Email2Vmail — An Email Reader

Abstract

This paper describes the results of the email2vmail project whose main goal was to modify the Dixi+ text-to-speech (TTS) system for reading email messages. Although a TTS system is usually able to read any type of text, the normal audio rendering may not be appropriate for the contents of an e-mail message. For instance, there are differences in the usefulness of the information in the header, body and signature. E-mail messages also include specific uses of sequences of symbols for e-mail and web addresses, for quoting other messages or for graphical and formatting purposes. Another difficulty is that the message does not contain any indication on the language it is written in and, in some cases, it even includes text in multiple languages. This work tries to handle most of these problems by marking-up the message text with clues for how it should be read by the TTS system.