Recent advances in mobile devices and video streaming services have motivated large-scale media consumption both in fixed and mobile environments. In this context, the main objective of network operators and service providers is to improve the Quality of Experience (QoE) of the end-users by providing high quality services and mechanisms for seamless adaptation to the specific network conditions of each user. However, the initial design of the Internet as a best-effort network makes it inappropriate for high-volume and bandwidth-intensive applications, like video streaming, and the centralised architectures employed by the network operators lead to long, non-optimal communication paths between clients and servers, waste of network resources and increased delays. The improvement of the QoE of multimedia communication services require novel network architectures and cross-layer adaptation mechanisms. Specifically, at the network layer, Software Defined Networking (SDN) enables the virtualisation of the network functions so that the network operators implement their own rules and policies in software and deploy them in an abstracted and virtualised network infrastructure. At the application layer, the Dynamic Adaptive Streaming over HTTP (DASH) approach enables the seamless adaptation of the video client to the specific network conditions of each user. The understanding of how the network parameters affect the human perception is a key factor in optimizing the functions in the end-to-end delivery chain. The V-SDN project aims at developing a QoE-driven media delivery platform based on network virtualisation functionalities, which takes into account the service utility functions, network topology, link capacities and the specific QoE requirements of each application. The developed QoE-driven media delivery platform will be implemented and demonstrated with an innovative media delivery system employing a novel DASH client, a further key development of the project.

Deep neural networks (DNNs) have become a critical tool in natural language processing (NLP) for a wide variety of language technologies, from syntax to semantics to pragmatics. In particular, in the field of natural language inference (NLI), DNNs have become the de-facto model, providing significantly better results than previous paradigms. Their power lies in their ability to embed complex language ambiguities in high dimensional spaces coupled with non-linear compositional transformations learned to directly optimize task-specific objective functions. We propose to adapt Deep NLI techniques to the biomedical domain, specifically investigating question answering, information extraction and synthesis. The biomedical domain presents many key challenges and a critical impact that standard NLI challenges do not posses. First, while standard NLI data sets requires a system to model basic world knowledge (e.g., that ‘soccer’ is a ‘sport’), they do not presume a rich domain knowledge encoded in various and often heterogeneous resources such as scientific articles, textbooks and structured databases. Second, while standard NLI data sets presume that the answer/inference is encoded in a single utterance, the ability to reason and extract information from biomedical domains often requires information synthesis from multiple utterances, paragraphs, and even documents. Finally, whereas standard NLI is a broad challenge aimed at testing whether computers can make general inferences in language, biomedical texts are a grounded and impactful domain where progress in automated reasoning will directly impact the efficacy of researchers, physicians, publishers and policy makers.