Knowledge graph based visual interpretation of web content

Web content nowadays can also be accessed through new generation Internet Connected TVs. However these products failed to change users' behavior for consuming online content. Users still prefer their personal computers instead of their TVs when they access Web content. Certainly, most of the online content is still designed to be presented with a personal computer or mobile devices. In order to overcome usability problem of Web content consumption on TVs, this thesis presents Videolization, a knowledge graph based visual interpretation system that automatically interprets visually given Turkish or English textual Web content by using Semantic Web based technologies. The system visualizes textual Web content by utilizing visual representations of extracted entities from the content. The generated visual interpretation of a given Web content could be automatically converted into a video by using Computer Graphics based technologies. Therefore the main focus of this study is entity linking, which is the most critical task in Content Curation process. Entity linking is the generation of assignments from knowledge graph entities to documents. In contrast to many successful applications for English, there is currently no publicly available entity linking system for Turkish. In order to visualize Turkish content, this thesis presents Thinker, a novel entity linking system for linking Turkish text content with entities defined in the Turkish dictionary or Turkish Wikipedia. The effectiveness of Videolization is validated empirically over opinion surveys and the effectiveness of Thinker is validated empirically over generated data sets. The experimental results show that Thinker greatly outperforms previous methods in terms of disambiguation performance.