Text documents similarity has been always an active research topic because of the numerous applications it has. And as the dominating keyword similarity search is not always sufficient for detecting similar content in text documents, especially when this content is expressed in different forms( using synonyms, pronouns, etc.), current research is moving towards more intelligent techniques in document similarity detection. We define a new notion of semantic similarity between documents based on simple RDF-facts similarity rather than keyword similarity and a step further from semantic similarity of concepts. A general framework for semantic similarity detection is also proposed to meet the needs of the various use cases of text documents similarity including plagiarism detection, suggesting similar news articles and research papers as well as a user driven document similarity search.