This work presents the design and implementation of the DocEmoX system for the automated typography-derived emotional extraction and annotation of printed and electronic documents. The DocEmoX system targets the Design-for-All based multimodal accessibility of documents. The methodology is based on the results derived from a number of readers' emotional state response experiments that model the mapping of any combination of typographic elements into specific analogous variations of the three emotional dimensions (Valence/Pleasure, Arousal and Potency/ Dominance) using a set of Emotional Rules. DocEmoX implements these Emotional Rules in XSL format and produces the annotated output document following the ODF standard and the W3C EmotionML recommendations.