In this paper we describe a constraint analyser for contextual rules. Contextual rules constitute a rule-based formalism that allows rewriting of terminals and/or non terminal sequences taking in account their context. The formalism allows also to refer to portions of text by means of exclusion zones, that is, patterns that are only specified by a maximun length and a set of not allowed categories. The constraint approach reveals particulary useful for this type of rules, as decisions can be taken under the hypothesis of non existence of the excluded categories. If these categories are finally deduced, all other categories inferred upon the non existence of the former ones are automatically eliminated. The parser has been implemented using Constraint Handling Rules. Some results with a set of rules oriented to the segmentation of texts in propositions are shown.