A Statistical and Structural Approach for Symbol Recognition, Using XML Modelling

International audienceThis paper deals with the problem of symbol recognition in technical document interpretation. We present a system using a statistical and structural approach. This system uses two interpretation levels. In a first level, the system extracts and recognizes the loops of symbols. In the second level, it relies on proximity relations between the loops in order to rebuild loop graphs, and then to recognize the complete symbols. Our aim is to build a generic device, so we have tried to outsource models descriptions and tools parameters. Data manipulated by our system are modelling in XML. This gives the system the ability to interface tools using different communication data structures, and to create graphic representation of process results