We present an experimental software system to design, specify and interpret customized visual languages for different applications. The system consists of two major subsystems: the visual language compiler and the visual language generator. The experimental system is capable of parsing and interpreting visual sentences, and generating a customized visual language from sample visual sentences. Three parsers are supported: a general visual language parser, a fuzzy visual language parser, and a positional grammar parser. Two approaches of specifying semantics are supported: by semantic routines, or by semantic tables. The experimental system has been implemented on IBM PCs and SUN workstations, and applied to visual queries for information retrieval, and the design of iconic languages for augmentative communication.