Çizge madenciliği tekniklerini kullanarak haber ile ilgili metinlerden bilgi çıkarımı

The increase in data availability and the progress in natural language processing provide a basis for the development of new techniques for information extraction. Event detection from textual content using the text mining concepts is a well-researched field in the literature. However, graph embedding techniques in recent years provide an opportunity to represent textual contents in graphs because texts can be enriched with additional attributes in graphs, and the complex relationships within graphs can be modeled better. In this thesis work, graph-based representations of textual resources such as news are examined for pattern extraction, and a method is proposed for news prediction. As the first step, graph representation techniques are investigated. Afterward, frequent subgraph mining and sequential rule mining algorithms are applied. We consider that subgraphs contain the main story of the contents, and sequential rules indicate the subgraph patterns' temporal relationships. Finally, the sequential patterns are used for recommendation according to their similarity scores. In order to measure the similarity, various graph embedding techniques are also examined.