Many recent approaches towards neural information retrieval mitigate their computational costs by using a multi-stage ranking pipeline.In the first stage, a number of potentially relevant candidates are retrieved using an efficient retrieval model such as BM25.Although BM25 has proven decent performance as a first-stage ranker, it tends to miss relevant passages.In this context we propose CoRT, a simple neural first-stage ranking model that leverages contextual representations from pretrained language models such as BERT to complement term-based ranking functions while causing no significant delay at query time.Using the MS MARCO dataset, we show that CoRT significantly increases the candidate recall by complementing BM25 with missing candidates.Consequently, we find subsequent re-rankers achieve superior results with less candidates.We further demonstrate that passage retrieval using CoRT can be realized with surprisingly low latencies.

Visualization of inter-document similarities is widely used for the exploration of document collections and interactive retrieval. However, similarity relationships between documents are multifaceted and measured distances by a given metric often do not match the perceived similarity of human beings. Furthermore, the user's notion of similarity can drastically change with the exploration objective or task at hand. Therefore, this research proposes to investigate online adjustments to the similarity model using feedback generated during exploration or exploratory search. In this course, rich visualizations and interactions will support users to give valuable feedback. Based on this, metric learning methodologies will be applied to adjust a similarity model in order to improve the exploration experience. At the same time, trained models are considered as valuable outcomes whose benefits for similarity-based tasks such as query-by-example retrieval or classification will be tested.

The great scientific advance due to Convolutional Neural Networks (ConvNets) for image recognition problems encouraged many researchers to apply ConvNets on video understanding tasks such as human action recognition. However, related state-of-the-art approaches differ in various aspects, making it difficult to compare their results. This work compares some of those approaches in a shared environment using a standard protocol. The results give indication about the effectiveness of fundamental ideas behind proposed approaches and other influencing factors. Based on these findings several adaptations of the parameters and methods have been implemented and tested. Human action recognition problems are commonly approached by facing two complementary aspects of vision. The first one relies on appearance of shown objects, scenes and human poses. It can be considered a regular image recognition task. The second utilizes optical flow estimations to exploit motion information. Since image recognition already is a well-researched area, it is the temporal aspect which is in need of further investigation. The studies were therefore focused on optical flow, which allowed to deeper investigate the less researched sub-discipline.