Out of the Box Phrase Indexing

We present a method for optimizing inverted index based search engines with respect to phrase querying performance. Our approach adds carefully selected two-term phrases to an existing index. While competitive previous work is mainly based on the analysis of query logs, our approach comes out of the box and uses just the information already contained in the index. Even so, our method can compete with previous work in terms of querying performance and actually, it can get ahead of those for difficult queries. Moreover, our selection process gives performance guarantees for arbitrary queries. In a further step, we propose to use a phrase index as a substitute for the positional index of an in-memory search engine containing just short documents. We confirm all of our considerations by experiments on a high-performance mainmemory search engine. However, we believe that our approach can be applied to classical disk based systems as well.