This project has presented a methodology for retrieving information graphics
relevant to user queries; it takes into account a graphic's structural content
and the graphic's high-level intended message. Our experimental results show
that our methodology improves significantly over a baseline method that treats
graphics and queries as single bags of words. To our knowledge, our work is the
first to investigate the use of structural and message information in the retrieval
of infographics. We are currently extending the variety of relevance features and
exploring widely used learning-to-rank algorithms to achieve higher retrieval
performance. We will also explore question-answering from relevant graphics.