| Abstract [eng] |
Every day in websites, news portals the huge amount of information is often redundant. In a multitude of information it is very easy to get lost and not find the necessary information. Automatically analysing websites, news portals, published Lithuanian articles face with a problem that is inserted into the ontology recurring information. In semantically annotated texts, find the necessary information would be much easier and more convenient if the ontology would be given the opportunity to automatically identify similar ontology individuals and link them by similarity. In this master's degree work an ontology is analysed which was created by Informatics Systems Department during \"Syntactic-semantic analysis and search system for Lithuanian Internet, corpus and public sector applications „SemantikaLT“\" carried out by Lithuanian Economics growth actions program 3rd priority \"Information Society for Everyone\" realization means Num. VP2-3.1-IVPK-12-K \"Lithuanian Language in Information Society\". This ontology individuals of Person class which were automatically created by analysing article texts in Lithuanian language published in the Internet. Creating ontology Persons individuals this way, duplicated Persons were created too. The master goal is to make possible to automatically identify similar ontology individuals, to link them and thereby minimize information of ontology duplication. Master thesis objective algorithm was developed and written in SPARQL query that includes the ability to automatically identify similar ontology individuals and link them. This helps to reduce the duplication of ontology information. About specimen collected and analysed information the result is displayed on the ontology that searching every time do not need to analyse the ontology individual’s similarity. Created solution easily adoptable for this type of ontology. In ontology structure there is label_lemma (a generic form of the word), using label_lemma reduced ontology individuals repetitions, merging individuals by similarity for the purpose to provide concrete, structured information. |
