Advertising is a huge market. Each year 70 billion Euro are spent only in the European Union. Analysis of advertising sector is a task of media monitoring companies, which is currently mainly performed manually. Recently performed research projects have contributed to automate parts of the workflow, however, technologies are concentrating on multi-modal low level analysis and fusion of those results. Semantic knowledge, e.g. driven by domain specific media monitoring ontologies, has not been considered so far. Systems for highly automated trend analysis, identification of new business areas, product / company analysis are currently not available. On the other hand media monitoring companies and market research institutes would greatly benefit from such tools in order to ease their work and make results faster available.

We will perform research into a semantic media analysis framework (SEMAF) capable of taking as input low level feature metadata received from visual analysis of advertisements and public knowledge (i.e. company website info, stock market info, etc.). SEMAF will be a generic framework of modules with open interfaces which can be tailored easily by instantiation to specific domains of customer queries – e.g. trends for advertising expenditure in automotive sector as domain. Within the UAd application area specific ontologies and rules will be developed and applied together with the public knowledge in order to relate and homogenize information from heterogeneous sources and generate an answer to the user query. Main challenges and at the same time innovations will be to take context information into consideration - such as ontologies, rules, public knowledge and low level feature results - next to a dynamic adaptation of the knowledge model represented by the UAd ontologies by application of the rules. Furthermore a major challenge will be also to take user feedback into account.