遇见数据集

NGT video stimuli, with and without mouth actions

收藏
CESSDA2025-02-11 更新2025-04-12 收录
官方服务:

资源简介:

This repository consists of 908 high-quality videos in .mov format of single signs in Nederlandse Gebarentaal (NGT); i.e., Dutch Sign Language, also called Sign Language of the Netherlands. These NGT signs were created as experimental stimuli for a study led by principal investigator, Ellen Ormel (Radboud University). The sign model in all videos is Ms. Merel van Zuilen, who is filmed with a dark long-sleeved shirt on a plain blue background. The video file sizes range from 16 megabytes to over 40 megabytes. <br><br> There are two overall types of videos as part of the experimental design: (1) those in which the manual sign is accompanied by a mouth action (449 videos) and (2) those articulated without any mouth action (459 videos). "Mouth action" refers to both "mouthings" that map to the articulations in a spoken word and "mouth gestures" that are linguistic elements that accompany the movement of the hands and body. All together, there are 425 unique manual signs in the dataset and most of these manual signs have both types of videos: a manual sign with a mouth action and the same manual sign without a mouth action. There are also phonetic variants on some signs, which increases the overall number of videos. <br> <br> The filenaming convention for the files is as follows. The filename starts with a gloss translation in Dutch in capital letters (e.g., APPEL) followed by an optional dash separator and letter to indicate a lexical variant (e.g., KROON-A, KROON-B), followed by an optional underscore and number to indicate a phonetic variant (e.g., ONTDOOIEN_1, ONTDOOIEN_2, PRATEN-A_1, PRATEN-A_2), followed by an obligatory underscore and two-letter code for the mouth actions status of the sign: MM, GM, or ZM. 'MM' for <i>met mondbeeld</i> indicates a mouthing that typically mirrors a spoken word in Dutch; 'GM' for <i>gesture mondbeeld</i> indicates a mouth gesture (including mouth morphemes); and ZM for <i> zonder mondbeeld </i> indicates no mouth action at all. For example, AARDBEI_MM.mov is a video of the NGT sign for strawberry (<i>aardbei</i>) produced with mouthing, while KAMMEN-A_1_ZM.mov is a video of the first phonetic type of the lexical variant A of the sign for to comb/combing (<I>kammen</i>) and is produced without mouth movement. <br> <br> Four other documents are in the repository: (1) an Excel spreadsheet showing the 425 unique manual NGT signs with index number, Dutch gloss, indication of a matching sign in the NGT dataset in Global Signbank (https://signbank.cls.ru.nl), the number of sub-variants, whether there is a video with and without a mouth action, and the total number of videos associated with the manual sign; (2) an 8-page PDF of the Excel spreadsheet; (3) a comma-delimited document [original format] that lists all video filenames for those signs with a mouth action (NGT_Signs_With_Mouth_Action-MM_GM_449signs.csv); (4) a comma-delimited document [original format] that lists all video filenames for those signs without any mouth action (NGT_Signs_Without_Mouth_Action_ZM_459signs.csv). Tags are applied to each document type for custom searches: <i>Data</i>, <i>Documentation</i>, <i>+mouth</i>, <i>-mouth</i>, <i>mouth_gesture</i>.

本数据集仓库包含908条格式为.mov的高质量单手势视频,对应荷兰手势语(Nederlandse Gebarentaal,NGT),即荷兰手语(又称荷兰国家手语)。这些NGT手势被用作由拉德堡德大学首席研究员艾伦·奥梅尔(Ellen Ormel)主导的一项研究的实验刺激材料。所有视频中的手势演示者均为梅雷尔·范·祖伊伦(Merel van Zuilen)女士,拍摄时她身着深色长袖上衣,背景为纯蓝色。视频文件大小介于16兆字节至40兆字节之间。 本数据集的实验设计包含两类视频:(1)手部手势伴随口部动作的视频(共449条);(2)无任何口部动作的手势视频(共459条)。“口部动作”涵盖两类:一是与口语发音对应的“口型匹配”(mouthings),二是伴随手部与身体动作的语言性“口部手势”(mouth gestures)。本数据集共包含425个独特的手部手势,其中绝大多数手势均配有两类视频:带口部动作版与无口部动作版。部分手势还存在语音变体,这使得总视频数量有所增加。 视频文件的命名规则如下:文件名以大写的荷兰语释义词(gloss)开头(例如APPEL);其后可添加连字符与字母,用于表示词汇变体(例如KROON-A、KROON-B);接着可添加下划线与数字,用于表示语音变体(例如ONTDOOIEN_1、ONTDOOIEN_2、PRATEN-A_1、PRATEN-A_2);最后必须添加下划线与表示口部动作状态的两位字母代码:MM、GM或ZM。其中,<i>met mondbeeld</i>对应的MM表示与荷兰语口语发音匹配的口型;<i>gesture mondbeeld</i>对应的GM表示口部手势(包含口部语素);<i>zonder mondbeeld</i>对应的ZM表示完全无口部动作。例如,AARDBEI_MM.mov为搭配口型的荷兰语“草莓”(<i>aardbei</i>)对应的NGT手势视频;而KAMMEN-A_1_ZM.mov则为“梳理(头发)”(<i>kammen</i>)的词汇变体A的第一种语音类型的无口部动作手势视频。 本数据集仓库还包含4份附加文档:(1)一份Excel表格,涵盖425个独特NGT手部手势的相关信息,包括索引编号、荷兰语释义词、全球手势语库(Global Signbank,https://signbank.cls.ru.nl)中对应NGT手势的匹配标识、子变体数量、是否存在带口部动作与无口部动作的视频,以及该手势关联的总视频数;(2)一份对应上述Excel表格的8页PDF文档;(3)一份逗号分隔值格式的原始文档,列出所有带口部动作的手势对应的视频文件名(文件名为NGT_Signs_With_Mouth_Action-MM_GM_449signs.csv);(4)一份逗号分隔值格式的原始文档,列出所有无口部动作的手势对应的视频文件名(文件名为NGT_Signs_Without_Mouth_Action_ZM_459signs.csv)。所有文档均已添加自定义搜索标签:<i>Data</i>、<i>Documentation</i>、<i>+mouth</i>、<i>-mouth</i>以及<i>mouth_gesture</i>。

二维码
社区交流群
二维码
科研交流群
商业服务