Linguistic summarizaton on heterogeneous information networks

With the spread of information technologies, the volume and diversity of data is rapidly increasing. Therefore, the processes of obtaining useful information from data gain importance. By linguistic summarization, which is one of the descriptive methods, natural language-based summaries are produced from large volumes of data. Real world networks are rich in terms of data volume and diversity. Due to the nature of objects and relations having different characteristics, heterogeneous information networks have emerged that take into account the difference of node and connection types. In this thesis, firstly, real case studies were conducted in the bicycle industry with the summary forms proposed for social networks. Linguistic summaries have been generated to support the decision maker for supply chain networks to be designed for specific purposes. Then, linguistic summarization techniques proposed on homogeneous networks were extended to heterogeneous information networks. Linguistic summary structures have been developed on the concept of metapath in heterogeneous information networks and truth degrees have been calculated with the iteration-based method. This time, the effect of supply chain partnership relations on supply chain network performance is examined with an application in the automotive industry. The interpretability of the produced linguistic summaries was tested on different sectors and different problems in both applications.