Norma is a translation-based [text] normaliser (TBN). This is a toolkit for building text normalisation models based on open-source tools, is able to learn how to normalise Non-Standard Words (NSW) for a Text-To-Speech system (TTS) in any language. It is a data-driven toolkit, where the developer provides the system with a set of example NSWs and how they normalise. These examples are then used to train an automatic translation system which can then be integrated in the front-end of the TTS system in order to process, not only the provided examples, but also other cases through the generalisation capabilities of the translation-based normaliser.
Compatibility: Linux
[Download]