The article presents the usage of the semantic network for storing and retrieving information from unstructured sources. The authors describe the semantic network model based on labeled oriented graph, consider basic semantic network elements (concepts, relations and attributes) and basic relation types between elements (hyponymic and meronimic as well as class-instance relations), and define main single and group operations on the network. Besides the method of storing the concept instances and related wordforms and glosses for automated information retrieval from semi-structured sources is provided. The designed semantic network is decomposed into atomic concepts providing absence of stored information redundancy without necessity to apply the normalization procedure. The developed model was applied in several practical tasks, for example in real estate information parsing, the address information system input validation, and information extraction from spreadsheet data. The model was improved and extended based on experimental results.
Keywords: semantic network, graph model, concept, relation, attribute, instance, semi-structured data