A Prototype System of Search: Finding Short Material for Science Education in Long and High-Definition Documentary Videos

High-definition (HD) video material for science education is always welcomed in elementary or secondary schools. It not only helps teachers to convey concepts but also impresses pupils. However, it is difficult for teachers to find customized HD video material besides the unified compact disks companied with teaching reference books. The teacher-customized HD video material has two requirements: (1) short (less than 3–5 min) and (2) fit to the current knowledge map. One potential pool of HD video material is HD documentaries. Nevertheless, a HD documentary is usually very long (around 45–90 min) and has its own agenda. In this chapter, a prototype system named SEARCH (Seeking Excerpted educAtional Resource in Collections of High-definition documentaries) was proposed to help teachers find short material for science education in long and high-definition documentary videos. SEARCH consists of three vital components: knowledge map extraction, documentary subtitle tagging, and hit re-ranking. The knowledge map extraction component is to extract knowledge map from teaching reference books, assigning different weights to concepts in different positions in a knowledge map. The documentary subtitle tagging component is to tag subtitles with concepts extracted in the former component, using a deep learning technique: long short-term memory. The hit re-ranking component is to re-rank multiple hits returned by scanning tags, according to different demands, such as relevance first or most viewed first. Preliminary results indicate that it can facilitate teachers to prepare their courses.