Paradigmatic and Syntagmatic Relations in Information Systems over Ontological Graphs

People make sense of a text by identifying semantic relations which connect the entities or concepts described by a text (cf. [2]). Therefore, in the search for smarter, more human-like, computer tools, we need to equip such tools with ability to identify and utilize semantic relations in processing the texts. In [4], we have dealt with the problem of mining real-estate listings. In this problem, a special attention should be focused on preprocessing steps transforming advertisements in the textual form into information systems. Information systems have been proposed by Z. Pawlak as knowledge representation systems (cf. [5]). Any information system can be represented in a tabular form, i.e., as a data table called information table. Columns of the table are labeled with attributes, rows are labeled with objects, and entries of the table are values of the information function. Many machine learning and computational intelligence methods and algorithms, especially those based on rough set theory, need to capture and represent data in a machine-friendly format. Therefore, they work with data stored in information tables. In the paper, we would like to show the main idea of incorporating paradig-matic and syntagmatic relations into mining data stored in information tables. In [6], F. de Saussure distinguished between syntagmatic and paradigmatic (as-sociative) relations. Paradigmatic relations hold between concepts belonging to the same grammatical category. Paradigmatic relations cover a wide variety of associations between words, including morphological and phonetic. Syntagmatic relations hold between two or more words co-present in a sequence. Combinations based on sequentiality are called syntagmas. The notion of a syntagma applies among others to group of words and to complex units of every size and kind, for example, phrases, sentences. In [3], ontologies, represented by means of graph structures (ontological graphs), were incorporated into information systems to deal with paradigmatic relations in data mining processes. In the ontological graph, each node represents one concept from a given ontology, whereas each edge represents a semantic relation between two concepts. There are a lot of paradigmatic relations defined in the literature. For example, WordNet [1] represents around a dozen paradig-matic relations between concepts, including: synonymy, antonymy, hyponymy, hyperonymy, meronymy, and holonymy. Information systems, in which attribute values are concepts from ontological graphs assigned to attributes, are called