BCC: Collaborative Research: Community Building for Research on Mathematics Learning Using Data-Intensive Sources

视频数据在STEM教育研究和开发中无处不在。然而，这些数据在最初开发项目之外使用的潜力受到了劳动密集型方法的限制，这些方法以支持外部研究人员使用的方式对视频片段进行索引和分类。这个项目将建立社区，以确定分类，视频可能被定性。研究人员和一个强大的顾问委员会将与InVideo的程序员合作，InVideo是一种用于索引视频数据集中的语言和非语言对象的工具和软件。该项目使用罗伯特·B·戴维斯学习研究所的学生和教师在数学教室里的视频，将通过连续引入更复杂的视频样本来确定集合中哪些类别的视频可以进行索引。该项目将推动计算机科学算法在视频搜索方面的应用，使计算机科学界和数学教育研究和发展界都受益。目前还没有一种自动方法来访问锁定在视频格式中的内容数据，这种格式具有成本效益并可用于教育目的。参与该项目的研究人员将确定是否有可能应用自动索引来生成高精度和高召回率的检索；哪些人工增强的注释可以生成更有效的搜索；以及获得IRB批准以在更广泛的研究人员社区中共享视频资源以促进对其他数据集进行细粒度分析的可取做法。该项目将培训inVideo的音频到文本翻译功能，以便使用目前存在的翻译识别视频中记录的人的语音模式。该项目的第二阶段将涉及测试新的视频样本的过程，这些视频样本不存在人工生成的抄本。该项目还将为inVideo开发一个网络界面，以方便远程用户进行文字记录编辑、带有时间戳的评论和标记。这项研究的结果将提高教育视频对更广泛的研究人员的实用性。

Video data is ubiquitous in STEM education research and development. Yet the potential for these data to be used outside the initial project in which they are developed has been limited by the labor intensive methodology to index and categorize video segments in ways that support use by external researchers. This project will build the community to determine the taxonomy by which videos might be characterized. The researchers and a strong advisory board will work with the programmers of inVideo, a tool and software for indexing language and non-language objects in video data sets. Working with video of students and teachers in mathematics classrooms from the Robert B. Davis Institute for Learning, the project will determine which categories of videos from the collection are amenable for indexing by introducing successively more complex video samples. The project will push the boundaries of computer science algorithms for video search, benefiting both the computer science community and the mathematics education research and development communities.There is currently no automated method for accessing the content data locked within a video format that is cost effective and available for education purposes. The researchers in this project will determine whether application of automated indexing to generate high precision and high recall retrievals is possible; what human-augmented annotation generates towards more effective searching; and, advisable practices for obtaining IRB approval to share video resources among a wider community of researchers to facilitate a fine-grained analysis on other data sets. The project will train the audio-to-text translation function of inVideo to recognize speech patterns of people recorded in the videos using translations that currently exist. The second stage of the project will address the process of testing new samples of videos for which human-generated transcripts do not exist. The project will also develop a web interface for inVideo to facilitate transcript editing, time-stamped commenting and tagging by remote users. The results of this study will enhance the utility of educational video to a wide community of researchers.