The DAISY 3 MathML Modular Extension allows all of the features of a Digital Talking Book to work for math, just as they do for literary text. But this extension does not specify details about advanced MathML players on the localization (i.e. support of a specific native language), support of multilingual mathematical or textual content, cultural differences or user preferences. In this work we present the methodology, we have developed to evaluate in a systematic way the correctness of the Math-to-Speech rendering by investigating specific steps during the production and rendering phases for non-English DAISY Digital Talking Math Books. We define four types of errors in the math expressions: deletions (D), insertions (I), substitutions (S) and translations (T). More-over, we propose an appropriate corpus of mathematical expressions that facilitates the evaluation of the proposed methodology. Then, we have applied the above in the case of the Greek language, a not widespread inflectional natural language. The results of the evaluation for the Greek language and the Math-Player working with the ReadHear player are for each category of math expressions in the whole corpus: trigonometry = 2D, fractions = 3D, matrices = 18T & 4D, integrals = 10T, 12S & 4D, limits = 2T, 12S & 4D, powers-indicators = 8T, 4S & 1D and sets = 9T, 1S, 5D & 3I.
CITATION STYLE
Riga, P., Kouroupetroglou, G., & Ioannidou, P. P. (2016). An evaluation methodology of math-to-speech in non-english DAISY digital talking books. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 9758, pp. 27–34). Springer Verlag. https://doi.org/10.1007/978-3-319-41264-1_4
Mendeley helps you to discover research relevant for your work.