The paper deals with a computational resources that is being developed within the FunGramKB Project. Generally speaking, the full natural language processing is almost impossible to achieve and only two main tasks are performed where knowledge is applied to the input text: parsing, e.g. syntactic disambiguation and spell checking, and partial understanding, e.g. document classification, lexical ambiguity resolution, etc. This is mainly due to the fact that deep semantics for NLP is highly limited because it relies on limited sources of information such as Wordnet, GUM or SUMO whose weakness is due to their being lexically-driven. A system which is conceptually-driven and exploits a robust knowledge base and a powerful inference component, as advocated by Vossen (2003), can help to achieve a successful development of NLP systems. In order to pursue this goal, FunGramKB integrates the three types of common-sense knowledge – i.e. semantic (cognitive information about words), procedural (information about how events are performend in ordinary situations) and episodic (information about specific biographic situations or events) (see Tulving 1985) – with three corresponding knowledge schemata – i.e. 'meaning postulates' in the Ontology, 'cognitive macrostructures' in the Cognicon, and 'cases' in the Onomasticon.