Enrichment of news show videos with multimodal semi-automatic analysis

Enriching linear videos by offering continuative and related information via, e.g.,  audiostreams, webpages, as well as other videos, is typically hampered by its demand for massive editorial work. While there exist several (semi-)automatic methods that analyse audio/video content, one needs to decide which method offers appropriate information for an intended use-case scenario. In this paper, we present the news show scenario as defined within the LinkedTV project, and derive its necessities based on expected user archetypes.
We then proceed to review the technology options for video analysis that we have access to, and describe which training material we opted for to feed our algorithms. Finally, we offer preliminary quality feedback results and give an outlook on the next steps within the project.