With the arrival of new tools for information retrieval, extraction, organization and integration, the need has risen for these tools to filter relevance. Not the amount of information, but the access at the right time and in the right form determines the value. Moreover the increasing availability of new information-transmission technologies demands more personal information filtering. One such approach could work on information sources dynamically adaptable to changing application needs. As a consequence the interaction between application domains and information sources needs to be stressed. Natural Language Processing tools may be helpful in identifying relevant aspects of documents providing not only the identification of morphological or structural properties of the texts but also stressing what are salient passages in them. The identification of subsentences capturing relevant terms may be a first step to be acquainted with specific topics of a document. The availability of a system able to capture such terms, in different languages and application domains, could optimise users’ requirements.