RBJPRice Bowl Journals Podcast
References in periodicals archive ?
This process in RbJP is not covered by the logical uncertainty principle, since there is no revision of the prior probability.
Notice that RbCP is a "normalized" form of RbJP. In fact, according to Probability Theory we have
In order to perform some performance testing with the four retrieval models, RbJP, RbCP, RbLI, and RbGLI, we have three requirements:
It can be easily seen that the average precision increases consistently from RbJP to RbGLI in all three document collections, although the increase rate is variable.
Collection RbJP RbLI RbCP Cranfield 1400 24.3 27.6 (+12.0%) 31.8 (+13.3%) CACM 27.1 33.2 (+16.8%) 37.1 (+10.6%) NPL 22.4 29.8 (+24.8%) 38.1 (+21.9%) Collection RbGLI Cranfield 1400 36.2 (+12.1%) CACM 42.8 (+13.4%) NPL 42.1 (+9.5%) The results displayed in the Recall/Precision graphs in Figures 7, 8, and 9 show that the performance of RbGLI are slightly higher than those obtained by any other model, with RbJP at the lowest level of performance.
--any model inducing a probability transfer (like RbLI, RbCP, and RbGLI) performs better than any model that does not induce such transfer (RbJP);