Predict Ranking of Object Summaries with Hidden Markov Model

Ranking of object summaries proposed a keyword search paradigm which produces, as a query result, a ranked list of object summaries (OSs) in top-k and size-l; each OS summaries all data held in the relational database about a particular data subject (DS). This paper further investigates the volatility of the ranking position, and a robust, adaptive model is developed with probability terms basing on the hidden Markov model (HMM) approach. The parameters of HMM are trained by calculating the rank scores of each tuple in time series, and then this model is used to guide the ranking of OSs for further high accuracy depending on probability estimations. Preliminary experimental evaluation on Microsoft Northwind and DBLP Databases are presented, which proves that HMM has superior discriminative properties.