Knowledge graphs & search engines

This Friday we'll have two talks followed by drinks.

Our industrial speaker is Panos Alexopoulos, head of Ontology at Textkernel B.V., in Amsterdam, Netherlands, leading a multidisciplinary team of data professionals (Linguists, Data Analysts, Data Engineers, NLP/ML Scientists) in developing, delivering and maintaining a large multilingual Knowledge Graph about the HR and Recruitment domain.

Our academic speaker is Thomas Kipf, PhD student at the University of Amsterdam supervised by Prof. Max Welling. His main area of interest is large-scale inference for structured data and semi-supervised learning. He is also interested in reasoning and multi-agent reinforcement learning.

Program:

16:00-16:30 Panos Alexopoulos

16:30-17:00 Thomas Kipf

17:00-18:00 Drinks and snacks

Details:

Panos Alexopoulos - Search Engines and Knowledge Graphs: It's complicated!

Modern search engines and natural language personal assistants use increasingly knowledge graphs to better understand and fulfil the information needs of their users. However, enhancing a search engine with domain knowledge, either engineered in-house or derived from external sources, is not a so straightforward task. In this talk I describe 3 important pitfalls that knowledge-based search engine designers often fall into and suggest ways to avoid them:
P1: Not knowing or not describing the semantics of your knowledge graph accurately enough.
P2: Not using the right type and amount of knowledge for the task at hand.
P3: Not harvesting your domain knowledge from the right data or in the right way.

Thomas Kipf - Deep learning on graphs with graph convolutional networks

Neural networks on graphs have gained renewed interest in the machine learning community. Recent results have shown that end-to-end trainable neural network models that operate directly on graphs can challenge well-established classical approaches, such as kernel-based methods or methods that rely on graph embeddings (e.g. DeepWalk). In this talk, I will motivate such an approach from an analogy to traditional convolutional neural networks and introduce our recent variant of graph convolutional networks (GCNs) that achieves promising results on a number of semi-supervised node classification tasks. I will further introduce two extensions to this basic framework, namely: graph auto-encoders and relational GCNs. While graph auto-encoders provide a novel way of approaching problems like link prediction and clustering, relational GCNs allow for efficient modeling of directed, relational graphs, such as knowledge bases (e.g. Freebase).