Using language models in question answering

Abstract

In this thesis, we describe a language model based approach to parts of a complete Question Answering (QA) system. It includes the processing of the natural language query as well as the retrieval of relevant documents, passages and sentences.\nThe results show that the language model based modules in our QA system perform equally well or even better than current state-of-the-art systems. Due to the heavy use of fast statistical algorithms the main advantage of our system is an efficiency gain compared to the slower deep analysis linguistic methods used in other approaches. A second benefit of using language models is the ability to train them for new languages.