Frequentist and Bayesian Approach to Information Retrieval

摘要: We introduce the hypergeometric models KL, DLH and DLLH using DFR approach, we compare these to other relevant of IR. The are based on probability observing two probabilities: relative within-document term frequency entire collection frequency. Hypergeometric parameter-free Experiments show that have an excellent performance with small very large collections. provide their foundations from same IR space language modelling (LM). finally discuss difference between LM. Briefly, is a frequentist (Type I), or combinatorial whilst use Bayesian II) approach for mixing probabilities, being thus inherently parametric in its nature.