Collaborative Research: Automatically Annotated Repository of Digital Video and Audio Resources Community (AARDVARC)

来自未被充分研究的语言的音频和视频数据对语言学家、人类学家、教育家和对视觉动作提取、语音技术或软件本地化感兴趣的计算机科学家很有用。自从通过手持设备进行简单的数字记录以来，文献语言学家已经大量收集了tb级的此类数据。作为消失的语言和文化的记录，视频和音频记录比纸质记录更丰富、更迷人，但它们需要被索引和转录，以充分发挥它们作为研究工具的潜力。当前的项目AARDVARC（自动注释数字音视频资源社区资源库）将通过建立一个由语言学家、人类学家和计算机科学家组成的跨学科社区来共享知识，并就资源库规范和一套工具进行协作，以促进转录，从而解决未转录的问题，因此无法获得未被研究语言的文档。它将提供两个讲习班和一个研讨会，设计一个“取一留一”的存储库，并探索语音和视频处理的最新进展，这将使人类学家和语言学家打破语言数据的“转录瓶颈”。即使是部分自动化也将极大地促进分析人员的工作，并大大增加多学科研究人员可用的转录音频和视频的数量。

Audio and video data from understudied languages is useful to linguists, anthropologists, educators, and computer scientists interested in visual action extraction, speech technology or software localization. Terabytes of such data exist, having been collected in large amounts by documentary linguists since the advent of easy digital recording via handheld devices. As records of vanishing languages and cultures, video and audio records are far richer and more captivating than paper records, but they need to be indexed and transcribed so that they reach their full potential as research tools. The current project, AARDVARC (Automatically Annotated Repository of Digital Audio and Video Resources Community) will address the problem of untranscribed, and therefore unavailable, documentation of understudied languages by building an interdisciplinary community of linguists, anthropologists, and computer scientists to share knowledge and collaborate on the specification of a repository and suite of tools to facilitate transcription. It will provide for two workshops and a symposium to design a "take one leave one" repository and to explore recent advances in speech and video processing that will allow anthropologists and linguists to break the 'transcription bottleneck' for language data. Even partial automation will greatly facilitate the work of the analyst and dramatically increase the amount of transcribed audio and video available to researchers in multiple disciplines.