An adaptive and adaptable multi-purpose math-to-speech translation system is proposed in the paper. Along with the detailed presentation and design approach for the core math-to-speech translation system, exemplary output and tests are discussed. A scripting extension, providing flexibility of the system and enabling the user to adjust the output translations to his/hers preferences is incorporated into the presented solution. Some unit-tests and adaptable versions of translation rule sets are elaborated and evaluated by means of standard measures of machine translation quality. Designed as a flexible self contained solution, the presented system may be applied for various purposes such as e-learning audio presentation systems, educational tools for non-native speakers or visually impaired people, and may be easily adjusted to different national languages.

People with visual impairments suffer because of the lack of well prepared educational resources, especially in the field of mathematics. Existing translation software provides a way to convert basic math documents into Braille language or speak them through Text-To-Speech (TTS) engines, but it requires constant users's interaction and detailed preparations. In this paper we describe a self-contained, adaptive and adaptable Math-To-Speech Interface. We explain the assumptions and design of the system and present exemplary test data. Since the results are quite promising the developed system may be applied in automatic audio presentation tools of mathematical content to sighted as well as visually impaired audience.