From document retrieval to question answering

is one of the most valuable goods in modern society. With the rise of computers, storing huge amounts of data has become efficient and inexpensive. Although we are now in a position where we have unprecedented amounts of information at our finger tips, the question arises how to access these large amounts of data to find the information one is interested in. The issue of developing methods and tools for finding automatically relevant information is addressed by the research area of information retrieval, and, over the last decades, sophisticated document retrieval systems have been developed. One particular branch of information retrieval is question answering. Question answering systems enable users to pose full natural language questions, as opposed to keyword-based queries, which are commonly used in document retrieval. In recent years, question answering has witnessed a renaissance, which is mainly due to the availability of large corpora. Current question answering systems depend strongly on document retrieval as a means for identifying documents that are likely to contain answer to a given question. This thesis investigates the usefulness of different standard and novel document retrieval approaches in the context of question answering. More specifically, it compares them with respect to their ability to identify documents containing a correct answer. In addition, we also investigate to what extent the quality of a particular document retrieval approach has an impact on the overall performance of a specific question answering system.