The world of multimedia is rapidly growing and reshaping. Multimedia has become ubiquitous, and the tipping point has arrived with the improvements in ease of use and a simple media search. However, video annotation and content-based search is still a great challange. In this talk I will present our work towards this goal. Using automatic annotation tools, a multimodal video search system allows to combine speech, OCR, image-based and visual concepts into a single multimodal query. Mutual relevance feedback is used to assist an expert user with query formulation.

I will give an introduction to the NIST TRECVID international benchmark and present our evaluation results. In the last part of the talk I will present ViaScribe - a project that aims to assist disabled employees with access to corporate information.