Supporting search engines with knowledge and context

Search engines leverage knowledge to improve information access. Such comes in different forms: unstructured (e.g., textual documents) and structured relationships between real-world objects topics). In order effectively knowledge, search should account for context, i.e., about the user query. this thesis, we aim support leveraging while accounting context. first part of study how make more accessible when engine proactively provides such as context enrich results. As a task, retrieve descriptions facts from text corpus. Next, automatically generate fact descriptions. And finally, contextualize facts, that is, find related query fact. second interactive gathering. We focus on conversational search, where interacts with gather over large repositories. multi-turn passage retrieval an instance resolution, add missing conversation history current turn. model resolution term classification task propose method address it. final professional writers news domain. create event-narratives by exploring corpus articles. dataset construction procedure relies existing articles simulate incomplete narratives relevant performance multiple rankers, lexical semantic, provide insights into characteristics task. Awarded : University Amsterdam, The Netherlands 5 February 2021. Supervised Maarten de Rijke. Available at https://hdl.handle.net/11245.1/78187b29-2403-4711-800a-0f92fcb9b15c.