-studied\nmetric: the accuracy in detecting the true target among the decoys.\n  The paper makes several contributions: an effective and extensible mechanism\nfor generating decoys from (human-created) image captions; an instance of\napplying this mechanism, yielding a large-scale machine comprehension dataset\n(based on the COCO images and captions) that we make publicly available; human\nevaluation results on this dataset, informing a performance upper-bound; and\nseveral baseline and competitive learning approaches that illustrate the\nutility of the proposed task and dataset in advancing both image and language\ncomprehension. We also show that, in a multi-task learning setting, the\nperformance on the proposed task is positively correlated with the end-to-end\ntask of image captioning.","url_abs":"http://arxiv.org/abs/1612.07833v1","url_pdf":"http://arxiv.org/pdf/1612.07833v1.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":[],"tasks":[{"task_slug":"image-captioning","task_name":"Image Captioning"},{"task_slug":"multi-task-learning","task_name":"Multi-Task Learning"},{"task_slug":"reading-comprehension","task_name":"Reading Comprehension"}],"methods":[],"datasets_introduced":[{"slug":"mcic-coco","name":"MCIC-COCO","full_name":""}],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1612.07833","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}