Optimum query Seeks to find a query that maximizes similarity to elevant documents while minimizing similarity with non-relevant documents
Let
Cr denote set of relevant docs, Let
Cnr denote set of non-relevant docs.
qopt=argmaxp[sim(q,Cr)−sim(q,Crn)]qopt=∣Cr∣∑d∈Crdi−∣Crn∣∑d∈Crndi Compute the weight for term i using the log-odds ratio:
wi=log(qi(1−pi)pi(1−qi))The updated query vector is then given by:
qBRF=q0+[w1,w2,…,wn]