DRAFT: Orthoepikon: a toolbox to build reading-aloud assistants

Orthoepikon is a toolbox which allows for the automatic construction of systems to assist people to read texts aloud correctly. Each system is generated from a linguistic data file (containing a dictionary and rules) written in a standard format based on XML. The systems read plain text, HTML or RTF, and add to them annotations which are easy to interpret and indicate the correct pronunciation in difficult or exceptional cases. Annotation is almost instantaneous thanks to the use of finitestate techniques. The application of the described tools to a particular language is briefly described.