Empirical Term Weighting

Our system used an empirical method for estimating term weights directly from relevance judgements, avoiding various standard but potentially troublesome assumptions. It is common to assume, for example, that weights vary with term frequency ( ) and inverse document frequency ( ) in a particular way, e.g., , but the fact that there are so many variants of this formula in the literature suggests that there remains considerable uncertainty about these assumptions. Our method is a kind of regression method where labeled relevance judgements are fit as a linear combination of (transforms of) , , etc. Training methods not only improve performance, but also extend naturally to include additional factor, that is burstiness. The proposed histogram-based training method provides a simple way to model complicated interactions among factors such as and .