Towards a Rule Based System for Automatic Simplification of Texts

The CogFLUX system is based on transformation rules used to reduce complexity of texts. The rules were compiled by Anna Decker (2003) based on studies of corpora of easy to read texts and normal texts. She has identified 25 general syntactic transformation rules for simplification. The rules can be grouped into two subsets of rules; 1) rules that remove or replace sub phrases and 2) rules that add new syntactical information to the text. An example of a rule from the first category is: np(det+ap+n) → np(n). This rule will replace any nominal phrase containing a determiner, an adjective phrase and a noun with a nominal phrase containing only the noun. CogFLUX implements the first subset of Decker's rules. The system consist of several modules, eg, GRANSKA tagger for part of speech tagging and Malt parser for syntactic analysis, and a module to replace abbreviations with its extended form.