The idea-oriented search is a new pattern for search engines whose databases are composed by text files. This pattern sets out to solve a usual problem for search engines: demanding that users foresee which words are contained in the desired document, keeping them from focusing on the information they are indeed seeking. To solve this problem, the linguistic bases for the development of a theoretical model that can add semantical data to the text files are laid down. In this model, the semantical information of a text is represented by what has been referred to as structure of concepts. The main goal of the structure of concepts is to give one single representation to sentences that express the same meaning. Thus, paraphrases present the same structure of concepts. In this dissertation, the first elements of this model are exposed in its semantical, syntactic and textual parts. Also present are the integration of these elements. A small case study is presented as well, with the intention of illustrating the development of an application for text files databases search engines in which this technology is used.