Language Independent Ranked Retrieval with NeWT

In this paper, we present a novel approach to language independent, ranked document retrieval using our new self-index search engine, Newt. To our knowledge, this is the first experimental study of ranked self-indexing for multilingual Information Retrieval tasks. We evaluate the query effectiveness of our indexes using Japanese and English. We explore the impact that linguistic processing, stemming and stopping have on our character-aligned indexes, and present advantages and challenges discovered during our initial evaluation.