Mathematical knowledge contained in scientific digital publications poses a challenge for intelligent retrieval mechanisms. Many current approaches use statistical (e.g. Google) or natural language processing methods to find correlations in texts and annotate texts semantically. However both kinds of approaches face the problem of extracting and processing knowledge from mathematical equations. The presented system is based on natural language processing techniques, and benefits from characteristic linguistic structures defined by the language used in mathematical texts. It accumulates extracted information snippets from texts, symbols, and equations in knowledge bases. These knowledge bases provide the foundation for the information retrieval. This article describes the concepts and the prototypical technical implementation.