SignStream: A Multimedia Tool for Language Research

该项目的目标是：1)开发“符号流”，这是一个用于分析视频语言数据的数据库工具；2)使用SignStream创建美国手语话语的大型数据库，包括原始视频和细粒度的语言转录；3)开发用于将电子文档链接到SignStream数据库的交互式格式（允许从文本到视频、编码数据和SignStream工具的直接访问）。SignStream提供了一个单一的计算环境，通过在多个领域输入语言信息并将其链接到特定的视频帧序列来处理视频和转录话语，从而产生细粒度，多层次，临时校准的转录。来自各个字段的信息在屏幕上直观地对齐，以反映时间关系。目标是使SignStream和数据库以及开发的交互式文档工具对一般研究社区开放。对于手语的语言学研究（迄今为止，由于无法获得用于检查的数据和缺乏科学分析的工具而受到阻碍）以及依赖于视频数据分析的相关学科来说，迫切需要这样的工具。

The project's objectives are: 1) development of "SignStream," a database tool for analysis of linguistic data captured on video; 2) use of SignStream for creation of a large database of utterances from American Sign Language, consisting of raw video and fine-grained linguistic transcription; 3) development of an interactive format for linking electronic documents to SignStream databases (allowing direct access from texts to video, encoded data, and SignStream tools). SignStream provides a single computing environment for manipulating video and transcribing utterances by entering linguistic information in multiple fields and linking this to specific video frame sequences, resulting in a fine-grained, multi-level, temporally calibrated transcription. Information from the various fields is aligned visually on screen to reflect temporal relations. The goal is to make SignStream and the databases and interactive document tools developed accessible to the general research community. There is a pressing need for such tools, for linguistic research on signed languages (which has to date been hampered by the inaccessibility of data for inspection, and the lack of tools for scientific analysis thereof) as well as for related disciplines relying on analysis of video data.