Opening the Knowledge Tombs - Web Based Text Mining as Approach for Re-evaluation of Machine Learning Rules

Growth of internet usage and content provides us with large amounts of free text information, which could be used to extend our data mining capabilities and to collect specialist knowledge from different reliable sources. In this paper we explore the possibility for a reuse of 'old' data mining results, which seemed to be well exploited at the time of their formation, but are now laying stored in so called knowledge tombs. By using the web based text mined knowledge we are going to verify knowledge, gathered in the knowledge tombs. We focused on re-evaluation of rules, coming from symbolic machine learning (ML) approaches, like decision trees, rough sets, association rules and ensemble approaches. The knowledge source for ML rule evaluation is the web based text mined knowledge, aimed to complement and sometimes replace the domain expert in the early stages.