TITLE: 
Understanding the Web using Large-Scale Knowledge Repositories

ABSTRACT:
Web pages are often annotated by their authors to facilitate automated processing of the text, under the assumption that the consuming machine is not capable of deeply reading natural language text. These annotations are usually quite shallow and offer limited semantics, for instance, there is no disambiguation and it is generally not possible to establish whether two pages refer to the same entity. Fortunately, many of these limitations can be alleviated thanks to the recent advances in machine reading and in construction of large-scale knowledge repositories, such as Freebase. Automatic annotation of texts with respect to these repositories will benefit from a consistent vocabulary of entities as well as knowledge of their properties. On the other hand, reliable identification of entities and their relations in texts will allow further expansion of the existing repositories with new knowledge. In this talk we will discuss the frontiers of research in knowledge discovery on the Web, as well as new functionalities that become possible due to deeper text understanding, including proactively fetching relevant information and entity-based services.