Institute of Formal and Applied Linguistics

Question answering has always been an interesting and challenging task in computer science. In this talk, I present the result of our work, a state-of-the-art question answering system, and elaborate on its different components. I report its results using different approaches which we have taken for training, from simple logistic regression and multi-layer neural networks to more deep neural models including a convolutional neural network. To make our approach more clear, I initially have a touch on the structure of knowledge graphs and Integer Linear Programming approach both of which are crucial elements of our system.