[MA] Text2Turtle (T2T) -The transformation of Wikipedia texts into a Knowledge Graph

Start: 01.12.2020

End: 29.06.2021

Type: Master thesis

Student: Christian Klose

Supervisor: Prof. Dr. Andreas Harth, Zhou Gui

Abstract: With the rise of voice assistants and semantic search comes a need for Knowledge Graphs as they are the fundamental building block enabling these technologies. This thesis deals with the automated Knowledge Graph Construction (KGC) from unstructured data. Predominantly, the focus is on Open Information Extraction (Open IE), an unsupervised learning approach that attempts to extract triples from text independent of their domain and, hence, it is the first step towards automated Knowledge Graph Construction. Previous work mainly applied Open IE Systems to English texts. In this thesis, the focus is on German texts. Due to the lack of German Open Information Extraction datasets, a dataset will be created. Two novel Open Information Extraction Systems for German will be introduced. A naive attempt to create a Knowledge Graph from the extracted triples is described. The performance of the systems is evaluated. Finally, the results of the evaluation as well as for the transformation process will be reported.